Transcript · October 18, 2024 · 19.6K words

Overtime: Gen AI for Creators | Pulse Check: Midjourney, Runway, Minimax, Krea, Flux & More

Auto-generated captions, lightly cleaned. Expect the odd misheard word. Tap any timestamp to hear it on YouTube, or go back to the episode for the summary, receipts and chapters.

0:00:00Introductions and community

Play from 0:00:00 [Music] let me close out of some here so I don't I don't crash the stream I got presentation I think we are live welcome everybody we don't have our usual co-hosts Here Andrew and uh Max are both away teaching as usual but it happened to actually come on the podcast day um which is unfortunate this week but we have the co-hosts of one of the best Midjourney podcasts that you have to check out on YouTube Midjourney Fast Hours podcast with Rory and Drew what's up man how are you I'm so good I'm not I'm not used to somebody else doing the intro here

Play from 0:00:46 Chris yeah it's weird being the host isn't it and then being the co-host the dynamic the dynamic changes I love it I love tuning into you guys' podcast it's honestly the best one and um all of the examples that you guys together as well do so well on LinkedIn so you may be familiar with uh some of Rory's like minia Max image to video stuff the most recent stuff you know and Drew's doing the same just showcasing different workflows as soon as they come out they do a better job than these companies do of actually marketing themselves I told you that last time it's kind of funny um so we got tons to watch this uh this week it's been a bit of a watch party we've kind of changed this format we used to do it very news-based now it's kind of like a watch party half and then

Play from 0:01:32 news in the first half and uh people drop in and out so Dave might come and watch his own music video later which is one of the videos we're gonna watch and uh yeah we take questions from all over hey we got Jeremy an absolute AI legend in the building watching he just did a really cool video that I might load up in the background when we're watching other things uh where he put his face into like a dynamically moving video uh really really really cool we' got Glenn in New York City baby got Eric in the house I don't know you guys are from New York what what part of New York are you from drop in the comments yeah I'm you know I like to be these two are like absolute goats these guys have been following us from

Play from 0:02:18 the start years now Glenn's been like tuning into the audio sessions and Eric constantly asking me when we're going live ask him for the link he's an absolute Legends we got lucer in the house as well another AI let us know where you guys are from I forgot to say that in the very beginning I I usually shout out where you guys are from and I'm going to shout offen because I'm a bit sick this week as well you might hear it from my voice so you guys introduce yourselves and uh tell us more about uh what you're doing in ai go for it oh legit we got some Upper West Side Manhattan all right hey we've got leads in the building South Africa in the house Saudi Arabia in the building upper West Side it turns out in New York this

Play from 0:03:03 is you know what's crazy I was talking about this the other day someone it's just like that I never expected something so with such like Global I know translation right it's it's so so easy to translate globally right it's not something that we just like are stuck with here in the US and it's like this little bubble same thing with the UK it's like everywhere which is so so cool you know what that goes on the sentiment of um I spoke to Mato who's like heading up I at Google and he said that content is liquid it's fluid everything is becoming fluid in natural language it's going to change to language you know we talked about lip dub a few weeks ago actually I've got an example of lip dub um because it was one of the pieces uh actually let's just watch this as the first video anyone

0:03:24Advancements in lip syncing

Play from 0:03:48 that's not appreciative of this Rory and Drew will know when we first got lip sync it was terrible it was no teeth barely any movement in the lips it was no emotional on the face it was barely legible right then it graduated to ah you can see some nuances it kind of matches but still no emotion in the face yet and now it's like the full package of everything combined where uploading audio to an image actually makes it feel like you know the real thing so this clip is from uh Dave's Battalion which was a f minute AI film that he made and he redubbed this with lip duub it makes me think more we even fighting for what Freedom were we fighting

Play from 0:04:35 for this ain't Freedom what world did I bring him into I was part of the barrage balloon Battalion I remember storming the beach that day hundreds of [Music] soldiers it's crazy how far it's coming there's there's also such subtle Nuance in that shot where it's just subtly arcing and rotating to the left around like you know around the face and also you know the thing that I always hated about the lip dub or not lip dub in general but like the lip sync was that like there was so many unnecessary head motions right it was like compensate it's like yeah we yeah totally and then

Play from 0:05:21 it remember when it didn't blink and then it blinked too much as well because it was like the overcompensation for it's not blinking enough was to constantly blink and then it just looked unnatural so the Nuance is they're really when you really get into it there's a lot of them but when you really you know crack it you're like wow it really makes such a difference to see that type of like advancement in that specific area of gen right that's the interesting thing too like there's I've been reached out to by a couple production companies right they're just like they're the one question they ask more than anything is is is the limp lip sync ready yet I mean I still think it's got a little bit of a way to go considering you know exactly how a face moves and how it should look it's very uncanny valley a lot of stuff it's uh

Play from 0:06:08 that's definitely the feature I feel like is most I don't want to say Most Wanted but most like anticipated of it being really like translatable which this is close this is as close as it can get I sent someone the other day I was like Dave got it as close as it could possibly get right now I think so right like it's like teetering on Commercial that's the thing that's like it's really really really close but yeah it still takes a lot of you know post- production and stuff like Luc is pointing out here and absolutely you know color grading everything else in the post production to get really polished finished stuff that's of that commercial value but yeah like I wasn't even expecting to talk about that that was actually from uh last week's but I just wanted to you know seing a few new faces in here new this week so I thought like just to catch up on when we are lip

Play from 0:06:55 dub it's just straight in the Deep End right like that is exactly what people want especially for music videos you can get Talking Heads you can get personalities avatars it goes with everything um the thing that I loved about that clip is that he's not centralized remember when he had to be really up close centralized and looking towards the camera he's slightly tilted head and everything's still in check and that really matters as well like that you don't have to be facing the camera because it's a dead giveaway that it's AI done then you know 100% 100% motion the camera right like you have even with the motion camera stays kind of locked and fixed and masked on that exactly compressor and when in cinematic does the character look straight into the camera you know very rarely unless it's like you know a flashback or like a

Play from 0:07:40 moment outside of the third wall or whatever is very unlikely to happen so it has to be the other way in order for it to be taken seriously again hey where the hell is Les Soo that's amazing talking that's a small country in Africa that's incredible swiland just everywhere like so cool Paris Slovenia Saudi Arabia it's like you said it's so liquid cont it is wild it's so cool to see and welcome anyone new um these guys like I said Absolut Legends for stepping in last minute usually I'd have two other co-hosts here and Dave might join in who's been regularly dropping in just someone that we started collaborating with a lot recently Dave Clark just incredible filmmaker and we've got loads

Play from 0:08:28 and loads of stuff to show so I'm not going to blab on too much I just got back from Adobe Max that's the big news event of the week really they're really so much you know meta movie gen is kind of going to be happening next year it seems like in January so really Adobe stuff came out as soon as the event went live all these stuff that are popping off like um Adobe fireflies now video tool and it's in beta and uh you've also got generative extend in Premier Pro so that adds like clips that didn't even exist to pre-existing Clips or generative Clips as well so just multiple multiple features you guys use Adobe don't you uh the suite of tools yeah so it's interesting isn't it now

0:08:30Adobe Max and new AI tools

Play from 0:09:13 we're getting them in the mainstream creative software that we've always expected they were coming but at the scale where it's everything's everything's starting to come into one place you're starting to kind of see that you know rather than there be 12 tools there's one you know and I think that's that's probably the race that's happening it's interesting I I I love that adobe piece with the compositing uh aspect to it where it changes the color I was just like damn you know like that changes so many lives right like yeah yeah I'm not a compositing expert by any means but you just think about doing those images one at a time which I've I've done and I'm sure several people have done it could take a lot of time to make it look crisp tracking hair and stuff as it swooshes like all that like

Play from 0:09:59 the in stuff was big from meta's announcement wasn't it it's like just highlight a character and then it will track them and then you can remove them and do all this other stuff like that type of like uh action or feature within Adobe will save creators hours and keep them in the platform I'm seeing Korea actually trying to do this as well yeah that was like under the clo of Darkness like it just like I announce it it was crazy and the other thing they have in there too now which I haven't even seen announcement maybe I'm not paying attention was they have an existing image editor so you can just upload an image in there and edit yeah like they had image image and upscale didn't they like it seems like there's two camps there's kind of the will have any API

Play from 0:10:44 like free pick and other things like that and then there's the no we do sort of like you know blackbox stuff but it's Unique to us and so that's the use case and and I'm seeing you know even the open source stuff like com youi that hosts all of the open source ones and other things like that so it's a real battle of the like business models isn't it when it comes to AI like subscription tokens what's the way to go like how to get the users how to scale not easy stuff minia Max though God the most impressed I've been all year by any AI tool seriously I'm just blown away by the fact that this is by such a small team and it's so good it's top four at

Play from 0:11:29 worst you know and to say that is like just not they went from zero to top four in the text of video and now the image to video is maybe putting it top two it's hard to know because cing seems like very good at movements of people natural movements Runway seems really good at VFX Minx seems like a blend of the two you know almost like a hybrid of the two models and then lum's always done its first features you know first frame last frame they started that now runway's got that like what's your feeling on all of this it's it's to me like the just like you said there's different uses for different tools like a Mini Max I've been super impressed with with some of the natural motion like some stuff you just can't get in

