Cut any interview into scroll-stopping Reels

ClipEmber scores an entire interview for quotable moments and renders each as a captioned 9:16 Reel with a face-aware crop keeping the speaker centered. Per-scene active-speaker switching is not in yet — one speaker stays centered per clip.

Guest dropped a great line? ClipEmber scores the whole conversation, keeps the speaker centered and captions every word for silent scrolling.

Get clips

Open the app to run it — 5 free videos every month, no card required.

1

Drop in the interview

Two-person or panel — paste a link or upload the file.

2

AI ranks the moments

Hot takes and payoffs are surfaced with a hook title for each clip.

3

Restyle and post

Nine caption styles, per-word highlight color, then schedule across platforms.

A real run

5 clips from 15:00 – 18:00 of a real video

Run on 2026-08-17 · source: Mike Tyson Asks Sadhguru Some Hard-hitting Questions by Sadhguru · Creative Commons Attribution (CC BY)

Scores the model gave
8079787675
Time per stage · 57 min 26 s end to end

Not a fair timing: this run was one of six we queued at once and it went last, so most of that hour was spent waiting in the queue. Our quiet-queue runs on this site finish in 4 to 5 minutes.

  • Fetch source3 min 36 s
  • Transcribe3 min 14 s
  • Score moments6 min 29 s
  • Plan the crop15 min 13 s
  • Render 5 clips28 min 54 s

Every clip this run picked landed between 17 and 20 seconds, and the five scores sat inside a 5-point band, 80 down to 75 — the same tight pattern as our sermon run, and the opposite of our Zoom lecture run, where the spread ran 71 down to 52. A conversation where each answer is already a complete thought gives the scorer very little to separate. One limit to know before you rely on this: the crop keeps a single speaker centred for the whole clip. It does not cut back and forth between interviewer and guest inside one clip.

Request 62172c40-dd54-42b2-a2dc-5bcd07eae7c4 — every number above comes from that run, not from an average.

FAQ

Does it follow the active speaker?

Face-aware crop keeps a speaker centered in the 9:16 frame; per-scene speaker tracking is on the roadmap.

Can I change the highlight color?

Yes — pick any color for the spoken word in the editor's Style panel.

What languages work?

Transcription covers 90+ languages; captions keep the original language.