# tiktok-video-review

Quick utility skill to review TikTok videos by sampling evenly distributed frames across the full video timeline.

## What it does

- Downloads a TikTok video with `yt-dlp`
- Measures duration with `ffprobe`
- Extracts evenly spaced frames with `ffmpeg`
- Lets you quickly inspect the visual flow without full playback

Default sampling count is `14` frames.

## Files

- `SKILL.md` - skill instructions for OpenClaw/Codex
- `scripts/extract-frames.sh` - frame extraction helper script
- `_meta.json` - skill metadata

## Requirements

- `yt-dlp`
- `ffprobe` (from FFmpeg)
- `ffmpeg`
- `python3`

## Usage

```bash
# 1) Download a TikTok video
yt-dlp -o /tmp/tiktok_review.mp4 "<tiktok-url>"

# 2) Extract 14 evenly distributed frames
./scripts/extract-frames.sh \
  --input /tmp/tiktok_review.mp4 \
  --outdir /tmp/tiktok_review_frames \
  --count 14
```

The script outputs files like:

- `/tmp/tiktok_review_frames/frame_01.jpg`
- ...
- `/tmp/tiktok_review_frames/frame_14.jpg`

## Notes

- Sampling is centered and uniform across the entire duration.
- For very short videos, near-duplicate frames are expected.
- Any summary from this workflow is based on sampled frames, not every frame of the video.