Play from 0:12:17 Runway like I love Runway text a video way easier to get the motion that you're looking for um you know it's probably my go-to tool it's also so damn fast it's hard not to use it with the you know Gen 3 Turbo um but Mini Max some of the stuff has been hilarious but also you know really good like I yeah like surprisingly good right yes it did like this fight video the other day it's like totally warped and ridiculous but it's it's impressive nonetheless no other tool can get that level of generation speed or that level of movement and like dynamic dynamic with the camera Motion Plus the character movement plus the action you know it's that's the only one that can do it it feels like it's trained differently I don't know that's just my my own weird opinion about it

Play from 0:13:02 but yeah I like kind of be quite honest kind of put Luma on the back burner like I don't I don't have a use for it with the key frames now unless it's just something that I can't get on Runway because right what I don't like about it number one I've said this on our podcast it's like can we just get a reroll button like is that so hard yeah to have to constantly load in the pictures again and start the whole problem so frustrating right like I that doesn't seem that hard that's so funny because we talked about friction a couple of weeks ago you know like you don't have to be the best as long as you're the most frictionless this is like some of the photo booth dream Booth apps that have appeared where prompt engineering was the gateway then they're

Play from 0:13:47 doing incredibly well you know like on a subscription model too like 90s AI yearbook or Harry Potter this and they're pre-loaded in prompts probably built on stable diffusion or something like that with luras and you know uploaded mini databases and triggering from that but they're charging like $2.99 every generation it's insane and people are paying for it in the thousands it's so many mobile apps there's so many exactly I got I get an ad for it every time I open my phone some new one oh you yeah you right it's like this one I saw was just animate your animate your photos right like animate your old photos it's like they're just using the API from one of these tools and it's probably again like you said pre-loaded prompts and like put the picture of you and your family in there and then it's like animates it

Play from 0:14:33 it's like okay you know they charging for that yeah right let's go let's get into some of these videos I'm gonna showcase uh two that you've put together recently Rory I think this one um let me just that's not the right one this one uh last frame key frame [Music]

0:14:39Showcasing creative workflows

Play from 0:15:18 so cool these are just music cool I love that I love that one with the egg sandwich love that one that one's crazy I was surprised how that rendered um so funny because we talk about like you sent me remember the one where you were like oh the donkey One's next level in WhatsApp and I was totally in agreement when I watched the series of images that you sent for that one that clip I was like my God the donkey the way that it nestled the head against the daughter and they kind of the ears moved all perfectly it was all in sync you know like every single Nuance was like Checkmate for that one and like it's fascinating that we can see these now and go whoa that one still blows my mind you know I wonder when that will stop I

Play from 0:16:05 wonder when we'll just be like okay this is just yeah it's it's got to be like you know this one the one because everyone does the Aging right like that's always been like a Comon like expose or like a way to show the tools but this one I was like that one was a real that one took a little bit more post-production from like a like getting the two key frames right than actually having it just be like I generated this image I generated this image because right with like Runway right I think there's with the key frames I don't know if anyone else has noticed this um like you you have to have some variance in the images but they can't be too strong because if it's too strong it doesn't necessarily work that well like it doesn't connect like two disparate images together very well that's just my experience so it was like

Play from 0:16:52 with that person right I had that guy and I'm like I want to age him so I had to remix it in Midjourney to get the the Aging look right so it just like pretty much just changed the age and the remix prompt and it makes him look older but then it was all right I have to create some motion or else it's just going to be him like standing there static aging so it's like I had to get go take the background scale it down you know generative fill around that to fill in the padding cut the guy out put him like forward in the image and then uh you know really like blend and kind of you know feather around him to make him look like he's integrated into the scene so it would give that effect of him like moving forward so it's like you know there's a little bit more like Consciousness into the key frames I think in Runway that you need to kind of be aware of the egg the egg stand that

Play from 0:17:38 one just totally like one hey hey Rory for for the people that are still at the beginning stages of video curious if you have like a top three top five sort of like key tips right to like sort of FastTrack that growth or flatten the learning curve a little bit yeah there's always there's always like the I keep going back to this when you're creating your base images I like to have like a little bit of motion baked in to it already cuz it just gives you a leg up and kind of like starts halfway so if someone's walking instead of having straight up so you mean like keyw is in the prompt as in like in motion or some sort of dynamic movement that's brilliant walking towards us motion blur that's um you know driving to the left

0:17:39Prompting and production tips

Play from 0:18:25 like cars I say this one all the time like use spinning wheels in the prompt so wheels are already spinning cuz if it's just static like Wheels like in freeze motion the car tends to slide in the video Generations instead of the wheel spinning so if you get that wheel fascinating I wonder if it draws a harder line around it you know when it is like very still in the image I wonder if the outline you know in the sort of like that's fascinating to think about never considered that what a brilliant tip because there's you know when I was doing I did a video a while back about the these people dancing right and uh you know the dancing it was harder when the people were kind of like standing with their arms down and maybe in a normal position but when they had like their arms like this like they were already in a position motion yeah up the motion and went with it like right in

Play from 0:19:11 the middle of the movement is where you want them to be that's brilliant which is that's probably the biggest piece of it I I always go back to the base images unless you're doing text to video which I think requires a totally different prompting set like text to video you have more variables that you can control right like you have the camera movement you you have the the object or character movement you have lighting you have the composition you have you know the like what kind of style you're trying to do it in so there's a lot more variables to control just straight text to video imageability on the art Direction and everything else like that like knowing what's actually gonna be there and stuff yeah totally so it's all about sacrifices and it's like quality versus control these are the sort of things that I constantly see as the dynamic

Play from 0:19:56 battle going on amongst these different like text to video image to video even but different models competing you know like Dary 3 seem to get great prompt coherence but not overall quality like Midjourney and Midjourney is raw quality but less prompt coherence you know so it's just fascinating to see like how again like and then shaping your business model around that fact you know like that's different customer that wants that because you know someone that wants scale and quality and someone that wants I use udio because of its raw quality but it is a lot harder to control than sunno which is much easier for creating music you can make two minute beats with one prompt and bang you're ready to go export it and then send it to your mate in udio you're doing 30 second things it's very much

Play from 0:20:42 like knowing music composition and raw raw power but less control and you have to harness the Beast you know it's crazy I still have to get on udio I've been I've been sleeping on it so you know it's just so it is easy cuz when I'm trying to make this stuff for like a video real quick I just like know what I'm thinking in my head yeah that's what I mean but that's because you know sound's not your background soo's made for you that's the way like udio was made in the same way that like no doubt like Midjourney is made for a more artistic director than say chat BT you know version of do 3 like you know like that interaction where it sort of rewrites your prompts calls upon an image model Services you the model that's the like don't have a Graphic Design background you know but want a quick image for social or what have you

Play from 0:21:29 yeah I believe sunno just released if I'm correct maybe I saw this wrong this morning but I think they just uh released like either image to song or video to song where it like fits the so interesting multimodal is the name of the year like everything is going multimodal right at the Adobe Max thing it was uh sound effects to sound effects so the guy was like screaming into microphone on stage and then transform that into a growl of a monster what so cool yeah actually I think I've got that clip loaded up uh let's watch that one all right so the moment I hits this little red button please hush hush

0:22:10Multimodal AI sound design

Play from 0:22:15 everyone cuz we're going to be recording all right here we go all right so we got the growls down now let's see what we get back I'm going to type in Monster growl and I'm going to hit generate and it's going to be using the precise timing Dynamics energy of my voice to drive the generation and so we'll give it a couple more seconds to do it magic all

Play from 0:23:01 right you know the world growls there's a lot to take in there those are good grows take in there oh there we go we got it and of course we get um let's crank up the volume I want to see what we got all right are we ready everyone yes let's see what we get wow whoa how about the guy having to growl in front of I that exact same thing I was like imagine the tech themo where it just doesn't work because it's an alpha

Play from 0:23:48 and you're on a stage in front of that many people and it just freezes and you're just like I'm in an aring to make it wait and oh that would be so nerve said he was going to growl on the next presentation he get a live demo of AI tools growling that is some that is some you know coones right there if that doesn't work and it just comes out like a little Meep like you know instead of like a Big Growl it's like oh that lost all the power that we were trying to go with right there it's funny because did you ever see 11 Labs video to sound effects where you could upload like a 10-sec clip and it would try to automatically add sound effects I never found that one work quite well but I felt like sunno when you could do like hum to audio or

Play from 0:24:33 hum to music that was the like you know thing to sound effects why not having a sound effect right well man like they're always doing crazy stuff to find the right sound and then they're kind of like exactly exact they call it Foley sounds and like you can you can go and just Bas you know like horseshoes is like coconuts and you just clack the Coconuts together like this the halves of the coconut and it makes it horse noise like click clu click clck click it's amazing you find all these ambient things that like make the same noise as the thing that you want to make a noise for or you run it through a sound effects like machine afterwards you sort of raw digitally record the thing like a train station and then pitch it way down and you get like an ambient noise or you get like something else it's so fun oh

Play from 0:25:20 man I'm so excited by that honestly like now it's the full Suite of tools you know udio can provide the background music now we can Prov ride our own sound effects we can be all of our own characters we can voice them all Dave did all of that you know voiced every character in his thing it's insane I I love the Gen generative extend piece that sometimes is so nice for even just like half a second you don't even need like you know like a full minute sometimes just to like fit with the rest of the composition you need like that little extra quarter second half second you know to make it work which I think is let's hope it doesn't drop off dramatically that's my worry about that you know is that it just grades really quickly after the video you know extend start so you don't get but I suppose you could just iterate and iterate until you

