Two ways to export. An overlay is just the animation on a transparent background — you drop it over your video in your editor, so the video itself is never re-encoded and loses no quality. A composite burns the overlay onto a copy of your video here.
Drop your draft video here or
The video is transcribed (Whisper) so chapter timings can be estimated and AI can suggest chapters.
How do you want to get your chapter list?
Timings are estimated from the transcript automatically as you
write titles. “@” sets a field to the current playhead. Drag the markers
on the timeline for coarse moves; type exact times (m:ss.s) for precision.
Keyboard: Space plays/pauses, ←/→ steps a frame (Shift = 1 s), Ctrl+Z undoes.
How the intro rows arrive — “default” is the reference slide-in from the right. Pick one chapter above to give just that row its own entrance. The chapter-start rebuild stays a shared move; tune it under “Timing & animation”.
How do you want to get your build lists?
Suggested build lists — enumerations the episode walks through step by step. Add the ones you want.
A build list is a full-screen card that returns each time you reach the next point; earlier items sit static while the new one slides in (like the “five practical tips”). Each item's time = when its card cuts in; “hold” sets how long that step stays up (blank = the default). Drag an item's number badge to reorder the list.
How each item lands on the card — applies to every item of the selected list (or of all lists). “Default” is the reference slide-in from the right; the card itself hard-cuts in and out.
How do you want to get your chapter cards?
Suggested cards — the standout beats worth showing full-screen. Add the ones you want.
A translucent full-frame title over the footage (hard cut in/out, like the Met Gala video).
Entrance and exit for the full-frame card. “Default” is the reference hard cut in and out.
How do you want to get your definitions?
Suggested terms — ✕ any you don't want, then add. Placement times come from where the video explains each concept.
A big keyword lower-third over the footage (no tint), with an optional smaller definition line beneath — like “SIGNALING” in the Met Gala video.
Entrance and exit for the keyword lower-third. “Default” is the reference 0.3 s fade in and hard cut out. A slide from the bottom pairs well with its bottom anchoring.
Upload a chart image (usually a PNG) — or drop one on this panel, or straight onto the video frame where you want it to sit — and it plays opaquely over the footage at the time you set. Position, size and opacity are in the Style tab. Drag its marker on the timeline, or press “@” to stamp the current playhead.
Entrance and exit for the chart (its backing panel moves with it). “Default” is a gentle 0.3 s fade in and out.
How do you want to get your questions?
Suggested questions — ✕ any you don't want, then add. Each appears as the interviewer gets to the heart of the question and hides shortly after the answer starts.
This transcript has no speaker labels (it predates diarization, or transcription ran without an OpenAI key). AI can still find questions from the wording alone, but telling interviewer from interviewee is less reliable — re-run transcription via ⟳ Replace video for a diarized transcript.
A full-width band that slides up from the bottom with a short, direct version of the interviewer's question — for Q&A episodes. “at” is when the band appears; set “answer at” (or let AI find it) and the band hides a few seconds after the answer starts, else it stays up for the default hold.
Entrance and exit for the question band. “Default” is a 0.5 s slide up from the bottom edge and a 0.35 s slide back out.
A speaker lower third on a surf panel, bottom-left: a big name line and a smaller title line beneath. The letters decode in with a scatter-and-blur, the panel sweeps in behind them, and it all dissolves back out — the burrito-gate look. New rows start as the reference lower third (name, title, appearing at 0:05.2 for 5.1 s) — edit any of it. “at” is when it appears; “hold” overrides the default on-screen time.
“Default” is the signature decode — letters scatter in with a motion smear while the panel sweeps in from the right, then it all dissolves back out. The durations stretch or tighten that whole choreography (name, panel and title keep their relative stagger). Pick cut, fade, slide or wipe instead to replace the decode on either side.
Paste your chapter list in any shape — numbering, timestamps and stray notes are fine. AI tidies it into clean chapter titles (the on-screen list draws its own numbers).
Edits apply to everyone using this instance and take effect on the
next AI call. Everything here is git-backed: saving commits the prompt texts to
prompts/*.md, the model choice to prompts/ai-settings.json,
and your saved style defaults (cog → other tabs) to
prompts/style-defaults.json in the app's GitHub repo — full history +
rollback; you can also edit those files directly and push. “Reset all” restores the
shipped defaults.
Default style for overlays in new videos. The example below re-renders as you edit; the handwritten labels name each part, and hovering a heading below points its part out.