메뉴
FFmpeg media processing — video/audio conversion, trimming, encoding, filters, streaming. CLI and ffmpeg.wasm patterns.
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/Tibsfox/gsd-skill-creator --skill ffmpeg-media명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
Use the moment a request is underspecified and you are about to act on an ASSUMED goal. Enumerate the candidate goals the request could mean, project the action or artifact each one implies, and check whether they diverge: if two or more plausible goals lead to materially different actions or artifacts, ask exactly ONE targeted clarifying question; otherwise proceed on the most-likely goal and state the assumption in one line. Clarification is an evidence-producing action, not a delay — but one question, never an interrogation. Distinct from gsd-spec-phase (GSD-phase-bound, emits a heavy SPEC.md) and intent-router (fetch strategy, not goal disambiguation). Backed by agent goal-state inference (arxiv 2606.16813v1). Triggers on acting under an assumed goal when the request admits more than one materially different reading.
Use at write time to vector/embedding memory — Grove content-addressed insertion, chroma/pgvector upserts, memory-consolidation promoting session traces to MEMORY.md, or embedding externally-ingested content. Scores each candidate record against a fixed panel of sentinel queries and quarantines any record that would become the nearest neighbor of too many unrelated queries — a hub — whether from adversarial poisoning or accidental over-generality. This is the memory-record-side sibling of skill-injection-guardian (file-side) and the write-side complement of memory-use-warrant (read-side). Quarantine, never silently drop; a human reviews. Backed by the admission-time hubness gate (arxiv 2606.19692v1). Triggers on inserting into vector memory, consolidating memory, and embedding stranger content.
Run this appropriateness check the moment you are about to integrate a retrieved long-term memory — a Grove content-addressed hit, a chroma/pgvector neighbour, a memory-consolidation digest, or a MEMORY.md line — into a response, especially anything touching private origins, Fox Companies IP, credentials, or Center Camp / consent-governed content. It answers a question intent-router never asks: not WHAT to fetch or HOW, but WHETHER a correctly-retrieved item should reach output. Relevance is not appropriateness — a perfect similarity match can still be a boundary violation. Default is FAIL-CLOSED: if in-context authorization is uncertain, the memory may inform behaviour but must not be surfaced. Backed by RBI-Eval (arxiv 2606.06055v1). Triggers on surfacing recalled sensitive memory into a response.
SKILL.md 표시 중
SOC 직업 분류 기준
| name | ffmpeg-media |
| description | FFmpeg media processing — video/audio conversion, trimming, encoding, filters, streaming. CLI and ffmpeg.wasm patterns. |
| activationKeywords | ["ffmpeg","video","audio","transcode","encode","codec","h264","h265","hevc","webm","mp4","mkv","wav","mp3","aac","opus","thumbnail","waveform","concat","trim","crop","scale","filter","subtitle","hls","dash","streaming"] |
| type | skill |
| category | media |
| status | stable |
| origin | tibsfox |
| modified | false |
| first_seen | "2026-04-11T00:00:00.000Z" |
| first_path | examples/skills/media/ffmpeg-media/SKILL.md |
| superseded_by | null |
Dedicated to sudokku, creator of micro-ffmpeg — the in-browser video editor that proved ffmpeg.wasm is production-ready. Built in 2 days, 2,489 LOC, 93 commits. The patterns in this skill are drawn from that work.
Expert-level ffmpeg command construction, media pipeline design, and ffmpeg.wasm integration patterns. Covers CLI ffmpeg, Node.js ffmpeg bindings, and browser-side ffmpeg.wasm.
Build the filter graph first, then wrap it in the ffmpeg command. Never guess at flags — construct from codec specs.