Play from 0:26:06 get one that works you know as long as it's not too costly these are things I want to know how much is meta movie genen gonna be how much is Sora gonna be how much is you know Midjourney's teased a video tour like there's going to be loads of them available but are we all going to pay $30 a month $50 a month subscription cost credits like it's going to become a lot like uh yeah spoil alert Roy will Ro we will because we've got a professional use case for it you know like I think that's the thing that's the key is like how do you monetize this stuff if you can't directly sell it or any copyrights to it are you doing creative services are you you know doing what you guys do amazingly digital downloads and like courses and workshops and podcasts and

Play from 0:26:51 what else are you doing to monetize this space that we're all in it's it's it's it's a tough one isn't it I wonder if Adobe is going to bring in anything I mean I wonder if they're just all working on proprietary stuff they're going to bring in some apis for some things did they mention any of that I didn't listen so it's interesting because that's like the Korea play is perfect like to me it's perfect you like Korea and you like the real-time editor and now you have access to everything and I you know I was looking in there the other day you can do you can upload lauras you can go and use flux within there you know in in terms of like it's amazing building your own things out it's crazy I just didn't even realize that that happened and I went in there one day for you know I forget what I was trying to do and I was like where did all this stuff come from like when did this happen they went from a few of the Rus

0:26:56Industry trends and future

Play from 0:27:36 boat tools and a few cool apis to everything I saw that and I was just like what in Earth and and it feels like that's the next move isn't it is where's the dashboard we're all going to be in because everyone who's got an API available will Port through to everywhere else and we've seen Acquisitions happened like Leonardo got pull by canvas and it's directly in there Runway you can use through canva through the AP you know like all of these people are going to be vying for the attention and these models like if they don't have our attention they better be the best because if you're using them somewhere else you're not going to choose it you know like that's the thing it's really tough tough position for these guys I mean with video it's just I don't know like I hate to use the word like race to the bottom because that's not the right term but it's just like you almost have to one up

Play from 0:28:22 every single time and it's it's such a constant thing right now and obviously everybody in this room probably in a little bit of a bubble in the sense that we're in these tools more than the average population by far I just you know like that to me is just like how do people like us even find the time to experiment with these let alone you know I know started I mean it's just it's impossible and you can't I know can't but I'm constantly thinking about that in terms of all these tools the consolidation that's going to have to happen um I do think like two like organization is a a big area that could be solved across the board you know just in terms of like the flow and keeping things very efficient and easy to find

Play from 0:29:09 and all those components from a ux standpoint definitely I'm just fascinated man things are moving so so quick I'm obviously also thinking about this from Midjourney's perspective right they're probably just gonna go so far the other direction um they can't play that game of trying to keep up with every little change they don't have have the power they don't have the financial backing although they are bringing in a lot of money ARR wise so maybe that's a little bit up to how they they hire but um yeah you know just everything from the the Midjourney side of things just seems like they're going to go in a completely other direction which obviously just piques your curiosity um but I did hear on office hours yesterday they're talking about maybe V7 makes it to this year promise that was happening they

Play from 0:29:57 even said they were training 7.1 didn't they and that the 3D space as well that they're looking into and you know what Dave said ultimately this all boils down to World building ultimately that's what it is you know like all this Gan Splat all this 3D all this like this is going to be like you're going to be able to walk into a picture soon that's what's going to be like totally possible right I think they're going to work on Hardware too so that you can actually go and do it I think that's they never given any sort of like indication as to what that Hardware would be but what else could it be I mean you know I would I would love to have some I mean is it is it AR is it VR is it something cool we don't even know

Play from 0:30:44 about well I said this to Andrew because Andrew said look all of this hyp because he's an OG right he was the head of social media at Myspace so he saw the rise of social media he saw the rise of virtual reality AR XR all of the acronyms he said that the key thing that was missing from the metaverse driving it all was AI so this is going to power all of this stuff you know like have you guys seen the examples of like um uh Mario that's overlaid with Midjourney uh no sorry with Runway so it looks like you know an overlay of Graphics you can sort of take a 10-second clip of a game of Mario in the Nintendo 64bit style upload that to Midjourney and get live overlay in any style they've done it with Grand Theft

Play from 0:31:29 Auto as well they made it look photo realistic so it's just some playing Grand Theft Auto and then then it you know there ching and then all of a sudden it's a totally different skin and then Ching it's a totally different skin like that feels like what the metaverse was lacking you know like that level of immersion interactivity and like it's going to probably power all of these to accelerate again arvr metaverse it'll all come back with a bigger bang because of the Gen uh Evolution I suppose it's wild to think about it is wild to think about yeah like I'm so glad that also a fascinating form of Foley yes thank you it's honestly if you know anything about this sort of Sound Engineering sound

Play from 0:32:15 design this is the geeky stuff to get excited about sound effects creating your own sound effects from your own noises into your microphone for years I've been wanting to do some sort of transformative sound work like that just hasn't been possible here so really crazy times that's that's a good one like that's one that you probably don't you know you think of imagees video text right like that's kind of like the things you immediately default too but something like this it's so it's so usable that so usable for so many things um want experts to dip their toes in for the first time too get them excited about making sound effects that's a totally cool thing especially stock sounds it's really hard to find something you like it takes forever you know that's never never find something totally fits right it's just that's my

Play from 0:33:00 experience at least with it but I like that was it how long was this how was it like seven hours worth of a presentation from Adobe how long was the actual event oh well it's the three-day thing okay I only went for day one and uh what what it was was it was all kicking off in Miami and similarly like it seems that these guys do this now they have kind of international watch stations so they' hired out a big Cinema in London which must cost them tens of thousands and there was like a 100 creators met up for a private screening of the Adobe Max where we talked to Adobe staff and evangelists and people like that and suggest workflows and stuff like that so it was a bit more than just like watching the same thing everyone else was watching online we had some interaction with the company themselves

Play from 0:33:47 but yeah man over 150 gen announcements over three days across all the tools illustrator Premier Pro all of it every single one has got now geni infused as well as nearly everything they said had the word AI in it you know like it was like nothing was traditional creative workflows anymore it was now here's what we can do with geni it's crazy almost AI is almost at the point of it's completely desensitized you know what I mean it's just like mainstream isn't it it's totally not the like it's being almost gonna go you know everything Andrew say is is so good at these things like um everything goes in waves you know we we're never in the Middle with culture we're never in a balance or an

Play from 0:34:34 equilibrium it's either twoo extreme or one extreme the other so what will happen is everyone will get sick of AI and then it'll sort of go down a little bit in hype and then we'll go back to the middle and then it'll go back the other way again you know and that's the sort of cyclical nature of these sort of like explosion events you know like we're at the tip of the iceberg right now but then the middle has got carved out a little bit and then it'll go really big again and that that I can just see it happening yeah it has because we mentioned that Drew and I mentioned this on the other you know the other day on a podcast it's like there's so much development in the video tools right like it's happening so quick and so many new features and I don't know if that's because it was behind the image tools and there just a lot to catch up with and maybe you know it was just like the next evolution of it but it doesn't seem like much new has been happening on the

0:34:57Leveraging AI for creative ads

Play from 0:35:21 image side I think that's probably where can step in because they have so many you know really insane features to begin with and so much granular level of control it's kind of like you know when you open like the tools feature in ad in Photoshop you could be you know to try and learn that stuff that's like months you know I remember going through it and you could do it you could get to the end result one million different ways it's not just one tool can do one thing you can use a combination of different workflows to get there I would love to have like an AI do they have that like I would love to have like an AI assistant within Photoshop sometimes to like kind of oh like a clippy like like a Microsoft clippy but not an annoying one that would be amazing so I could like incredible have you seen have you seen

Play from 0:36:07 no have you seen notion they launched something like that where they've got that it's a little it's a little guy that looks like clippy a little bit wow it's really hard to know when it's not going to annoy you I think that's the reason they're not more prolific because it really bugs some people you know and it's just constantly there you don't know how to switch it off but then it's really useful in some use cases you know to have like I think LinkedIn wants one of these it's a hard platform to navigate in the sales Navigator and stuff like it could probably assist you as to how you should build your ideal Prospect lists and all of that stuff you know it's interesting like fascinating God linkedin's such a show I mean they really do need some more Forward Thinking people to execute on that I mean just the way that they've

Play from 0:36:53 handled their platform expansion over the last four or five years when it's just gone on this meteoric rise and the lack of new things you've been able to do um the ux right like all these little beta features they test most of them don't turn into anything and then they'll roll something out that isn't a test and then they'll just take it away without saying a word about it yeah remember remember the pin comments come on oh my God yeah how can you get something so basic so wrong and now like we had LinkedIn stories they disappeared obviously in Instagram clone now they've got YouTube videos which they're really pushing but it seems to be that it's not really algorithmically you know making

Play from 0:37:39 sense like it's certainly not the most interesting videos on that feed for me on the mobile experience compared to YouTube or Tik Tok or anywhere else like this not even close right so like they need to really know about what they're trying to do like I wish they would segregate it more and have like hashtags and stuff to bring back content that's more relevant to one one another you know but difficult times it's it's interesting with the video stuff because it obviously if you if you do something on the feed that you break a little bit of the the monotony like it does reward you on LinkedIn like if you have something on video that is different than the Tik Tok style talking that's really green screened whatever like the stuff that I've been doing I've been getting rewarded for it because I I I

