Pinned Tweet
guys, rea=Reverse Edit Anything for now.
you can do this with Claude Opus 5.5 and a single prompt.
drop a comment and i'll dm you how to do this.
a reaction shot isnt filler between generated lines. sometimes its the only place the audience gets to decide what the scene means.
a useful alternate take keeps the timing. change the performance and the duration together, and every downstream cut becomes a new problem.
editors cutting generated dialogue: do you lock the audio before making coverage, or let the performances change the cut?
an ai close-up can look perfect and still be unusable. if the eyeline points the wrong way, the scene stops making sense.
making money with ai video? clients dont pay for the generation. they pay for the revision round that doesnt break everything else.
the best ai shorts this year were won in the edit, not the prompt. hold length, where you cut, what you leave out. same as it ever was.
ai footage cuts badly for the same reason stock footage does. nobody shot it for your edit. generate for the cut, not for the clip.
I asked Claude Opus 5.5 to recreate Apple's Creator Studio promo. One prompt.
The test wasn't whether it could make a nice clip. It was whether it could follow a specific edit.
Original above. Recreation below.
rerolling a whole shot to fix one line read is the most expensive habit in ai video. change one thing, keep the rest.
the model stopped being the bottleneck. a clean 30s take is one prompt now. the slow part is getting the one beat you meant without rerolling the whole shot.
I spent weeks trying to clean up GPT image noise. This is the best result I’ve gotten.
Reply "shoty" and I’ll DM you exactly how I did it.
Using Higgsfield, ChatGPT, and Premiere separately?
Shoty brings them all into one workspace.
Here’s the demo.
This Post is from an account that no longer exists. Learn more

