Research: ffmpeg

Evidence only

Research cache for catalog enrichment. Not authoritative for install or trust decisions.

Back to catalog page

Quick Answer

Problem: Probe, trim, transcode, concat, and extract local A/V with ffmpeg/ffprobe. Use when converting or inspecting media files on disk. NOT for AI image generation (draw-thing), UI design (design), or networked downloads (yt-dlp).

Stack / assumptions: ffmpeg and ffprobe on PATH; recipe-only mutations from references/; probe-before-mutate via scripts/probe_media.py and scripts/doctor.py.

Comparable alternative: yt-dlp for URL fetch/transcript; draw-thing for generative local media.

Repo summary:

Repo-owned CLI skill with doctor/probe helpers and approved trim/transcode/thumbnail/GIF recipes. Prefers new output paths over in-place overwrite; redirects download and generative requests to sibling skills.

> Grounded in repository skills/ffmpeg/SKILL.md; treat as evidence, not authority.

Viz / MCP Notes

• MCP media-work: Prefer /ffmpeg for deterministic local transforms when the harness has shell access; MCP docling/fetch remain for document ingestion, not ffmpeg-style edits.
• Wave 9 integration: Skill-first over multimedia MCP servers per integrate-custom-media-pentest-skills design.