Back
NoteDec ’242 min read

Generation grew a timeline

Sora shipped with a storyboard, Re-cut, Remix and Loop. The editing tools are the real launch: a generative tool gets useful when you can change one thing and keep the rest.

Typeface

OpenAI released Sora to ChatGPT Plus and Pro subscribers on December 9. It makes clips of up to 20 seconds at up to 1080p, in widescreen, vertical or square, and it isn’t available yet in the EU, the UK or Switzerland. Plus subscribers get up to 50 videos a month at 480p.

The clips got the attention. The part I found more interesting was the set of tools around them. According to Maginative’s rundown, Sora ships with Storyboard, which gives “frame-by-frame control with a timeline-based interface,” plus Re-cut, Remix, Loop and Blend.

Those are an editor’s verbs. A video generator launched with a timeline.

Most generative tools so far have had one verb: generate. If the result is almost right, you change the prompt and roll again, and the new result is different everywhere, including the parts you liked. The tool can make anything, and it has no way to keep what you liked.

A generative tool gets useful at the moment you can change one thing and keep everything else. Re-cut is that idea for video: keep this shot, extend it, drop the rest. A storyboard is that idea for a sequence: fix the order and the beats, then generate inside them.

I learned how much this matters from a very different tool. On Uber’s experimentation platform, the clearest win after the redesign was rollbacks. They got faster, and they could be scoped to the one segment that was hurting while the rest of the experiment kept running. Being able to change exactly one thing made people willing to change things at all.

People treat creative tools the same way. Someone who can only reroll will stop at the first acceptable result, because trying again risks losing it. Someone who can change one part without breaking the rest keeps going until it’s good.

A timeline doesn’t make the clips better. The model still decides most of what’s in the frame, and a storyboard only constrains it. I’d expect the next generation of these tools to be judged less on what they can generate and more on how precisely they let you change it. That’s where editors have always competed, and I think generative tools are about to find out why.