Sashimii
A rough cut is only half the edit. The other half is everything on screen: a title here, a zoom there, a graphic that explains the point you just made. Now you can ask for all of it while you watch, and your AI agent builds it.
Why We Built It
Directing graphics usually means writing a list, switching to another tool, building each one by hand, then checking it against the video and starting over when it’s off. We wanted that to happen in the same pass as the cut: watch once, say what you want where you want it, and review the results right where they play.
What It Does
- Notes on the words themselves. Select some words, press N and say what should happen there. Tag it as text on screen, B-roll, a zoom, picture-in-picture, a chapter or a custom graphic. Dictation works too.
- Your agent picks them up. Sashimii works with the AI agent you already use, such as Claude Code or Codex. It reads your open notes and makes the graphics, animations and captions they ask for.
- Review them in place. Each graphic shows up on its own lane above the cut, at the moment it belongs. Play it back over your video, drag it to move or retime it, and press A to approve it.
- Send feedback, not instructions from scratch. If a graphic isn’t right, leave a note on it. The note goes back to your agent with the graphic attached, and the new version replaces the old one on the lane.
You stay the director. The agent does the building, and every graphic stays editable until you say it’s done.