Play from 0:38:27 think it's different though but I think it's just everyone's here's a talking head of me it's like when everyone was doing selfies and that was like the the guru everyone was like you got to use selfies they engage it you know this better it was like all right I'm not gonna do selfies because I don't want to do selfies it's weird and everyone does selfies so yeah let's just put up a picture of like a you know flaming neon skull and see if everyone likes that or or what about what about this is your reminder to take your PTO what about that kind of Po oh my God or or um you know just oh God they're just so many bad ones I said that I was like you nailed this format like um like let's watch another one [Music]

Play from 0:39:20 [Music] donkey my mind it still BLS my mind first gen first first generation not not even a not even a reroll on that one I was like get out of here with this that was when I got huged with a front on Mini in on that donkey first gen that is and that's another thing Andrew asked he said as well as the cost how many times how many times you have to generate to get something usable because if the credits are the way you know and it takes 10 generations to get the one you want that's 10 times the cost you have to factor in even if it's cheap one

Play from 0:40:06 $1 generation or whatever it's $10 you've wasted so then uh fascinating right like that is unbelievable that it got it first go though because that was their Rory test things as soon as they come out so this is like probably a day or maybe days into that being a feature that was available through this model so that means you know super early without user testing human feedback and it's just nailed it like that's wild I was I was shocked number one on that one specifically but what I really like about Minimax is because it still feels really raw is that if the prompt is off like the video is so far off like that's just like all right adjust that a lot which is like good for you know when

Play from 0:40:52 you're testing something me because Runway sometimes you can get stuff pretty close without you even like having a good text prompt in there and it works but it's not what you want you'll probably roll it a bunch of times like Minx with I had some prompts and it was just like uh garbage and I was like okay that's not GNA work I have to like readjust how I'm doing things so it's nice to get that feedback loop every once in a while with the tools and they're newely that you just know that again I have to change up what I'm doing consistently to get there but I think everything goes back to that first tip you made I just want to H back to that if you've got the motion in the image Runway without a will probably give you the right motion because of the way you're in the middle of the movement and then everything is a feed forward from there you know everything's going to get more destructive if you don't start the

Play from 0:41:39 LA the first key frame correctly which is such a great tip I can't believe I haven't really tried that more I'm actually shocked Haven you know what this could be very interesting can we dig into that just a second Rory like how how do you prompt for those particular scenarios when you're trying to add motion to a static image right because this this is you know prompting is a is a language of its own and you do need to have sort of like this vocabulary or or token repository in mind or just things that you've seen work right so when you're trying to develop or create an image that you are thinking about using for video are there some certain tokens or phrases that seem to work well for you that translate from from those platforms and

Play from 0:42:25 you do them in a certain order I'm going to ask as well Rory like do you hierarchical matters right doesn't it motion like typically motion blur because it's like a general holding term like that'll work really well and it doesn't have to be you know you can use subtle motion blur like extreme motion blur like that'll I'm typically putting that first like that's you know we always talk about the hierarchical piece of it is it might not be like realistic in this sense but I tend to believe after doing this a million times that whatever you have in the beginning of your prompt gets weighted heavier than the end so like put motion blur at the beginning like that's the initial token that it goes to find and that's what it's going to do right that's just how my brain um so like motion blur number one dynamic dynamic motion I like

Play from 0:43:11 depending on what they're doing like if it's if it's you know dancing like I'm trying to add like a different a different adjective to describe instead of just dancing like it's like lavishly dancing or enthusiastically or jubilantly or facial expression same thing like you're adding that extra adjective can kind of like juice it a little bit but depending on the shots that I'm envisioning too because sometimes you know Midjourney throws you curveballs right and like a lot of times the curveballs are awesome and I'm like I want to use that but that wasn't what I was going for um because everyone I use a little bit of chaos on on these generations for the video oh that's a really nice tip as well yeah because dash dash C and then a random value between zero and 100 but I think like don't overdo it on the right I keep it

Play from 0:43:58 pretty low like six to 10 like just to see how it affects it yeah because you get some good stuff from that and you get some really interesting unique stuff and like you know then when I'm thinking about it it's like oh am I going to get this from am I trying to zoom in am I trying to zoom out um you know zooming in a little bit easier in the tools because the the pixels are already generated when you're zooming out you have to create right yeah point the person in the maybe it's me right here and I'm like this but that's I want it to be zooming in on me doing that so then I'll you know back it out in editor so it's like much to consider isn't that it's fascinating it's cool because you can do a million different things right like the key frames now to me this is where it gets really interesting because the

Play from 0:44:44 key frames you have to think about that very like you have to consider a lot of things when it goes from IM like image to image right like and it's filling in the middle space so like I did this that that in that video the one with the um the one with the Jet right yeah like I couldn't get a jet for whatever reason to be at the angle that I wanted it to be because I wanted to show and turning so I generated it flat then I took it into Photoshop and I went GE SM generated around then I shrunk that down and put it down in the corner and generated round of it so I know because it's like it has to get smaller to go away into the distance right if I'm directing the motion of where I want it to go like that's kind of how you have to think about it now because if you do the same size like the same scale like

Play from 0:45:30 the motion's going to look weird you have to like Dynamic motion which is that's funny because I saw that in the Runway video actually they uh photoshopped like a giant hand and then composited it into the back of a street to make it look like an evil Monster Hand you know and that's really smart like because that's the manual control that we're lacking right now you know so you're doing a great work around in tools that you know in order to achieve the like angle and movement that you want it to be at would you then use that as the star reference and then generate again or would you just upload that as the last key frame I would use what I did in Photoshop so it's consistent oh really so it's actually like altered image is your last key frame each time in phot wow that's another great tool so the other way you can do it like that that living room which is I think this is probably a good tip for everyone

Play from 0:46:15 that's using Midjourney especially the editor feature like yeah when you edit right when you go to the editor feature you paint around stuff right typically I don't get like it doesn't add artifacts to me really well like it doesn't add things well according to my text proms so what I'll do is I'll generate like I knew I wanted to show like the room coming together with all the stuff in it so I was like all right let me do let me generate a blank like basically an empty room I want to generate an empty room and then I'll use that room as the base like the initial key frame and then I'll paint you know I'll paint in the stuff when I painted in the stuff it didn't work so I was like okay let me go generate kind of with the same prompt that I used for the bass image let me go separate generate the room that I'm looking for when I go into editor repaint it and

Play from 0:47:04 then use image reference because the image reference really helps the editor and then bang that it like kind of all populates inside there like I'm using gener to fill right so then you can show everything stays consistent like the bricks you know that's the thing with key frames these little tiny things like if all the bricks on the wall was a different room but it was kind of similar you'd see all the Bricks Like morphing and moving like it was you know an LSD trip or something um but that's kind of how you can consider these key frames and like what you're thinking to control yeah it's sometimes hard to control on the video tools what you exactly want with just your tax proms oh thank you so much for the pro tips honestly Dave gave away some amazing ones about how he thinks about building scenes and stuff and this was kind of lacking the minia max Dynamic motion you were talking about with a

Play from 0:47:49 camera that you can achieve now and the amount of props coherence they actually give so less and less those workarounds you know will have to be but because you're already doing them you're sort of sideswiping what's currently capable and going a little bit ahead which is why your videos are standing out so much on LinkedIn if you ask me you know like really Next Level stuff it's fun we've got some backtack videos to watch now throwback Thursday so this is one that was Andrew's favorite from last week and I don't know if you guys have seen this but it's basically a 90s Nik ad spec Commercial 45 seconds and in this case if you haven't seen it I want to do some sort of critiquing on it you know what you feel the pacing was like the generations were like the vibe was like we'll give it like an overall really quick score here we go

0:48:08Showcasing AI film projects

Play from 0:48:38 [Music] can I kick it yes you can can I kick it yes you can can I kick it yes you can can I kick it yes you can can I kick it can I kick [Music] it so good so fun right so good so fun

Play from 0:49:26 now there's certain things things that we notice you know like the basketball was slightly off you could notice it from the like lines and stuff but the editing added so much to that and the like style reference type like continuity of the 90s palette and like the beat and it was so nice right handheld cam kind of looked yeah exactly they're the sort of keywords that really get you into the world that's the World building thing everyone's talking about now like that was a world we just entered like a 90s world yeah I like that like that you know I don't know do you know what that was created with from the from the image side or was that that's a really great question I think it was a mixture of cing and minia Max but um that was a very recent one so um someone actually asked that in the comments earlier they were like what's

Play from 0:50:12 your preference like minia Max or clling which one would you pick I think lots of people are using them for very slightly nuanced differences right now and then combining the files and then making videos like that you know like even I'm sure that lumer and others were used in some sort of like some motion stuff but it's it's fascinating isn't it like they've all got slightly nuanced looks and like even now like the aspect ratio thing is an issue like Runway can only take two forms of aspect ratio versus cling and others can take more so then people are preferring like the taller videos and then doing it through cing and others so not having to upscale them as much or crop them or doing anything like that like so yeah it's kind of like what people are using is kind of down to the medium they want to distribute it in

Play from 0:50:59 like is it a YouTube short or is it like you know a YouTube normal video in which case it's 16 by9 versus 9 by 16 you know like these are the sort of differences that people want to to know but yeah I really like that one like as a spec ad like quickly thrown together by one person on X didn't get a lot of Engagement had less than uh 200 likes on that so really yeah really felt like that was like underappreciated you know see that's that's where I think it's like if you see that on LinkedIn and probably just it's just a there's a more open field there but you know also the pacing right because Minimax has got that faster 100% video motion right so I feel like that video specifically it took advantage of the slow parts and then also using probably Minx for some

