Score a video or narration with music timed to it
No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.
What counts as done
One instrumental track, or one per segment, at each requested target length, with no vocals unless asked, downloadable as MP3 or WAV, under a licence that covers the platform the video goes to.
Measured, for our own tests: each track is within 1 second of its target length, and the job takes at most 5 minutes of wall time per minute of music. No docs we read state these figures, so they don't count against a listing here.
How agents ask for it: add background music to this video
, make a music bed for my launch video
, score this explainer to the voiceover
, put music under the narration
Does the job, with a limit 11
It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.
| Listing | Why |
|---|---|
| Beatoven.ai API F Beatoven Private Limited (with subsidiary Beatoven Inc.) | Compose returns a task id and then WAV, MP3 or AAC with four stems, and the terms license sync with video. Length goes only in the prompt text, so the bed must be cut or looped, and whether the 2024 terms cover API output isn't stated. source |
| ElevenLabs Music API BB Eleven Labs Inc. | music_length_ms, force_instrumental, a composition plan whose section durations v2 and v2.5 always enforce, and a video-to-music endpoint return MP3 or PCM (acc-01, acc-26). The Music Terms exclude film, TV and radio below Enterprise Music, and wall time isn't stated. source |
| fal music models BB Features & Labels, Inc. | elevenlabs/music/v2.5 on fal takes music_length_ms, force_instrumental and a chunk composition plan behind a queue request id, and Sonilo has video-to-music. The catalogue tags endpoints 'commercial' without saying whether ElevenLabs' film and TV exclusions carry over, and wall time isn't stated. source |
| Google Lyria D Google LLC | MP3 or WAV comes back inline, but length is set only in the prompt or by timestamps (Clip is always 30 seconds) and instrumental only by prompt, so the track must be cut or looped to the cut, the blocker the map names. source |
| Loudly Music API F Loudly GmbH | duration (30 to 420 seconds), genre, BPM, key and preset structures return an instrumental with MP3 and WAV URLs. Nothing under 30 seconds, no timed section plan, and the licence agreement excludes TV broadcast, VOD, cinema and commercial streaming services. source |
| Mubert API F Mubert Inc. | duration with stocked values (5 to 300 seconds) that return instantly, track, loop and jingle modes, an async track id, and a score-a-video MCP prompt. Download links expire after 900 seconds, there's no free API tier, and broadcast terms weren't read. source |
| MusicGPT API D MusicGPT Ltd | make_instrumental, a task id with webhook, MP3 and WAV paths and commercial use on paid plans fit a bed under narration (acc-29), but /v2/MusicAI has no length parameter, so the track must be cut or looped to the cut. source |
| MusicGen on Replicate B Replicate, LLC | An integer duration (default 8 seconds) and WAV or MP3 output fit a bed to a cut, but the weights are CC-BY-NC 4.0, so only non-commercial videos are covered, and there's no section plan or maximum length stated. source |
| SOUNDRAW API F SOUNDRAW Inc. | Compose takes a required length in seconds with mood, genre, theme and tempo, returns a request_id to poll, and links last 72 hours. The only request description is a vendor blog post, m4a is the recommended format, and a contract and token are needed before any call. source |
| Soundverse API E Soundverse Inc. | An instrumental-only generate_music route, a Sync licence tier for film, TV, ads and games, an async job with Idempotency-Key and a signed download link. No duration field on music generation, so the bed must be cut or looped to the cut. source |
| Stable Audio API F Stability AI Ltd (UK, EU, Switzerland) and Stability AI US Services Corporation (elsewhere) | duration (1 to 380 seconds on Stable Audio 3, 190 on 2.x) and MP3 or WAV, an async id on 3.0, and the terms assign output to the user. No instrumental flag or section plan, wall time isn't stated, and acc-14 found arrangement must be described in the prompt. source |
Couldn't check 1
The docs were closed to us, or silent. These carry no mark either way.
What stops agents finishing
- no length parameter, so the track has to be cut or looped to fit the cut
- SDK wrapper missing the video-to-music method (acc-18)
- music API disabled on the free plan
- MCP plugin fails before any API call (acc-17)
Around it
Before: video-generation.render-clip, text-to-speech.long-form-narration. After: video-generation.compose-with-audio, social-media.schedule-post.
Where this comes from
Weight 19 of 100 in Music generation: 8 of 43 account-to-use-case links (38 accounts, five counted in two use cases): HN launch-video and marketing-pipeline posts and GitHub issues that score narration to measured windows
The same as data: /api/v1/use-cases/music-generation.video-score.json, or the MCP tool get_use_case with id music-generation.video-score. Other jobs in this category: /api/v1/niches?category=music-generation.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.