Ootto-AI

Ootto Watch — Development skill for Claude Code

Development community

Let Claude actually watch a video — frames + transcript, aligned on one timeline.

How to install Ootto Watch

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Ootto-AI/ootto-watch and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Ootto Watch does

Let Claude actually watch a video — frames + transcript, aligned on one timeline. A Claude skill.

Alternatives in Development

Brainstorm OKRs — Brainstorm team OKRs aligned with company objectives 7.8k ★
Setup Video Vision — Interactive setup wizard for claude-video-vision — configure backend, whisper, frames, and verify dependencies 1.3k ★
07 Timeline Gap Analysis — Timeline Gap Analysis 1.1k ★

README

ootto-watch 👀 — let Claude actually watch a video

A Claude skill that gives Claude **eyes + ears** on any online video. Most tools only read the transcript and miss everything visual — this hands Claude the **frames and the words, aligned on one timeline**, so it can reason about cuts, on-screen text, demos, hooks, and pacing.

Paste a link → it downloads the video, samples key frames, transcribes the audio, and Claude reads both — then gives you a **full breakdown**: hook, format, storytelling structure, pacing and visual layout. Pass several links to **batch-analyze dozens at once** and surface the patterns the best ones share.

Install

**Claude Code** (drop the skill in):

git clone https://github.com/Ootto-AI/ootto-watch ~/.claude/skills/watch-video

**claude.ai (web):** zip this folder and add it under Settings → Capabilities → Skills.

Then just tell Claude: *"watch this video: "*.

How it works

Download — yt-dlp pulls the video (+ native captions if present) from YouTube, Instagram, X, Vimeo, and 1000+ sites.
Frames — ffmpeg samples key frames, duration-aware (a 30s clip gets ~30 frames; a 30-min lecture stays sparse) and scales them down for fast vision.
Transcript — native captions when available (free); otherwise local faster-whisper (also free, CPU). --no-whisper to skip.
Watch — Claude reads the frames as images alongside the timestamped transcript (index.json) and answers grounded in both.

Requirements

yt-dlp and ffmpeg/ffprobe on PATH
(optional) pip install faster-whisper — only needed for videos without captions

Analyze & batch

Analyze — after watching, Claude writes a structured breakdown (hook · format · structure · pacing · visual layout), every point grounded in a frame timestamp + transcript line. See SKILL.md → Analyze.
Batch — pass several URLs to watch them all, then compare across them for shared hook shapes, beat lengths and visual moves —