background-music
Skill136 starsChanged 18 days agoMITView on GitHub
Pick background music from the assets/music/ library and configure auto-ducking (music dips automatically under speech) via meta.json audio.music for the Remotion assembly layer. Read when the brief/edit prompt has a "Nhạc nền" (background music) section set to auto, or when the user asks to add/adjust background music on a video.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- PicsArt/gen-ai-skillstheir own version, first seen Nov 1, 2026
- notivn/AIEVword for word, first seen Oct 9, 2026
Other versions2 projects use a version of this.
Word for word · 1
- notivn/AIEV★ 132100% same
Edited · 0
None found.
Wrote their own version · 1
- PicsArt/gen-ai-skills★ 7same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository20 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- ai-illustrationsSkill
- auto-cutSkill
- color-gradingSkill
- gsapSkill
- hyperframes-cliSkill
- hyperframes-registrySkill
- hyperframesSkill
- key-layoutSkill
- make-a-videoSkill
- noti-tiktok-full-textSkill
- noti-tiktok-vnSkill
- noti-youtube-editSkill
- offline-voiceSkill
- remotion-assembleSkill
- short-form-videoSkill
- skill-authoringSkill
- video-pipelineSkill
- website-to-hyperframesSkill
- webui-designSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Read assets/music/library.json - each entry has file, tags (mood: nang-luong, chill, cam-hung, cang-thang, vui-ve...), durationMs, …
- Very specific: line 63
If in doubt about the levels, measure with ffmpeg -i <draft.mp4> -af volumedetect -f null NUL over a speech segment and a non-speech …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Background Music - music bed + auto-ducking
Background music is played by Remotion in the assembly layer (the MusicTrack component): the track loops, fades in over 0.5s at the start of the video, fades out over 1s at the end, and auto-ducks (smooth 0.4s volume dip) around every speech segment. Claude's job is to pick the track + generate the speech ranges + declare audio.music correctly in meta.json - everything else is deterministic inside Remotion.
1. Pick a track by mood
- Read
assets/music/library.json- each entry hasfile,tags(mood: nang-luong, chill, cam-hung, cang-thang, vui-ve...),durationMs,description. - Pick ONE track whose mood matches the video content (motivational video -> nang-luong/cam-hung; slow explainer -> chill...).
- A track shorter than the video is fine -
MusicTrackloops it automatically, no need to concatenate files with ffmpeg.
2. Generate speech ranges from the transcript
Speech ranges = the segments that DO contain speech (in seconds, on the already-cut composition timeline) - these are the moments the music gets ducked.
- Take the word timestamps from the transcript (already remapped if autoCut ran).
- Merge adjacent words into one continuous range.
- Merge two ranges if the gap between them is < 0.6s - a short breath between two sentences is not worth having the music swell back up and drop again.
- Result:
[[startSec, endSec], ...]- usually just a few ranges for a talking-head video (silent intro, body, outro).
3. Volume levels
| Context | volume | Notes |
Speech present (duckVolume) | 0.10-0.15 | Music sits ~18-20dB below the voice - the voice is ALWAYS clearer than the music |
No speech (volume): intro/outro/pauses | 0.30-0.40 | Music carries the emotion but does not bury the sfx |
Loud / bass-heavy tracks -> take the lower bound of both ranges (0.10 / 0.30).
4. Declare it in meta.json
Copy the music file into the project's assets/ (same as sfx) so the project stays portable - NEVER point directly at the shared library:
Copy-Item "assets/music/<file>" "video-projects/<id>/assets/music/<file>"Then declare it (path relative to the project root):
"audio": {
"voice": "...",
"sfx": [...],
"music": {
"file": "assets/music/<file>",
"volume": 0.35,
"duckVolume": 0.12,
"speech": [[1.2, 14.8], [16.1, 42.5]]
}
}The server stages the file into the Remotion staging directory during assembly - nothing else to do.
5. Verify (mandatory before final)
- Render a draft (assemble-draft) then listen at 3 points: start of the video (fade-in + intro level), middle of the video during speech (music must sit clearly under the voice), end of the video (clean fade-out).
- The only standard: the voice MUST be clearer than the music at every moment. If you have to strain to make out the words -> lower
duckVolume. - If in doubt about the levels, measure with
ffmpeg -i <draft.mp4> -af volumedetect -f null NULover a speech segment and a non-speech segment, and compare the two mean_volume values.
Mistakes to avoid
- Music louder than speech -
duckVolumeis too high, or the speech ranges are missing a segment. Re-check transcript coverage: every spoken segment must fall inside a range. - Forgetting the final fade-out -
MusicTrackfades out over 1s based ondurationInFrames; but if the composition's total duration is computed wrong, the music gets cut off abruptly. Verify the end point when listening to the draft. - Music with Vietnamese lyrics - it fights the speech, the listener's brain cannot separate two layers of words. ONLY use instrumental music.
- Downloading music from the internet - FORBIDDEN (copyright). Only use the
assets/music/library. If the library is empty -> skip background music and state that clearly in the report.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

