
Talking videos edited ~20× faster
In Reelcrew from 18 Oct 2026, at the latest 22 Oct 2026.
Give it a phone clip and get back a trimmed 9:16 short, several shorts from one long talk, or a clean long version with a subtitle file.
- Cuts silences, “uh, um” and repeated sentences on its own
- Karaoke subtitles that light up word by word
- Paper illustration cards and background music at the key points
- figuresThe 19:10 long version renders in 204 seconds (3.4 minutes); the old full edit took 64.5–76.7 minutes for 16:42
- figuresA 5-minute clip becomes a 75-second short in 84 seconds (cut + subtitles), 51 seconds more with 4 illustration cards and music
- figuresMeasured on an i5-10400 + GTX 1660
Cut this video into 3 shorts of 60 seconds, karaoke subtitles, 3 illustration cards in each short
Go deeper
No browser opens to render: it cuts straight on each frame with ffmpeg and the graphics card (NVENC), so a 65-second short with subtitles, cards, sound effects and music is done in about 15 seconds.
Cut points snap to a 30 frames per second grid, so phone clips (variable frame rate) keep their sound in sync after the cut.
Three cut levels: light / medium / strong. The vertical frame follows the face (picked from a still, or face tracking when the speaker moves).
Real limits: no colour grading, no round face frame, no maps. For those, use the full edit (next item).
























