EP 021 · 82 MIN MIDJOURNEY IS GETTING OUT-SHIPPED
Is Midjourney in Trouble? Updates, Discussions, and AMA
The short version
Episode 21 is a last-minute live stream and AMA, recorded after Rory and Drew's usual LinkedIn Live setup fell apart. They vent that Midjourney's V7 keeps slipping (now pushed to January) while Runway previews its Frames stills model and ships Act One motion capture live during the show. Rory trains a personal LoRA on Krea in ten minutes on air, and the hosts field audience questions on SREF organization, Kling's try-on feature, and ElevenLabs.
Key takeaways
- 01
SREF-mine efficiently: run one fixed test prompt (Rory's is 'angry samurai --sref random') and route results into a smart folder so codes stay organized.
- 02
Krea can train a usable personal LoRA from about 12-15 photos in roughly 10 minutes — use neutral clothing in training images so outfits stay editable later.
- 03
Hailuo (MiniMax) has the most creative motion, Runway is the most consistent all-rounder, and Kling is best when you need one very specific, controlled result.
- 04
Runway's slow, incremental feature drops (text-to-image, then image ref, then Act One) work as a teaching curve — Rory calls it smart versus dumping everything at once.
- 05
Pika is a cautionary tale: a viral effect bought it 48 hours of attention before users moved on, proving virality alone doesn't build platform loyalty.
- 06
Rory questions Sora's 60-second generations: the average theatrical shot runs about 15 seconds, so anything past 30 seconds may be overkill for most real editing.
- 07
Kling's clothing try-on loses image fidelity on render; Rory's workaround is mocking up the garment in Illustrator and re-uploading it instead of using try-on directly.
Receipts · said on the record
“It just seems like they're getting out-shipped left and right right now. … They're not VC backed. They don't have any funding coming in.”
“MiniMax to me still has the best motion — if we're talking creativity scale, MiniMax has the most creative ability. Runway is the most consistent; I can do about 90% of what I want there. And Kling is when I need something extremely specific.”
“That's where there's opportunity, right? You gotta build brand affinity. You've gotta increase your stickiness so people aren't just gonna choose what's latest and greatest.”
“Twelve to fifteen images, and it even gets my sunglasses — it nails those, the ones I was wearing that day. Some of these images, I could pass off. I could pass that off as an image.”
“The average shot length is 15 seconds in feature cinema. … They really only need to be 15, 20, 30 seconds. Sixty seconds to me feels like overkill — that starts to feel like video game more than an actual usable shot.”
“I'll create a smart folder within the Organization tab, and I'll use one token for mining SREFs — my prompt is angry samurai, dash-dash-SREF-random. I'll only do angry samurai when I'm generating SREF codes, so everything just filters in there and stays organized.”
Chapters
- 0:00:00 Weekly update and banter
- 0:03:55 LinkedIn live challenges
- 0:06:21 Midjourney V7 updates
- 0:11:31 OpenAI Sora expectations
- 0:13:59 Midjourney personalization
- 0:17:52 Runway frames model
- 0:24:26 Creative AI opportunities
- 0:29:16 Magnific editorial portraits
- 0:32:48 Macro photography AI
- 0:36:29 Photon and tool comparison
- 0:38:46 AI storytelling features
- 0:42:23 Kling and Luma updates
- 0:46:06 Krea Lora training
- 0:51:45 AI tool subscriptions
- 0:58:34 Storyboarding and Krea
- 1:02:53 Future live show plans
- 1:13:01 Midjourney technical tips
Questions this episode answers
Is Midjourney V7 delayed again?
Yes. As of this episode, recorded in early December 2024, Drew and Rory report that Midjourney confirmed V7 would not ship by end of year after all and is now expected in January, with a follow-up 7.1 planned for features that miss the initial release.
How do you keep Midjourney SREF codes organized?
Rory suggests running one fixed test prompt, like 'angry samurai --sref random', every time you're hunting for new codes, then routing the results into a smart folder (or an underused Room) so every SREF you discover stays filtered and easy to find later.
How good is Krea for LoRA training?
Rory trained a personal LoRA on Krea live during the show using about 12 to 15 of his own wedding photos, finishing in roughly 10 minutes. He recommends using neutral, plain clothing in training images so outfits stay easy to change afterward.
Which AI video tool is best, according to Fast Hours?
Rory says it depends on the job: Hailuo (MiniMax) has the most creative motion, Runway is the most consistent for everyday use, and Kling is best when you need one very specific, tightly controlled result.
What did Fast Hours think of Runway's Frames model preview?
Rory and Drew sat in on a briefing with a Runway co-founder and came away impressed by its cinematic, less 'plasticky' faces and natural motion cues, though they note it launched as text-to-image only, with style and character reference expected to follow.
Show notes
Midjourney is getting out-shipped left and right, and now video players like Runway and Luna have entered the AI image chat.
Is Midjourney in trouble?
We discuss that, the most recent Midjourney news, the latest changes in the AI creative landscape, and answer your questions live on the stream.
