What we're learning from AI music, voice, and video users

AI creative tools don't just need generation. They need workflow.

One thing we've been learning from AI music, voice, and video users is that "usable output" means very different things in each category.

With AI music, users can often keep part of the result — a hook, intro, chorus, or background track. With AI voice, the feedback loop is usually faster: generate, listen, adjust, retry.

AI video is different.

The upside is huge, but the path to a usable result is much less forgiving. If the shot, motion, character, or style is off, it is harder to fix than cutting a song or regenerating a voice.

So the real challenge may not be only better models.

It's helping users get from idea → usable creative output with fewer painful retries.

That's the workflow problem we're thinking about a lot right now.