Play from 0:51:46 of the the more real time uh video that took place in there too so I think that's like a good use case right now the way that I see it just from an outsider perspective and uh this one that we're going to jump back to back with these videos this one was made by Dave and it's his new music project really cool he used Luma for a lot of this because of the way he wanted to shoot it it's kind of like a Trippy thing that's sort of morphing a lot it's using a lot of Like You Know music video motifs and stuff like that so let's uh watch this one [Music]

Play from 0:52:35 [Music] oh [Music] oh [Music]

Play from 0:53:47 that's just wild isn't it what else there to say Dave thing wild I said this to Dave on I probably said it on Twitter DM or something like that I was like dude it's just like the the differentiating factor too in a lot of this stuff is the editing capability like you know 100% killer editing capability too because that stuff is really well seamless it's awesome so good do you know what I'm really proud of is that when I collaborated with Dave on the Battalion stuff I remixed um I think it was London Grammar he asked me to do the song Night call so I basically like ran the stems through like logic which is my Daw kind of exactly what you do with Photoshop you know manipulating it before you add it the generator so that you know complete control that's what I did I had complete control over what I was doing in udio and that was a

Play from 0:54:34 song He generated in udio and that's his sort of first for into generative music is that he was going to use a traditional music piece but then actually generated his own in the end how cool is that so he generated everything there the video the music everything no lyrics but like so cool does does are any of the like like I know I've probably asked you this before but has it increased since probably the last six months when I asked you this has like logic has that added any AI features to it it has it has and they're coming surprisingly in the mastering side of things which was what my expertise was in I actually did audio engineering as a master's degree so honestly it sad as sad as it sounds kind of my job like prior to being an agency

Play from 0:55:19 owner has been reduced to master with AI oneclick button which is insane because it will take the overall mastering of a track and dip out the frequencies that are damaging and give you like this idea of um generating to me is always felt like sculpting so instead of building Brick by Brick you get a giant piece of marble and then the AI sculpts the thing and you've got a kind of template to work with instead of starting from the ground up compositionally so like it's fascinating where the AI is coming in is in the mastering side rather than in the like programming of instruments or anything like that it's the overall Audio Dynamics and stuff like that so really helpful for people because that's the that's the hardest bit is the physics of mastering it's very difficult

Play from 0:56:05 to get Acoustics correct I mean who's got more who's got more audio data than they do with that kind of stuff them Ableton those types exactly Pro Tools Ableton they're the only ones that I can feel like reason used to be a big audio software but it's not so big anymore but yeah ableton's in the live space really really big as well so they've all got Mass audio Banks of stuff for sure that would be awesome for you know like Festival setting like big you know you can probably do you think about all the festivals that go on throughout the world or just any any club there's a lot to use there I mean it's it's fun too I mean imagine there's I'm sure you're going to be able to syn in in real time with waveform at some point like a I was thinking that and also contextualizing

Play from 0:56:52 it to every senior at you know imagine on like Taylor Swift's tour if every single stop could be generative you know like adding keywords from Miami or Florida or wherever she is these could be added to the sort of background things that are syncopated to the music bpms like the tempo is sort of like flashing with the visuals which is being made on the Fly based on stuff it would be amazing to see that every show would be different every single show that she ever went even if she did one in the same state it would look completely different you know so that's the sort of fluid thing that AI gives you isn't it is I said this about gaming imagine when Grand Theft Auto 7 has ChatGPT characters and you never have the same conversation twice kids will never know what it was

Play from 0:57:38 like to talk to NPCs with five stock CLI you know they like no idea what it was like like the radio stations also just generating music consistently exactly that was also fun I remember I forget which one that was when when they rolled that out that was just like the radio stations were epic whenever you were in the car yeah there some bangers on there too yeah that was FIFA is how I listen to half of my playlist the game FIFA and it's classic Tunes like so many jams like I beef for Classics we we talked about this R like you know we had a big I beef for like you know like 20s in our 20s like so many classic anthems ccock all of these people you know like came onto my radar because of games F crazy I

Play from 0:58:23 mean I would look back at some of the old like for us you know the Madden game games back in the day they had epic soundtracks like you you still listen to songs and you go that was mad in 2007 know it right like brilliant isn't it big deal right you know the movie soundtracks too you were always they were so big that actually obscure songs got made really popular did you notic that like few FIFA songs didn't even make the charts and then all of a sudden they got catapulted into massive Fame as the game got really really popular on the release stranger things like the came out of nowhere yeah back to number one wasn't it yeah that was insane that that was Wild Man we've got two more to show one is from the Culver cup which Dave was just a panelist at apparently

0:59:03Storytelling and distribution

Play from 0:59:08 it was hosted by AWS and it was all AI generative films um so there was a submission I really liked from that that didn't really fly as much on social media so I thought I'd give that a bit of a shout out another one was one I quickly found on Twitter but this one's really short this I'm not even quite sure how they've achieved this but you guys tell me what you think about [Music] this of the sea back to the block Snoop Dog Dog Mony the do quo on that ass but it's still same Long Beach is the spot where serve my follow me follow me follow me follow me but don't lose your g n a is there for me to up so I ain't holding nothing back and I got five on the 20 sack it's like that and as a matter of fact

Play from 0:59:55 cuz I never had a to put a niggaer on his back yeah so keep out the manuscript you see it this unless we name doggy what people are awesome it's like it's like stop motion claymation all mixed together with said right like the editing I is the one thing I really regret in my digital career is not picking up editing sooner I really think that's the Difference Maker when it comes to all this stuff right now 100% especially on the shorter side of things right I mean longer form of course too but really when you're trying to get something across in 30 seconds you know that that capability is is big time but that was I this this is why I

Play from 1:00:41 love people like people with AI is like that's just funny and awesome that someone thought that together you know because someone had that idea probably y two years ago three years ago that's just an idea and probably get shell never never see it again right that gets lost in the depth of your brain but now you know we have clation Snoop Dogg and I can't can't do anything about it you know like without these tools right like we were just talking about this with uh with Billy Bowman on on our podcast yeah the only reason I know that is because I was just I just got done editing it but we were just talking about that that exact thing right like all these doors are opening up for people like him that you know you just don't you can't do anything about it unless you had that

Play from 1:01:28 formal training or you went to school or you know somebody or this or that and the other right it just goes back to that same point of you know creativity being democratized which I hate I need a new phrase for that now because I feel like that's starting to be overused but it really is a powerful thing that anybody anybody regardless of where they're at right if you've got internet and you've got access to one of these tools you can do anything that's basically in your brain and we're still in the early Innings of that um Chris have you have you seen um are you familiar with nice aties that was another one we were talking about in our last pod oh yeah I have oh my gosh she does some trippy videos she does some great stuff too which which I think is like also interesting from the video side because you know there's a there's a lot

Play from 1:02:13 of uh definitely thinking about the right person it's the kind of like Pink Floyd level of trippy videos right like uh I don't know maybe I've seen one where it's like you know like all like sushi and grandmas and they're a little morphing together oh yeah the same person then yeah that's so what's fascinating by that right is just like she's using Dolly right and then she was doing that stuff early on with the video and it's like it's not perfect it's not meant to be perfect but the storytelling aspect the creativity that goes into it really is the unlock right so it's just like with where we're at with some of the tech and the tools in the video it's not perfect right but

Play from 1:03:00 it doesn't have to be perfect to tell a good story if you do it that's right that is the thing like um it's no longer that impressive you know Rory is always on the bleeding edge and we'll get you the impressive video no one else needs to make them other than Rory he's the best leave him to it seriously he's so good at making these things just just follow his newsletter and their guys' YouTube and seriously keep up to date because they're doing it and taking the time to do it it takes so so long to keep up with it every week and make demo videos and then process them and then post them and then you know curate them and trust me I know I do all of that stuff for artificial inspiration as well so really hats off to these guys like but the storytelling aspect is everything now because the capabilities are there you know no longer is that the

Play from 1:03:46 sort of boundary that we were beheld to and as you said dolly 3 I love that you know like that's like working with logic 4 instead of logic 10 no necessary use to logic 10 if you can get a great sound out of logic 4 and you know that's the thing is like um storytelling now being the barrier is really good and we're going to start to see things like because one of the questions Andrew had was if everyone's making their own movies how will we ever watch anything and the answer is distribution is king So like the greatest stories will rise to the top because of great storytelling and their sharability with with other humans you know essentially that's where AI will never beat the creative is that we humans are telling

Play from 1:04:32 human stories to other humans that resonate and so that's where the AI tools enable that at scale like you're saying Drew I don't like the word decentralized either it's kind of like one of those words that's always like unprecedented you know used out of context and whatever it's like we need a better word for democratizing creativity or you know just making it for the people but the best will rise to the top you know that's the thing is Dave is a standout you can tell he is a director you are great like there's so many people to know and look out for for their great storytelling this one is another one like never heard of this Twitter account so before we watch the quick Culver cup one this one again less than 100 likes

Play from 1:05:18 and it really stood out to me in the initial thing so let's just watch and have a quick uh recap on this [Music] we all breathe the same air we all cherish our children's future and we are all Mortal to the 103 [Music] 103 we have been given the scientific knowledge the technical ability and the materials to pursue the exploration of the universe 4 3 2 [Music]

