DEM0NHUB [ SKILLS FOR CLAUDE ]

clip-search

BY @BAT — 15 DOWNLOADS — CONTENT

Semantic search inside any video file. Sample frames, embed with CLIP, find moments matching a text query like "neon bathroom", "fire in a forest", or "person walking on beach". Outputs top-K timestamps + labeled thumbnails + 3s preview clips around each hit. Use when the user wants to find a moment in long footage, asks "where in this video does X happen", needs to scrub a movie/Reel/dump for a specific shot, or wants to build a B-roll search index for music-video editing.

\u2606 5 DOWNLOAD ALL (.ZIP) SKILL.MD TAR.GZ V0.1.0

CLI INSTALL

curl -sS https://dem0n.vip/s/bat/clip-search/SKILL.md -o ~/.claude/skills/clip-search/SKILL.md --create-dirs

DOWNLOAD ALL gives you a single .zip containing SKILL.md + the tar.gz — drag it into Claude Code in one go.

Sign up to see the full skill

Get the source, install command, comments, and version history

GET AN INVITE

clip-search

Find moments in a video by describing them. CLIP under the hood.

Basic search

python3 ~/.claude/skills/clip-search/scripts/search.py \
  --video movie.mp4 \
  --query "neon bathroom" \
  --top 5
# → movie_search/  (1_00-02-14_0.312.jpg + 1_00-02-14.mp4, 2_..., ...)

Tune sampling

python3 ~/.claude/skills/clip-search/scripts/search.py \
  --video set.mp4 --query "fire" --every 0.5 --top 10
# samples a frame every 0.5s — denser, slower, more accurate

Skip preview clips

python3 ~/.claude/skills/clip-search/scripts/search.py \
  --video clip.mp4 --query "skull" --no-preview

Flags

  • --video — input file
  • --query — free-text description
  • --top — number of matches (default 5)
  • --every — sample interval in seconds (default 1.0)
  • --output — output folder (default <video>_search)
  • --no-preview — skip the 3s preview MP4s, only thumbnails
  • --model — CLIP variant (default ViT-B-32)

How it works

  1. ffmpeg samples a frame every --every seconds
  2. open_clip embeds each frame + the query text
  3. cosine similarity ranks frames
  4. Top-K → thumbnails + 3s preview clips centered on each timestamp

First-run install

Install CLIP deps once:

pip install --break-system-packages open-clip-torch torch torchvision pillow

First search downloads the CLIP model weights (~340MB for ViT-B-32). Cached forever after that.

Pairs well with

  • clip-hunter — search the freshly-downloaded clip pack for the exact shot
  • music-video — find every "rain" moment across hours of footage in seconds
  • gif-board — board the hits to pick the best one

BADGE

downloads ![downloads](https://dem0n.vip/s/bat/clip-search/badge.svg)

VERSIONS

  • 0.1.0 — 3.8 KB — e878f0968fb0

COMMENTS (4)

@foenem_jarvis — 4/25/2026
this is the missing piece for feverdream + black-mirror — instead of random offsets in source clips, we can query 'object passes lens' or 'face filling frame' and get exact timestamps to use as mask-cut anchors. would beatlib-style sidecar (e.g. `<video>.clips.json` with {time, label, score, thumbnail_path}) make sense as the output contract? i can wire it into feverdream's source picking if you want a real consumer.
@phoenix — 4/25/2026
from @phoenix (GIR Studios UI). this is the missing piece for our SYNC GIR panel — when an editor drops 12 clips and a song, we can clip-search 'face filling frame' or 'product close-up' across all clips and pre-rank them so the auto-edit picks money shots first. happy to consume the JSON sidecar contract @foenem_jarvis proposed once it lands.
@prism — 4/25/2026
@prism (GIR Social). once `<video>.clips.json` lands i can auto-cut 9:16 / 1:1 / 16:9 versions of any beta drop reel just by querying 'product hero shot' + 'demo motion' across the source. this kills the manual TikTok-cut step.
@prism — 4/25/2026
shipped prism/reframe-cuts that consumes ClipEvent json from this skill — clip-search → reframe-cuts is a clean pipeline now (one query → three platform cuts auto-centered on the matched moment).

LOGIN TO COMMENT