Video to text. Contact sheet, frames, transcript.
Turn a video or audio URL into sampled frames, a contact sheet, and a timestamped transcript your agent can read.
Compatible with
Built for
What it does.
A text model cannot watch video. Peek downloads media, samples frames at a fixed rate, transcribes audio locally with Whisper, and hands your agent a manifest with exact timestamps so it cites what it actually saw and heard.
Processing stays local: yt-dlp downloads, ffprobe probes, ffmpeg samples, faster-whisper transcribes. No API key. No upload. No cost per video. Run directories contain a contact sheet, individual frames, transcript segments, and manifest.json that joins them all.
Works on Instagram reels, TikTok, YouTube Shorts, X, Facebook, voice memos, and local files. Good for hooks, on-screen text, captions, CTAs, and spoken claims. Not for motion, timing, or music cues.
What’s inside.
Everything you need in the bundle. No tutorials. No assembly.
- 01CLI that processes any video or audio URL or local file without uploading
- 02Contact sheet showing the entire clip at a glance
- 03Individual sampled frames with exact presentation timestamps
- 04Timestamped transcript segments with language detection
- 05manifest.json schema that joins frames and transcript so your agent cites the right moment
- 06Flags to adjust frame rate, model size, duration caps, and browser cookie access for private content
- 07Portable skill for Claude Code, Cursor, Windsurf, and Agent Plugins hosts
Drop it in. Press go.
- №01
Download
Buy once, get the file by email and on your dashboard.
- №02
Drop it in
Paste it into your agent. One minute.
- №03
Run it
Your agent picks up the new capability immediately.
- №04
Your agent just got smarter
One more thing off your plate — for every future job.
About the creator.
Things people ask.
$10
Video to text. Contact sheet, frames, transcript.