Play from 1:06:13 1 we can see actual constellations and um you know the Milky Way or we can see um City Lights brighter than ever [Music] [Music] stage two FL 30 seconds falc 9 is configured for flight 6 5 4 3 2 1 Zer take and lift

Play from 1:07:05 off the only thing we can be sure of about the future is that it will be absolutely [Music] fantastic what like mindblowing to get less that might be the best one I've seen editing editing incredible too if you notice how many different things are moving at once there as well wow but the shs were like everything felt right I didn't see anything that stood out like no smoke was moving in the wrong direction the bucket looked good when it took off you know everything looked like right on right and it had a feel to it

Play from 1:07:51 like it it had some color work post production to like give that like overall color continuity and stunning right amazing little Goosebump little Goosebumps on that one like the weaving in and out of the of the different you know radio voices and everything like that that's really everything looks so sharp even the like Tech you know the like white bot and everything in contrast to the like nature Landscapes and everything was just so sharp it was uh remark do you have the the Creator's name uh uh let me we all breathe the same I'm not sure

Play from 1:08:40 how to forward it to the end put it I'll put it into the comments afterwards because this is like as soon as we finish the live it'll be finished as like a recording on LinkedIn and then there'll be like live comments that I can go and add the creators to but yeah that was really remarkable right from an ex account again just found it randomly right before this session thought wow I didn't even watched the full thing I watched the first 30 seconds and thought I've got to show these guys this this is unbelievable unbelievable stuff humans right I mean like again okay somebody comments somebody just dropped the name in the comments there Claud oh awesome thank you so much for the people doing all of this I'm seeing people coming in from YouTube by the way and and Linkedin and everywhere and we got

Play from 1:09:26 plus people watching us live foot is okay my foot is doing good hey bro you had you had people really concerned about you I know it's like the stere nail in the foot you know it's like you hear that so many times like oh you got to get your tetnus shot you might step on a nail and it's like you know whoever actually steps on a nail in like public that's not work oh no yeah I did like just random day step on a nail do I have a t a shot I cannot remember but we're good there's no Gang Green it's [Music] a aren't they good for like 10 years too and then you're just like damn was it like 10 years ago like might something you don't remember

Play from 1:10:12 because you don't think you're ever stepping on a nail unless you work in that field right like was I'm working in renovation construction like when am I going to see a nail in public you know on a grassy null in New York City don't know so did oh man that's not fun there these comments are so on point by the way that was very Interstellar the soundtracks and the audio just on point like great editing sound design my God the PAC just pacing yep yeah we've got the right people here you know everyone sort of knows you know like it it's a feeling that even if you're not from a creative background you can sort of intuitively feel it that felt right right like there's something about it that permeated your background

Play from 1:10:59 and creativity and and critique ability just to go wow I was swept into that at one point I forgot it was AI That's one of the key things that I really feel like is like damn that just looks like a normal video like there was nothing AI about it really like stunning super good I don't even you know I think I have to go back to the lab now see what I can do just to like cuz that that got me all Juiced up now I boundaries gets me excited too Dave every time he sends me something I'm like dude my brain go into overdrive looking at this about what I should do and like you know he's inspired so much of what I've been doing lately it's you guys too this last one was his suggestion so the Culver cup like I said he was a panelist um really big

1:11:12Operational challenges in AI

Play from 1:11:47 hopefully we can get over there next year it's AWS sponsor so really big this is a 2 minute 22 last uh of the loaded in clips of this week I've got no idea what this is about I haven't seen this one so we're all going to be watching it for the first time you want to know what the craziest part of this whole thing is this was supposed to be my day off I wouldn't say I loved my job every day was the same the same boring customers eating the same boring food in the same boring Sun soaked tin can I'm just glad I didn't have to watch them scarf this

Play from 1:12:32 slop then one night someone screwed up we're a diner not a Jambo juice still couldn't let it go to waste I had to make sure there were no cucumbers I was afraid of them here I don't know what these things were but I doubt the health inspector would [Music] approve there were two of them and only one of me guess the customers were on their [Music] [Applause]

Play from 1:13:19 [Music] own what is your emergency this situation was getting stupider by the [Music] minute I'm pretty sure the sheriff just pissed it off what a freaking mess this dond Dash was going to pay order

Play from 1:14:06 up I was pretty pleased with myself until I remembered the sun was going to come up and melt all this in like an hour [Applause] yes like you said it doesn't need to be as you know the the one right before that maybe I should have left that till the end because of the stunning visuals but that was just fun you know like characters were there it was fun it was good pacing again good editing like a real story was told there was a protagonist cat and there were different animals as characters you know like interesting right again given someone else's tool they make a totally

Play from 1:14:52 different story like that the orang cat it's a that's I think that's an account on X that I follow and uh they did some good stuff for Runway uh runways gen 48 also oh yeah they've been entering that as well that's cool yeah I hadn't seen that one yet but I know I know I recognized the uh the handle yeah the overall yeah handle it's like good it's got a good mix of it's got a good mix of like well-timed humor like well-timed you know edits you know that slow motion like they know what they're doing right it's interesting they have a motif they kind of got built a character that's in every one of their videos then wow that's an interesting way to think about if I'm not wrong if I'm not wrong I've seen a bunch of videos with the orange cat so maybe I'm wrong sense yeah good

Play from 1:15:39 marketing especially because you got the brand there you know one thing I've noticed about AI creatives is that videoo video there's no sort of continuity it's just sort of everything and anything it's no real like personality infused or like character infused or like brand infused yet like fascinating space isn't it everyone's you know that's what um Dave actually said he said you would not know how many Hollywood producers are using pseudonames and doing this stuff because of the wrer strike and they don't want to get you know traditional budgets you know taken away from them but they're still experimenting with the bleeding eggs technology because of course you know we've seen people like um stability AI take on James Cameron doesn't get bigger the pioneer of CGI and graphics

Play from 1:16:25 and VFX in Avatar and places like this spent the most budget possible is now working on leading end products on the front edge of AI so like of course other directors are interested in it they just don't want to necessarily attach their name to it because of fear of the backlash but they are interested and they are doing it so like just be aware of that everybody you might not know who they are they might use an alternative name like I've seen Paul for a long time my co-founder was Jedi Master that was his sort of like like wter ego in his Instagram profile and it was cool because he had a Star Wars theme running through his Generations then we ended up talking to Disney not cool anymore right then we like well we're all official it's like Paul Parson's art because now you know the fun little handle that we had is

Play from 1:17:12 maybe going to upset a potential lead of the future so then it's like Okay so we've gone through iterations with this you know thinking about how to like obviously you guys um have the podcast and and other things that you do in personal brands have you like Dave's part of this thing called secret level which is a collective of AI people that are like answering briefs at scale and they do with everything like advertising film making and other things like this is really cool and I'm seeing little collectives formed together you know it's like a new wave of things isn't it like the micro teams of like two or three people can answer big briefs in tandem with other cool people and we're all sort of getting our special I'm getting in the sound design Dave's

1:17:20The future of image modeling

Play from 1:17:57 really getting into the cinematics Rory you're getting into like probably all of the advertising for all of these platforms they should be paying you thousands for the amount of exposure that you get them wild yeah right yeah please that would be great I I'll appreciate any donations thanks guys yeah you guys really need to partner with this and we want to get you know access of these tools out to our communities because we've all got really good communities that are interested in this Tech it's much bigger than us elves when we're like you know broadcasting this to other people like that Minx stuff they're a team of six people every bit of marketing dollars helps them you know like it's crazy these times we live in if anyone's looking for a little referral there Minimax if you go on their Twitter they just uh opened up

Play from 1:18:44 applications for their partner program so if people want idea go ahead get them get I'm so happy because these video platforms are expensive so Runway cling you know they should open up partner programs referral programs like it really makes a lot of sense and it's Community Building you get like long-term customers I really want to see like what Midjourney does in this space because um the photo realism when we cracked it I think that's the thing that like image models you said earlier Rory you know image models aren't advancing much more PA wondered how much more can they Advance after photo realism it's really hard because flux now you can type in stuff that literally looks like a photo you know like no further flux 1.1 I saw one and some hack they'd used where

Play from 1:19:31 they'd actually typed in IM mg like you would store a photo in your phone underscore and then a random number and then a genuinely almost real photo pop through that was the prompt it was just like how store an image in your phone which is almost creepy to think like was that a raw pull there I mean there are a couple of those out there like the the whole snap chat trick you could put Snapchat photo Instagram photo all those work pretty well we we've covered like from a cinematic point of view Rory and I have talked about sort of the Stills archive good one like shooting fish in a barrel like you could just literally put that in um and get stuff that that looks real right it looks it looks off you

Play from 1:20:16 know like you could even put in the the netflix.com the hbo.com Behind The Stills archive and you're going to get something that looks like it's a real show um so so where does go from there Drew that's a good question you know like where where can we progress to when it's almost like unidentifiable real you know like I wonder where pression can go I I really don't know I I just think like with yeah like it's it's all gonna I think just come together but I do think Midjourney's personalization or that ability to almost like get something that feels like what you would expect in like your preference as far as Aesthetics or camera angles or composition or all those little intangibles that you look