# Full probe (JSON output)
ffprobe -v quiet -print_format json -show_format -show_streams input.mp4
# Quick summary
ffprobe -v error -show_entries format=duration,bit_rate -show_entries stream=codec_name,width,height,r_frame_rate input.mp4
# MP4 (H.264 + AAC) — universal compatibility
ffmpeg -i input.mkv -c:v libx264 -preset medium -crf 23 -c:a aac -b:a 128k output.mp4
# WebM (VP9 + Opus) — web-optimized, smaller files
ffmpeg -i input.mp4 -c:v libvpx-vp9 -crf 30 -b:v 0 -c:a libopus -b:a 128k output.webm
# H.265/HEVC — 50% smaller than H.264 at same quality
ffmpeg -i input.mp4 -c:v libx265 -preset medium -crf 28 -c:a aac -b:a 128k output.mp4
# AV1 — best compression, slow encoding
ffmpeg -i input.mp4 -c:v libaom-av1 -crf 30 -b:v 0 -c:a libopus output.mkv
# Audio only (MP3)
ffmpeg -i input.mp4 -vn -c:a libmp3lame -q:a 2 output.mp3
# Audio only (high quality FLAC)
ffmpeg -i input.wav -c:a flac output.flac
# Trim from 00:01:30 for 60 seconds (fast, no re-encode)
ffmpeg -ss 00:01:30 -i input.mp4 -t 60 -c copy output.mp4
# Trim with re-encode (frame-accurate)
ffmpeg -i input.mp4 -ss 00:01:30 -to 00:02:30 -c:v libx264 -c:a aac output.mp4
# Scale to 1280x720 (maintain aspect ratio)
ffmpeg -i input.mp4 -vf "scale=1280:720:force_original_aspect_ratio=decrease,pad=1280:720:-1:-1" output.mp4
# Scale to 50%
ffmpeg -i input.mp4 -vf output.mp4
ffmpeg -i input.mp4 -vf output.mp4
# Blur
ffmpeg -i input.mp4 -vf "boxblur=5:1" output.mp4
# Brightness/Contrast/Saturation
ffmpeg -i input.mp4 -vf "eq=brightness=0.1:contrast=1.2:saturation=1.3" output.mp4
# Speed up 2x (video + audio)
ffmpeg -i input.mp4 -vf "setpts=0.5*PTS" -af "atempo=2.0" output.mp4
# Slow down 0.5x
ffmpeg -i input.mp4 -vf "setpts=2.0*PTS" -af "atempo=0.5" output.mp4
# Rotate 90 degrees clockwise
ffmpeg -i input.mp4 -vf "transpose=1" output.mp4
# Crop (center crop to 640x480)
ffmpeg -i input.mp4 -vf "crop=640:480" output.mp4
# Overlay watermark
ffmpeg -i input.mp4 -i watermark.png -filter_complex "overlay=W-w-10:H-h-10" output.mp4
# Text overlay
ffmpeg -i input.mp4 -vf "drawtext=text='Title':fontsize=48:fontcolor=white:x=(w-text_w)/2:y=50" output.mp4
# Create file list
echo "file 'clip1.mp4'" > list.txt
echo "file 'clip2.mp4'" >> list.txt
echo "file 'clip3.mp4'" >> list.txt
# Concat (same codec — fast)
ffmpeg -f concat -safe 0 -i list.txt -c copy output.mp4
# Concat (different codecs — re-encode)
ffmpeg -f concat -safe 0 -i list.txt -c:v libx264 -c:a aac output.mp4
# Single frame at 5 seconds
ffmpeg -i input.mp4 -ss 00:00:05 -frames:v 1 thumb.jpg
# Thumbnail grid (4x4)
ffmpeg -i input.mp4 -vf "select=not(mod(n\,300)),scale=320:180,tile=4x4" -frames:v 1 grid.jpg
# Extract frame every N seconds
ffmpeg -i input.mp4 -vf "fps=1/10" frames/frame_%04d.jpg
# PNG waveform
ffmpeg -i input.mp3 -filter_complex "showwavespic=s=1920x200:colors=white" -frames:v 1 waveform.png
# Video waveform
ffmpeg -i input.mp3 -filter_complex "showwaves=s=1280x720:mode=line:colors=white" waveform.mp4
# HLS output (adaptive bitrate)
ffmpeg -i input.mp4 -c:v libx264 -c:a aac \
-hls_time 6 -hls_playlist_type vod \
-hls_segment_filename "segment_%03d.ts" playlist.m3u8
# Multiple quality HLS
ffmpeg -i input.mp4 \
-map 0:v -map 0:a -map 0:v -map 0:a \
-c:v libx264 -c:a aac \
-b:v:0 5M -s:v:0 1920x1080 \
-b:v:1 1M -s:v:1 640x360 \
-var_stream_map "v:0,a:0 v:1,a:1" \
-hls_time 6 -hls_playlist_type vod \
-master_pl_name master.m3u8 \
-hls_segment_filename "v%v/seg_%03d.ts" "v%v/playlist.m3u8"
# Burn subtitles into video
ffmpeg -i input.mp4 -vf "subtitles=subs.srt" output.mp4
# Extract subtitles
ffmpeg -i input.mkv -map 0:s:0 subs.srt
// ffmpegSingleton.ts — main-thread singleton (no Comlink)
