EP 031 · 68 MIN RORY'S 30-SECOND CLIP FOOLED VFX ARTISTS
Midjourney: New Tricks, Wild Updates, and Deep Cuts
The short version
Episode 31 mixes 90s movie nostalgia with a Midjourney office-hours recap (V7's Omni-Reference, a rumored MiniMax video partnership, new hardware) before Rory walks through building a custom GPT to prompt Google's Veo, turning vintage F1 footage into a projection-mapped clip that professional VFX and 3D artists mistook for real production work. Drew separately shows Runway's reliable on-image text for a Goldcast rebrand and a print feature in Midjourney Magazine.
Key takeaways
- 01
Build a GPT around a tool's own suggested prompts: Rory fed 15 of Google Veo's example prompts into a custom GPT to reverse-engineer its narrative, emotion-heavy prompting style.
- 02
Rory's viral trick: projection-map vintage race footage onto a modern car in Veo with one text prompt, then keep re-running it since generations are cheap and fast.
- 03
For on-image text, skip Midjourney: Drew says Runway rendered a multi-word sign correctly roughly 9 or 10 times out of 10, with no cherry-picking of results.
- 04
Midjourney video generation struggles badly with sports and basketball; Rory says video-to-video is currently the only workable approach, not text-to-video.
- 05
Keep more than one AI video tool in rotation: Rory picks MiniMax for playful, unpredictable spins on a prompt and Kling when he needs precise, literal control.
- 06
Midjourney's office hours flagged Omni-Reference (faithfully carrying logos, characters or objects into new scenes) as the V7 feature to watch, alongside early wearable and large hardware projects.
- 07
Music carries a finished AI video further than expected; Drew estimates it's roughly 75% of what makes a clip feel like a polished, finished asset.
Receipts · said on the record
“I had a lot of people reach out to me who were VFX artists or 3D artists, like, did you know how long that would have taken? And I was like, I hope more than the 30 seconds it took to generate.”
“Kling is so damn precise to what you tell it to do. But there's also the fun aspect - I want the AI tools to take over a little bit and put their own spin on it.”
“The music is really 75% of it. You can have a great video, but to get it to be an elite sort of asset … the music does so much, it really does.”
“I've only found a few things it has trouble with: basketball, number one. Anything sports related, just don't even try. If you want that, go do video to video, because text to video is not going to be great.”
“There was a run of good movies between the mid-2000s and 2015 where everything was just funny, starting at Anchorman and Old School and going up until about 2016. Then it kind of just fell off a cliff. We don't get funny stuff anymore.”
Chapters
- 0:00:00 Travel and city reflections
- 0:02:36 Classic movie nostalgia
- 0:06:40 Movie talk and airplane habits
- 0:10:03 AI tool trends and updates
- 0:14:41 Midjourney V7 speculations
- 0:20:11 AI moderation and features
- 0:24:35 Future hardware development
- 0:30:36 Business leadership lessons
- 0:35:47 Community and project updates
- 0:38:41 Brand storytelling with AI
- 0:45:30 Prompt engineering for video
- 0:52:16 Creative workflow demonstration
- 1:02:54 Final thoughts and wrap-up
Questions this episode answers
How did Rory create the viral vintage-F1 projection-mapping video?
Rory built a custom GPT from 15 of Google Veo's own suggested prompts to learn its narrative, emotion-driven prompting style, then used it to write a prompt projection-mapping vintage racing footage onto a modern McLaren. It worked on the first generation, and other creators began recreating it on Twitter.
What does Fast Hours use for text inside AI-generated images?
Drew and Rory skip Midjourney for on-image text and use Runway instead. Drew says Runway rendered a multi-word sign correctly roughly 9 or 10 times out of 10 without cherry-picking results.
Which AI video tool is better for creative work, Kling or MiniMax?
Rory says Kling is more precise and literal about following a prompt, while MiniMax has more of a playful, unpredictable "creative streak" that adds its own spin, which is why he'd pick MiniMax for a rumored Midjourney partnership.
What did Midjourney's office hours reveal about V7?
The hosts relayed that Omni-Reference (faithfully carrying logos, characters or objects into new scenes) is the flagship feature under consideration, alongside multiple video-strategy paths (partnering externally, adapting open-source models, or building in-house) and early-stage wearable and large-footprint hardware projects.
Can AI video tools handle sports footage well?
Rory says no: basketball and other sports-related footage consistently gives AI video generators trouble, and video-to-video is currently the only workable approach rather than text-to-video.
Show notes
Fast Hours Podcast Ep. 31—Midjourney: New Tricks, Wild Updates, and Deep Cuts (with Drew Brucker & Rory Flynn)
Drew Brucker and Rory Flynn are back, and this episode is one giant creative rabbit hole.
From the Valentine’s Day chaos of Brooklyn (spoiler: crying and semi-trucks were involved) to Birmingham, Alabama’s undercover coolness, the duo kicks things off with some quality storytelling before diving headfirst into Midjourney madness, AI breakthroughs, and wild tech speculation.
They wax poetic on 90s movie nostalgia, aggressively defend Heavyweights and Dodgeball, and collectively mourn the death of genuinely dumb, hilarious comedy. But then—like true AI nerds—they pivot hard into a full-blown discussion on Midjourney’s updates, AI video tools, hardware rumors, and the next wave of creative technology.
Somewhere in the mix, they channel their inner Steve Jobs, dissect Nike’s early hustle, and debate whether AI-powered wearables will be revolutionary or just another overpriced gadget nobody asked for.
There’s also a completely unnecessary deep dive into AI-generated fishing footage (don’t ask), a rant about why Midjourney’s moderation is suddenly allergic to the word "mucus," and some genuinely useful insights on AI in branding, prompt engineering, and the future of digital creativity.
- Brooklyn on Valentine’s Day is pure chaos. - Birmingham might be cooler than we thought. - 90s movies had a magic we don’t get anymore. - Midjourney updates are coming… but when? - AI tools are evolving, but they still frustrate us. - AI hardware could be game-changing (or totally useless). - Steve Jobs was onto something. - Community feedback shapes tech more than we realize. - AI-generated video is getting dangerously realistic. - Midjourney’s biggest challenge? Keeping up with its own hype.
Buckle up—it’s a fast, weird, and wildly entertaining episode.
🎧 Listen now on YouTube, Spotify, and Apple Podcasts!
