Transcript · moment picking · framing on the speaker
What was wrong
Long videos — podcasts, interviews, talks — run for hours, but people watch them in Reels, TikTok and Shorts, 30–60 seconds at a time. Cutting them by hand is an evening in an editor for every recording: find the strong moments, reframe for 9:16, caption every word.
The start: drop a video, choose a file or paste a link.
What it does
Cloth takes a file or a YouTube link and hands back 3–20 vertical shorts. A model reads the transcript and suggests moments that end on a finished thought; the frame follows whoever is speaking, and the subtitles run word by word, karaoke-style, inside the safe zones of all three platforms. Every run stays in the library together with its transcript.
A finished run of an interview: three 9:16 clips with subtitles, the strongest marked «Post first».
A file (MP4, MOV) or a YouTube link — from a link it downloads only what goes into the clips
A model picks the moments from the transcript; tell it what to look for, or pick them yourself
Framing on whoever is speaking, and a subtitle colour per voice — the Mac itself recognises the speakers
The strongest clip is marked «Post first»
Pauses and «um»s cut with one switch
Settings: subtitles, speaker recognition and voice colours, pause trimming.