EP 023 · 74 MIN SORA'S BLEND FEATURE HAS NO RIVAL
Early Sora Tester Shares Everything
The short version
Episode 23 features Sora alpha tester Minh Do, a YouTuber and filmmaker who spent months testing OpenAI's video model before its public release. Minh walks through Sora's blend, remix, recut, and storyboard tools, argues its camera control and blend feature have no real rival, and describes how OpenAI's roughly 300-to-400-person alpha program actually ran. The conversation also covers Midjourney's new moodboards and Patchwork features and the industry's uneven shift toward AI video adoption.
Key takeaways
- 01
Sora's blend feature merges two separate generations into one continuous shot, then lets remix clean up the seam — Rory calls it the tool's most unique, uncopyable feature.
- 02
In Sora storyboards, be more specific, not less: Minh gets better results describing motion at every second (cube floats up, then down) than one broad instruction.
- 03
Place a Sora storyboard's text card about a second after the reference image, not right on it, for smoother transitions between beats.
- 04
Sora's recut lets you regenerate just the bad three seconds of a ten-second clip instead of rerolling the whole shot — a real production-time saver.
- 05
Being a known 'tinkerer' in the LA/SF AI art scene, not just a big channel, is what got Minh cold-emailed into OpenAI's 300-to-400-person Sora alpha.
- 06
Minh calls Hailuo and Kling genuinely strong for one-shot generations, but says only Sora can deliver certain camera moves, like a true gimbal-style orbit.
- 07
No one can out-train Google on video: Rory argues YouTube alone hands Google more raw footage than any competitor could assemble.
Receipts · said on the record
“It's insecurity, it's ego. You have to eliminate ego in a lot of ways to really change perspective, and be okay with these people that have already publicly shamed AI: it's going to be even harder for them to pivot.”
“They were just kind of cold-emailing, finding people out there who were experimenting and tinkering. … I know a bunch of people in the LA-SF AI art scenes, and so I was kind of a known entity.”
“If there is one differentiator that the other tools can't sort of compete with Sora yet, it's the blend and remix functionality. … Blend to me is like the most unique feature. There's no other tool that can do that.”
“It just felt like some other programs, it's just like, hey, you're in the program here, you get credits and alpha access to some features. … But yeah, I felt like that was disingenuous.”
“We're going to have all these different companies that have different personalities. Some of them aren't going to make it, frankly. … We might lose those personalities, or they become part of other larger personalities that we have to interact with creatively.”
“How does anyone compete with Google for training data? How is anyone going to compete with YouTube? There's billions of hours of content and they own all of them.”
Chapters
Questions this episode answers
How did Minh Do get access to OpenAI's Sora alpha program?
Minh says OpenAI cold-emailed known tinkerers in the LA/SF AI-art scene; he was one of roughly 300 to 400 testers who joined weekly calls and hands-on office hours with the Sora team.
What makes Sora different from Runway, Kling, or Hailuo (MiniMax)?
Minh says Hailuo and Kling are genuinely strong for one-shot generations, but argues Sora's blend, remix, recut, and storyboard tools, plus camera moves like gimbal-style orbits, aren't matched elsewhere yet.
What is Sora's blend feature and why does Fast Hours like it?
Blend merges two separate video generations into one shot, letting the user control how much comes from each clip and where the transition happens; Rory calls it the platform's most unique, uncopyable feature.
What did Fast Hours think of Midjourney's Patchwork feature?
Minh Do says he logged about 10 hours in Patchwork and still finds it hard to use, calling it a promising 'world building' canvas that needs more iteration before it's genuinely useful.
How should you write a prompt for a Sora storyboard?
Minh recommends being granular rather than vague: describe the motion happening in each individual second instead of one broad instruction, and place text cards about a second after the reference image for smoother transitions.
Show notes
MJ:FH Ep. 23—Early Sora Tester Shares Everything (with Drew Brucker & Rory Flynn)
Sora alpha tester Minh Do joins the show for an insider's look at OpenAI's groundbreaking video generation tool. Minh shares his three-month experience testing Sora, revealing previously unknown details about the tool's development, features, and the close-knit alpha testing community. The conversation explores Sora's unique capabilities including blend, remix, and storyboard features, while comparing them to other AI video tools in the market. The trio dives deep into practical applications, discussing how features like recut and camera movements are revolutionizing creative workflows. Minh also provides exclusive insights into Sora's development process, the weekly calls with other testers, and OpenAI's hands-on approach with artists.
Key topics include: Behind-the-scenes look at Sora's alpha testing program Detailed breakdown of Sora's unique features and capabilities Comparison of current AI video tools and their strengths The evolution of AI in advertising and creative industries Future predictions for AI video generation