Play from 1:21:02 for and getting that refined feels like a little bit of something special that there's still room for um but outside of that I don't I don't know I don't know if there is anything left on the image gen side I'd really have to think about that I wonder if like the 2D will power the 3D in the way that you know G in Splat will so I suppose like there's a lot more pixels to add in order to create a better 3D version of that 2D image you know maybe we're not even going to notice the difference ourselves from the 2D but once you add it to a 3D then it will be more immersible or something like that I just it's really hard to know what they're working on like what can Midjourney 7 do that you guys wouldn't like from a control point

Play from 1:21:48 of view totally understand you know if we had more control more ability to do things like maybe like continuity continuity yeah lauras it would be nice wouldn't they in Midjourney like uploading say 50 images of someone and then having it one for one that would be great like the character reference mid journe is now behind on I would expect that to get better the the continuity so like going back to their idea of Storytelling and that be becoming a focus where they've got a team that's just focused on that we don't know details but I would assume like if you're thinking about a single image how much better can it get but if you're thinking about a series of images I think that's where there's there's some creative room for

Play from 1:22:33 new features for some more consistency and cohesiveness which could also right I would imagine fuel whatever they got planned on the 3D and video side so I heard that the mid Jour is working on like uh giving you eight or more suggestions per prompt that seems like interesting you know it's kind of like what's your preference would you like the live prompt thing like Korea does you know where every word you type it changes or would you like eight versions of the exact prompt that you enter that's I want hear Rory's answer first on this yeah it's a tough one I'd like the live version that's me though I think so live one control freak like Paul Paul is exactly sculpted is prompt to I want to

Play from 1:23:19 see that please you know he doesn't want to see the process or the progress he knows exactly you know the he's yeah yeah I would want to I would really want like some of these other tools I've seen them on the open source side where it's like you know I can go and edit the face and like change the expression or like tilt the head or turn to the side you know like like that where it's like you know click and drag or however I want to prompt the face to be different but keep the same image because like remix is good but it doesn't necessarily work all the time you know the editor is good it doesn't work all the time like you know trying to do style reference or image you know reference to imagine if you combine live portrait with Midjourney then you've got last and end key frame really easy because you could just like move the eyebrows move the face and they drag it to where

Play from 1:24:05 the position that you want the last key frame to be and then I wonder you know will Midjourney do the interpolation between first and last key frame do you think if they do a video model obviously you're using Midjourney to generate the image why not natively do it in Midjourney to video as well if that's the best one it's tough to know isn't it they could toally get on this video sphere they've got a lot yeah I do too I don't know why because like look they they basically created with style reference they created a Laura with one image right like how crazy is that to think abouty wild right it's like that's just one image and you can do whatever you want with that one image from a style perspective now character reference doesn't work as well but I would say like this I feel like this video this

1:24:22Business models and ads in AI

Play from 1:24:51 video side everyone has like a kind of defined sort of look at what it's going to be like and this just it's just a hot take who knows if it's right who knows if it's wrong the the video side everything kind of looks like the same like everyone's doing the same stuff like key frames motion brush like everyone kind of took what Runway did initially and like put their own spin on it as we know Midjourney just does things differently so this might be totally different it might be a totally different model in terms of the way they generate videos um you know I'm pretty sure like if I've gone down the the rabbit hole appropriately on the video side you know they are generating all those frames individually in between you know start and end and then it's also referencing against like data set that is motion but then also diffusing it you

Play from 1:25:37 know on top of that right so it's very it's a very complex thing to try to like understand me personally but I've always they've always thought about things differently the way they even the way they combin the editor feature with the gener like the uh a few things that we were asking for they did it in a very smart way they combined a number of different like pain points we had and they made it like generative fil like it's they took like five like four or five steps that you would have to do individually and combined it into that one remember that Rory because it was like you could you could brush something out and then you could scale out and then you can change the composition you could you know you could even crop in which was which was UN you

Play from 1:26:23 know you couldn't do that before in in program uh in tool so it was like yeah you you could combine all these things which that was always like the maybe the one thing about Midjourney is just like some sometimes you have to take several steps to get somewhere and I think that's cool in some ways obviously but it's it's also going to be obsolete very soon so that was like a preview with editor that I was really really excited about because Rory and I would just sort of roast this uh you know commonality that you'd get with the Symmetry in a lot of the photos unless you like really went out of your way to prompt for asymmetrical or really give it compositional details and so forth so it would mirror the uh other side yeah I totally understand what you mean by that and it's funny because Paul asked is Midjourney trying to be Adobe as soon as

Play from 1:27:09 they did that you know because Jen Phil Photoshop was the only reason we used workflow of that at that time and then obviously when the editor came out you were like oh wow then like mid Jour is going for it they want you to be the editor the upscaler they want you to be other things too like everything's multi modal now everything's trying to be multimodal whether they want to or not is not really a choice anymore people demand it you know like it was funny Drew and I were talk we did this live and we were talking about it and like like what could what could Midjourney do with the like the stupid edit feature that wasn't an edit feature at that point it was like you know uh reframe or replace whatever the hell it was called i w we don't even remember no we were like repaint

Play from 1:27:55 very yeah trashing it and then he we were like what do you want and it was just like just give us generative fill exactly how that works like so we can expand the canvas we can like sculpt within there and it was like three days later it was like here's your janitor Phil canvas like awesome thanks guys exactly kind of how we wanted it that was uh because it it is a great like Photoshop I don't think it's the generative fill like ease of how how they like very from a first step concepted it very well like that to me was a first step concept that was very easy to use like especially with how complex Photoshop can be um when I first saw it I was like oh this is really this is like the simplest thing they've ever done like I can't remember the last simple thing that I that has been released on on Photoshop like camera of

Play from 1:28:44 raw editor like 10 years ago whenever that was like that I just that's such a good point actually because even people like me I'm not that expert in Photoshop but having lasso around something and then hitting a generate button and adding a prompt easy you know it's the other stuff in Photoshop all the other hundreds of tools and the like you know seven steps and the like background and all that other stuff that that's the daing part so you're right it's simple wins sometimes you know like solving problems in a simple ux as you said Drew that's going to be the key like Korea could do it others could do it we we're all sort of on a knife edge right now waiting for the place that's going to demand all of our attention because they have simply the suite that we want you

Play from 1:29:30 know and everything even LTX Studios and others they're trying to do this you know they're trying to make a new way of doing like storyboarding and they're sort of like having a dashboard that's storyboard orientated and so if that's your sort of flow that's maybe the one you're going to gravitate to it's just fascinating everyone's getting their Niche I think you look at the brands too right like Adobe has like a very loyal fan base I used to be one of them but now I'm kind of I'm swing back and forth on them I feel like it's always different but like figma to me is doing it in a very interesting way also the way right they they've a lot of their stuff is like uxui design like it's trained on that you can generate you know I think that's available now um

Play from 1:30:16 they released it a couple months ago I think it was in June but they had to like go back to the drawing board but that to me is crazy like how you can spacing King it's like a design system which is nuts because that's something that you know if for designers it's awesome it's a starting point like you can start there and then edit from there instead of just building every little asset piece by piece granted I know a lot of people like to do that me personally every once in a while if it was under the gun need to get something done with like a starting point that was great um so then you know that's really the other thing they've done is like these little there's like a subtle subtle change that I was just like these guys and these women and whoever's running that or making the decisions Whoever thought of this little issue like the layers inigma often times I'm

Play from 1:31:02 so unorganized in there which I know everyone is probably way better than I am it's you know you don't have anything layered it's just like image 4576 and like copy layer of text that's whatever they have an auto they have an auto name layer feature which is so simple in like you just hit the button and like it names all the layers and it's like now I can organize this instead of going back through 45 layers and like labeling each thing it's so such a simple fix that's how you know they they know their customer like that right therei that's so funny you mentioned that one too because my mind is a tool where you can oneclick thing to the cloud and it will AI orientate and Tagg it it's brilliant it's the best thing for a curator like me that's

Play from 1:31:48 running multiple communities because I'm essentially coming across hundreds of posts and I'm like okay I don't have the time to go keyword tag all of these and I just want to oneclick them and have it all orientated for me you know like oh that was a creative piece and it was a video format and it came from Instagram it should sit in this folder brilliant I've had that app on my phone forever and I've never opened it oh man if you are a person like me where like literally we run multiple communities you hoard bookmarks absolutely aren't we all you know other people I know they email themselves they WhatsApp themselves and I'm like no I just want to one click this to the cloud and have it all stored for me you know if there's any system to do so that was a system that was just made for me it was like exactly my

Play from 1:32:33 problem that I dealt with and it's the same with like you and stock libraries no doubt you know half your week spent doing this so it saves a god an ungodly amount of time to just one click everything to the cloud nowadays is and my attention too I noticed that like I can really be more focused instead of having to oh that's a great piece how do I store it what do I think of as a a caption and all of that stuff it's no just like click click click click click and then all the captions come in a different segment of the day or whatever you know I can like really change my routine because of this to fascinating how one problem solved and I'll pay those guys hundreds of dollars a month if they want me to they organize my life it's like my second brain so it's like it's ultimately exact I was thinking about that the other day I was thinking

Play from 1:33:20 gez that's a really good llm I'm building you know all of my thoughts and daily things like that's gonna be valuable one day people might want to PR against that you know like as a database I'm starting to build for myself it's really weird I never thought of it until I just clicked and I was like wait now this got 10,000 videos in it I wonder if that's valuable to someone of course it is that's a training data set right there it's a training dat right there I didn't even realize I was doing it I was just abstractly I feel like that's the recurring thought right now it kind of underpins like I'm always like lately especially the last month or two just kind of thinking about this problem okay I know I'm gonna solve this problem with AI you know and whether that's image gen

