# Media OS

> A Claude Code plugin that routes natural-language requests to production media pipelines. 96 tool-and-technique skills, 13 routed modes, 7 orchestrator agents, 4 safety hooks, 1 dispatch-audit hook, 3 PATH CLIs. Covers FFmpeg complete, OBS Studio, GStreamer, MediaMTX, broadcast IP (NDI, OpenTimelineIO, dovi_tool, hdr10plus_tool, Blackmagic DeckLink, gphoto2), control protocols (MIDI 1.0+2.0 UMP, OSC, DMX512/Art-Net/sACN, VISCA + ONVIF PTZ), system audio routing (PipeWire, JACK, Core Audio, WASAPI), VFX (USD, OpenEXR, OpenImageIO), computer vision (OpenCV, MediaPipe), WebRTC (Pion, mediasoup, LiveKit), and 2026 open-source AI media. Strict Apache-2 / MIT / BSD / GPL license filter on every AI model. MIT licensed.

## Install

- /plugin marketplace add damionrashford/media-os
- /plugin install media-os@media-os

## How dispatch works

User describes a media production task in natural language → router skill auto-loads on intent → reads `modes/_shared.md` + matched `modes/<mode>.md` from disk → composes prompt = shared + mode + user-ask → spawns specialist via Agent() → specialist runs in isolated context, writes artifact to deterministic path, returns summary → SubagentStop hook logs the dispatch.

## The 13 routed modes

- live-production (specialist: live) — OBS + NDI + DeckLink + PTZ + RTMP/SRT/RIST/WHIP
- streaming-distribution (delivery) — HLS / DASH / CMAF / LL-HLS + cbcs DRM + CDN upload
- broadcast-delivery (delivery, approval-gated) — DPP AS-11, Netflix IMF, ProRes, MXF OP1a, IMX50, XDCAM HD422
- editorial-interchange (architect) — Premiere ↔ Resolve ↔ Avid ↔ FCP via FCPXML, AAF, EDL, OTIO
- ai-enhancement (architect) — Real-ESRGAN, SwinIR, HAT upscale; RIFE, FILM interpolate; BiRefNet, rembg, RVM matte; Depth-Anything depth; LivePortrait lipsync
- ai-generation (architect) — FLUX-schnell, Kolors, Sana image; LTX-Video, CogVideoX, Mochi, Wan video; Kokoro, OpenVoice, Piper, StyleTTS2 TTS; Riffusion music
- podcast-pipeline (architect) — record-to-master, script-to-podcast (TTS), existing-to-master; EBU R128 loudness; whisper captions
- vfx-pipeline (architect) — EXR / DPX / USD through ACES + OCIO to ProRes 4444 or JPEG 2000 IMF
- hdr-mastering (hdr) — HDR10, HDR10+, Dolby Vision profiles 5 / 7 / 8.4, HLG, SDR tone-map
- vod-post-production (encoder) — H.264 / H.265 / AV1 / ProRes / DNxHR with VMAF gate
- analysis-quality (qc) — VMAF + SSIM + PSNR + loudness + freeze + black + silence detection
- audio-production (architect) — PipeWire / JACK / Core Audio / WASAPI routing, multitrack mix, DeepFilterNet repair, MIDI/OSC control bridge
- acquisition-archive (probe) — probe-batch, ingest-card, tether-capture (gphoto2), archive-verify

## The 7 specialists

- architect — designs end-to-end pipelines; serves editorial, AI, podcast, VFX, audio modes
- probe — forensic file inspection (color, HDR side-data, GOP, captions, timecode)
- qc — VMAF + SSIM + PSNR + loudness + freeze/black/silence gate
- hdr — HDR10, HDR10+, Dolby Vision, PQ↔HLG, ACES, OCIO
- encoder — rate control, pixel format, container flags, hwaccel
- live — OBS + RTMP/SRT/RIST/WHIP + NDI + DeckLink + PTZ
- delivery — HLS/DASH packaging + cbcs DRM + CDN upload + IMF/MXF

## CLI toolbelt (added to PATH on install)

- moprobe <file> — compact media inspection; --color for HDR/color summary; --json for ffprobe JSON
- moqc --ref <src> --out <encoded> [--vmaf-min N] — VMAF + SSIM + PSNR + loudness + freeze/black/silence gate
- mosafe ffmpeg ... — pre-flight lint of any FFmpeg command (catches missing -movflags +faststart, -sc_threshold 0, aac_adtstoasc, hvc1 vs hev1, zscale sandwich for PQ↔HLG, cbcs vs cenc DRM, in-place overwrites)

## Lifecycle hooks (fire automatically)

- SessionStart — probes installed CLIs + ffmpeg build flags (libvmaf, libzimg, libvidstab, librist, libplacebo, hwaccel)
- UserPromptSubmit — auto-probes any media path named in the prompt
- PreToolUse (Bash) — blocks FFmpeg foot-guns before they fire
- PostToolUse (Bash) — re-ffprobes every FFmpeg output for zero-duration / truncation
- SubagentStop — logs every dispatch (mode, specialist, duration, exit status) to ${MEDIA_WORK_DIR}/modes/dispatch.log

## License filter on AI skills (Layer 9)

Every model shipped is Apache-2 / MIT / BSD / GPL. Restricted models explicitly documented-and-dropped: XTTS-v2, F5-TTS, CodeFormer, DAIN, SVD, Wav2Lip, SadTalker, Surya, FLUX-dev, Meta MusicGen, SDXL / SD3 base.

## Source-of-truth links

- Repository: https://github.com/damionrashford/media-os
- Releases: https://github.com/damionrashford/media-os/releases
- Skills catalog (109): https://github.com/damionrashford/media-os/tree/main/skills
- Modes (13): https://github.com/damionrashford/media-os/tree/main/modes
- Agents (7): https://github.com/damionrashford/media-os/tree/main/agents
- CHANGELOG: https://github.com/damionrashford/media-os/blob/main/CHANGELOG.md
- Contributing: https://github.com/damionrashford/media-os/blob/main/CONTRIBUTING.md
- Modes pattern reference: https://github.com/damionrashford/modes
- Claude Code documentation: https://docs.claude.com/en/docs/claude-code/overview
- Claude Code plugins spec: https://docs.claude.com/en/docs/claude-code/plugins