import { FFmpeg } from '@ffmpeg/ffmpeg';
import { fetchFile, toBlobURL } from '@ffmpeg/util';
let ffmpeg: FFmpeg | null = null;
export async function getFFmpeg(): Promise<FFmpeg> {
if (ffmpeg) return ffmpeg;
ffmpeg = new FFmpeg();
await ffmpeg.load({
coreURL: await toBlobURL('/ffmpeg-core.js', 'text/javascript'),
wasmURL: await toBlobURL('/ffmpeg-core.wasm', 'application/wasm'),
});
return ffmpeg;
}
// Usage
const ff = await getFFmpeg();
await ff.writeFile('input.mp4', await fetchFile(file));
await ff.exec(['-i', 'input.mp4', '-vf', 'scale=640:360', 'output.mp4']);
const data = await ff.readFile('output.mp4');
ff.on('progress', ({ progress, time }) => {
console.log(`${Math.round(progress * 100)}% (${time}μs)`);
});
async function extractThumbnail(file: File, timeSeconds: number): Promise<Blob> {
const ff = await getFFmpeg();
await ff.writeFile('input', await fetchFile(file));
await ff.exec([
'-i', 'input',
'-ss', String(timeSeconds),
'-frames:v', '1',
'-f', 'image2',
'thumb.jpg'
]);
const data = await ff.readFile('thumb.jpg');
return new Blob([data], { type: 'image/jpeg' });
}
| Codec | Use Case | Quality/Size | Speed |
|---|---|---|---|
| H.264 (libx264) | Universal playback | Good/Medium | Fast |
| H.265 (libx265) | Storage, streaming | Better/Small | Slow |
| VP9 (libvpx-vp9) | Web delivery | Better/Small | Slow |
| AV1 (libaom-av1) | Archive, future web | Best/Smallest | Very slow |
| ProRes (prores_ks) | Editing intermediate | Lossless/Huge | Fast |
| AAC | Universal audio | Good | Fast |
| Opus | Web audio, VoIP | Better at low bitrates | Fast |
| FLAC | Lossless audio | Lossless/Large | Fast |
| CRF | Quality | Use Case |
|---|---|---|
| 0 | Lossless | Archival |
| 18 | Visually lossless | High-quality master |
| 23 | Default | Good balance |
| 28 | Lower quality | Web/mobile delivery |
| 35+ | Very compressed | Thumbnails, previews |
| Mistake | Fix |
|---|---|
-ss after -i (slow seek) | Put -ss before -i for fast seek |
Missing -movflags +faststart | Add for web-playable MP4 |
Using -b:v with CRF | CRF and bitrate are mutually exclusive |
Forgetting -pix_fmt yuv420p | Required for broad compatibility |
| Concat without same codecs | Re-encode or use filter_complex |
| Not setting thread count | Add -threads 0 for auto |
The gource-visualizer skill (skills/gource-visualizer/) produces PPM frame streams that pipe through ffmpeg. Key patterns:
# Standard Gource → ffmpeg pipeline (headless with xvfb)
xvfb-run -a gource --output-ppm-stream - \
--viewport 1920x1080 -s 0.5 --auto-skip-seconds 1 \
--stop-at-end /path/to/repo | \
ffmpeg -y -r 60 -f image2pipe -vcodec ppm -i - \
-vcodec libx264 -preset medium -crf 18 \
-pix_fmt yuv420p -movflags +faststart output.mp4
# Cinematic preset (from skills/gource-visualizer/configs/preset-cinematic.conf)
xvfb-run -a gource --load-config skills/gource-visualizer/configs/preset-cinematic.conf \
--output-ppm-stream - /path/to/repo | \
ffmpeg -y -r 60 -f image2pipe -vcodec ppm -i - \
-vcodec libx264 -preset slow -crf 15 \
-pix_fmt yuv420p -movflags +faststart cinematic.mp4
# Multi-repo combined visualization
gource --output-custom-log - /repo1 | sed "s|/|/repo1/|2" > /tmp/combined.log
gource --output-custom-log - /repo2 | sed "s|/|/repo2/|2" >> /tmp/combined.log
sort -t\| -k1,1 /tmp/combined.log > /tmp/sorted.log
xvfb-run -a gource /tmp/sorted.log --output-ppm-stream - | \
ffmpeg -y -r 60 -f image2pipe -vcodec ppm -i - \
-vcodec libx264 -preset medium -crf 18 -pix_fmt yuv420p output.mp4
Gource agents: skills/gource-visualizer/agents/ — installer, log-generator, renderer, deliverer.
Presets: quick (preview), standard (demo), cinematic (publication), thumbnail (still frame).