Play from 1:34:07 or whether that's an llm like chat gbt but there are all these specific use cases right and the way that you would do that now is you you gota upload all these docs you got to give you know instructions like I mean it's quick but it's not quick at the same time imagine when that becomes much more consistent and efficient like that is going to change the game that is super super interesting it feels like we're about 12 months away from not being able to buy any Tech where it's not software built you know all the latest iPhones Apple intelligence all the latest Microsoft computers it's going to be like co-pilot built in you know you're GNA have to opt off of it basically yeah it's going to be just there in the background I was thinking about building a second desktop you know just to have a

Play from 1:34:53 virtual desktop which is ready for the co-pilot to access because that will essentially be your clippy around your browser and everywhere you go on your computer like that's the thing that's coming isn't it is who's our Jasper who's our clippy who's our like I wonder if you can give him a name I want him to be a really cool Butler with like a little bow tie and then access access anything that you've come across or thought of that you know in modern time like you wouldn't be able to recall certain corners of of your brain where those things live right it's just it be so nice they switch on and off wouldn't it like have like though if you get then you might be able to find any memory of anything you've ever read at any time which would be like now we're getting sci-fi right my favorite part

Play from 1:35:39 about that was the guy like the first guy whatever his name was he was like a big Call of Duty guy and he's just like dude this game has become so easy because my Rea there's no reaction no reaction time right know he's just like owning everybody but yeah that's that's interesting I didn't even think about building a second desktop for you know it's it's again this is why I think this is why it's near future that's what I'm thinking about and it's just why this is you know I talked to a lot of people a lot of businesses I think very interesting how Nimble you can be with a Sixers team and how hard it is now with a 1500 person team to get integrated onto this stuff oh it's more than just infrastructure and tools like it's it's your whole operation like you need to figure out how this all plays in

Play from 1:36:25 from an operational standpoint you know because it it changes like the tools provide different skill sets now to different people you might be cross trained in a certain thing you might have to like legitimately change your operating structure like the how things get done because of the way that the tools function not saying it's bad like you're good bad you have to get rid of people it's just like skill sets change like can marketers now tell stories and do things that are a little bit more than they've done in the past a little bit more than data like that's what I mean so there's it's very very interesting for how nimble like you're saying like these little Collective teams can be with the tools while it's really hard at you know a really corporate ex you know highly infrastructure level because then also you're you're to the tools right like if

Play from 1:37:10 you're a Google company and you have 5,000 employees you're getting Gemini like personally you're not going to go switch to co- pilot like CH this we got to retrain 5,000 people on Google we got to get everything like migrated over to me that's a lot of working so you know it's like you're kind of tied with whatever decision you made via software which interesting to me also is like to me S like the from a bigger organization perspective right like let's talk about it like that like maybe we're talking your Coke level your Proctor and Gamble whatever something like that to me not once has software ever really had a competitive Advantage where it's like you use Google and you have email video meetings you know calendars

Play from 1:37:57 whatever same thing with Microsoft like ideally kind of in the same realm now like if Gemini is better than co-pilot like you have a competitive Advantage like I don't I don't know if that's happened before maybe I'm thinking about this the wrong way but like that's it's always just seemed like the same tools you're just choosing which company to go with right now it's like there is a competitive Advantage there is like a little bit of a race am I am I nuts with that thought or is that just no not at all that's a really good like nail on the head of where it's coming to and I think that's why initially um Microsoft and Bing and open Ai and the 10 billion was all because you know Google had a 95% sharehold of the search engine market and all of a sudden it started to be that oh there can be another player because of a tech Advantage you know

Play from 1:38:42 like essentially what they were looking at and now like all these other ones are popping up out of nowhere like anthropic clae and others like meta's got llama out there and then other people people are tackling it in a totally different way like I use um when I type uh just a new word now instead of opening a new browser on my Chrome tab I open perplexity because perplexity is better default now yeah isn't it amazing like because I realized three quarters of things I was Googling I'm going to end up on a quar forum and that's not the way I want to get the right answer I want perplexity to average out the answer and give me the answer please you know that's the way and site your sources all of that matters all of that matters that's just solving zero zero

Play from 1:39:30 CLI yeah zero click zero click Chris because like if you're curious about a company do you know do you know Drew no longer perplexity is shipping round right now for ads ads inside of perplexity so I I talked about this several weeks ago because we obviously come from an adver what do you think about that that's a really big play and it was ultimately coming because Andrew said the monetization and the cost of all of this training and like it's what it's why zukerberg is open sourcing everything meta I feel like it's the attention Play Because ads is where the money is it's not the $20 subscription it's the can I sell them more advertising because the data we have on them so like perplexity the deck was one

Play from 1:40:16 of the most impressive things I've ever seen in my whole career really it was 80% they they guess get this 80% of the users of perplexity they guesstimate have an undergraduate degree 80% damn it was the most like educated like customer base I'd ever seen and I was like imagine the like wealthy positions they're in and countries they're in and their buying power and like Advertiser is going to be all over that it's going to be the new Facebook advertis on the new Google because it won't be so competitive at the start Facebook clicks were one perer click I remember buying Facebook at one cents a

Play from 1:41:02 click that's when Facebook [Music] markets nowadays for that same click it's like $17 something you know like astronomic depends what keyword you're going for like but yeah if you're in the retail space you could be looking at hundreds if not thousands of dollars I wonder if perplexity had any aate marketing sort of plays there because they could probably be the best affiliate marketer of all time just like anything right like bam I mean like that's a that's an absolutely genius platform for automating the most money you could possibly ever create I don't know you don't there's the advertising model is very interesting to me on how you know how are they going to do the first time the other day I saw this on a custom GPT

Play from 1:41:48 I don't know maybe I just don't use a lot of them outside of my own but I looked for one and I found one and then like every prompt that I prompted in there at the bottom it it says like sponsored by X company like so someone oh seen that do you know how they came out um when you make the GPT you write that in as a rule yeah so um yeah I've seen say that again Chris so um when you make your custom GPT you have like things that you can upload as documents and rules that you can give it so it's basically baked in to every time you respond you know give a link hyperlink back link to the person whose response you got the thing from as a like advertising space so it's amazing isn't it like wow okay initially the lead magnet side of things was made by and

Play from 1:42:34 then the name and it could sometimes be hyperlinks but to get it in every response is like going to Google and the first responses are ads you know like I wonder though about the value of this because right now we're in the Golden Era the pre- advertising era of everything is like the place where you're not being sold to every fourth post or everything you can trust everything because it's not got an incentive to sell someone hasn't buy to play you know all of these words come up when advertising is involved so it could crash hard honestly on them or it could be the thing that everyone goes holy God they're making 30 times more money than me I need to include advertising in my AI product and then it clean sweeps and everyone's got it you know I've never

Play from 1:43:19 seen an alternative to advertising though in terms of Revenue generation and if for Big Data companies because this is what these guys are big data companies you know so advertisers are willing to pay the most money for the big data that they collect on all of us so unless someone else is willing to pay for it or they find another way to monetize it like large language model training or something like that it's hard to know that advertising will ever be beat to be honest it's tough It's like I I was waiting for that moment it popped up in ChatGPT like at some point I was just not expecting it as a as like a disclaimer at the bottom of a custom GPT which then I was thinking I'm like that's actually pretty smart and I had seen that before that's crazy but then I also thought we as marketers ruin everything that's how it goes you know we do yeah we ruin everything I know it

Play from 1:44:07 it's it's bastard especially the ones that don't care about their audience they're just thinking about themselves that's a great way to finish marketers ruin everything after all of the like storytelling and capabilities and stuff that we saying is so great thanks guys we stayed so much longer more and more people have been tuning in by the way we're now at like double what we were in the Midway point which is nuts to see you know these guys came in so last minute I can't thank them enough and we're gonna have them ongoing every Thursday we do this live and my co-hosts usually are like training I'm sometimes away as well so there'll be like roundtables of different people joining this live watch party and obviously they've hooked it up to X and YouTube and multiple sources are coming in through the comments so thank you so much guys you know we do this every

1:44:13Closing thoughts and outlook

Play from 1:44:52 Thursday at your local time note the time because I'm in the UK at 4 p.m. but I know these guys it's in the morning for them over in the states so uh you know wherever you're tuning in in the world note the time and then come back to this page and uh yeah we'll be doing this again live but thank you so much guys for coming in last minute seriously and how cool are those just as like a closing thing like mindblowing stuff is happening in the video space like mindblowing stuff like can't emphasize enough these are just random assortments that we showcase today you know there's hundreds more examples out there now I'm gonna shout out to all the comments too guys comments wild awesome

Play from 1:45:38 stuff in there seeing some familiar faces Glenn seeing you a bunch man thanks for thanks for joining Eric seeing you a bunch as well um and if you guys are just like bored and want some more information we did a podcast with Chris just two weeks ago yeah right pod first guest I was so honored I was so honored to guys first ever guest thank you so much for that invite as well yeah go and check these guys out on YouTube seriously they just Crush how much content they make it's actually daunting how much you guys like uh upload and stuff I need to get on my YouTube game dude come on over the the water's warm I I need to get on it mindblowing stuff yeah people are just going crazy

Play from 1:46:23 crazy about it yeah thank you so much guys right I'm going to end the stream there thanks guys thanks for having us thanks everyone for