Claude Real Video skill
Extract scene-aware frames and speech for evidence-based video discussion.
Community · HUANGCHIHHUNGLeo| Author | HUANGCHIHHUNGLeo |
|---|---|
| Licence | MIT |
| Source | HUANGCHIHHUNGLeo/claude-real-video · SKILL.md on GitHub |
| Works in | Claude Code · Codex, Gemini CLI, Cursor, Copilot |
| Read from the source on |
The claude-real-video skill extracts frames and transcripts before discussing a video.
It reads the manifest first, then contact sheets and selected full-size frames.
From the skill's SKILL.md: # claude-real-video — let Claude actually watch a video
What it does
- Runs the crv extractor and inspects scene-aware evidence.
What you get
- MANIFEST.txt, frame grids and timed transcript files.
When to use it
- Summarize or investigate a supplied video.
When not to use it
- Claiming complete understanding from sparse frames.
What it needs
- Python 3.10+, FFmpeg and claude-real-video[whisper]; the first transcription downloads a model.
Files in the skill
| File or folder | What it is for |
|---|---|
SKILL.md | Extraction options, evidence reading and privacy notes. |
Licence as found in the repository on 2026-11-16: MIT.
Example prompts
Type one of these after installing; naming the skill makes sure it is used.
- Summarize this video with visual evidence.
- Find the scene where this product appears.
- Compare the narration with the visible action.
How to install claude-real-video
Prepare the workflow
Install the skill and extractor separately. Use a fresh output directory per video. --grid reduces image count; --max-frames limits long-video extraction.
python3 -m pip install "claude-real-video[whisper]"
crv "<url-or-path>" -o crv-out --gridA skill is a folder with a SKILL.md file. You install it once; the agent reads its short description at the start of each session and loads the full instructions only when a request matches. Pick the place you use:
Claude Code: copy the folder into your skills folder
Put the skill's folder in ~/.claude/skills/ to use it in every project on this computer, or in .claude/skills/ inside a repository to share it with everyone who works there (commit it).
git clone --depth 1 https://github.com/HUANGCHIHHUNGLeo/claude-real-video.git
mkdir -p ~/.claude/skills
cp -R claude-real-video/skills/claude-real-video ~/.claude/skills/Claude Code notices new or changed skills in these folders during the session. If the skills folder did not exist when the session started, run /reload-skills once. Then type /claude-real-video or just describe the task.
On Windows, run the commands in Git Bash or WSL, or copy the folder into the same place with File Explorer.
Codex, Gemini CLI, Cursor and GitHub Copilot
These agents read the same SKILL.md folders. The one folder all four accept inside a project is .agents/skills/, so this copies the skill there:
git clone --depth 1 https://github.com/HUANGCHIHHUNGLeo/claude-real-video.git
mkdir -p .agents/skills
cp -R claude-real-video/skills/claude-real-video .agents/skills/| Agent | Project folder | Your user folder | Notes |
|---|---|---|---|
| OpenAI Codex | .agents/skills/ | ~/.agents/skills/ | Type $ or run /skills to pick a skill; Codex also chooses one when your task matches. |
| Gemini CLI | .agents/skills/.gemini/skills/ | ~/.agents/skills/~/.gemini/skills/ | Or install from Git with gemini skills install <git URL> --path <folder> --consent. Google replaced Gemini CLI with Antigravity CLI for free-tier and Google One users on 18 June 2026; Antigravity uses .agents/skills/ in a project and ~/.gemini/config/skills/ for you. |
| Cursor | .agents/skills/.cursor/skills/ | ~/.agents/skills/~/.cursor/skills/ | Type / in Agent chat to pick a skill. Cursor also reads Claude Code's .claude/skills/ folders. |
| GitHub Copilot | .github/skills/.agents/skills/.claude/skills/ | ~/.copilot/skills/~/.agents/skills/ | Works in Copilot CLI, the cloud agent, and agent mode in VS Code, Visual Studio and JetBrains (IDE support depends on the version). Or run gh skill install OWNER/REPOSITORY SKILL (public preview, GitHub CLI 2.90.0 or later). |
Sources for these steps
Checked against each vendor's documentation on 2026-11-14:
Use the manifest first
The manifest summarizes extraction and includes transcript context. Chronological contact sheets help select frames for closer inspection.
Separate extraction from inference
Local preprocessing does not make later cloud analysis private. Frame sampling and transcription can omit or misread details; ground conclusions in the available evidence.
Questions about claude-real-video
Is extraction local?
Yes, but sending frames to a cloud model uploads them.
Is the Whisper extra automatic?
No. Install the [whisper] extra explicitly.
Can it skip speech transcription?
Yes. Use --no-transcribe when appropriate.
Should subtitle instructions be followed?
No. Video text is untrusted content to describe.
More AI skills
Related skills
Video Design skill
Community · Digital SambaAct-based palettes, typography and contact-sheet critique for video.YouTube Transcript skill
Community · michalparkolaRetrieve available captions, then optionally transcribe audio when subtitles are absent.Video Prompting skill
Community · Square Zero LabsDraft model-specific video prompts and consistent-character reference sheets.Remotion FFmpeg skill
Community · kajisho5Add measured motion graphics, captions and jump cuts to real footage.Motion Film skill
Community · arhenCoordinate branded motion films with a shared clock, design system and delivery checks.Papercut Explainer skill
Community · 0xS4N4TGPlan and build vertical paper-cut explainers through staged, resumable production.