시작하기
Clip Picker: Long Video to Shorts from Transcript

Clip Picker: Long Video to Shorts from Transcript

Turn a podcast, interview, stream or talk you made into short clips for TikTok, Reels and Shorts. Send its transcript (.srt, .vtt or timestamped text) and get the moments that stand alone with exact times, titles, hooks, on-screen text and captions, plus caption files and a script that cuts each clip to 9:16.
#비디오#소셜 미디어
평점
더 많은 평가가 필요합니다
판매
0
사용 방법
Capafy에서 실행
외부 앱에서도 사용 가능
퍼블리셔 제공
Claude Sonnet 5

The shorts that are already inside your long video

A one-hour conversation usually holds five to ten moments that work on their own: a mistake with a number attached, a story with a turn, a line people will quote. Finding them means scrubbing through the whole thing, and cutting them means fiddling with timecodes.

Clip Picker reads the transcript, finds the windows that stand alone, and hands back everything needed to post them.

What it looks like

clip_map.webp

The clip map from a sample five-minute interview: five clips in time order, each with its times, length and the opening line that hooks.

What you send

  • The transcript with timestamps: .srt or .vtt from YouTube Studio (your video, Subtitles, download), from your podcast host, or from your editor's export. Text with a time at the start of each line also works.
  • (optional) the video file: if it uploads, the clips are cut here
  • (optional) how many clips and what length (default: up to ten, 20-55 seconds)

What you get

  • The clip list: start and end, length, a title, the hook (the speaker's own opening words), on-screen text, a caption, and why it was picked
  • clips.csv, one caption file (.srt) per clip, and clip-map.png
  • cut_clips.sh: one command per clip that cuts it to 1080x1920 with captions burned in, ready to run with ffmpeg on your computer
  • With the video attached: the clips themselves, cut here

How clips are chosen

A clip has to make sense with no earlier context, open on a question, a surprise, a mistake, a number or a turn, end on its payoff, and hold one idea. Intros, sponsor reads and housekeeping are skipped. If the transcript holds fewer good clips than you asked for, it says so - three strong clips beat ten padded ones.

What it does not do

  • Only your own content: links are not downloaded and no platform is logged into.
  • No speech recognition: the transcript's timing is what the cuts follow.
  • The 9:16 crop is a centre crop and does not track faces; check each clip before posting.
  • Hooks and captions use the speaker's words; nothing is added that was not said.