Video soundtrack API
Published November 2026
Three jobs over HTTP: one watches the video and returns a cue sheet, one renders the music you direct, one lays the mix on the picture.
A free account has 120 credits a month. The catalog, a quote and the OpenAPI document need no key.
Video in, scored video out
Make a key at Settings → API keys and send it as X-API-Key (or Authorization: Bearer):
export DOO_API_KEY=dsk_live_...
The whole pipeline, as one Python script. It watches the video, renders music for the first cue the video proposes, mixes it and lays it on the picture:
import os, requests
API = "https://api.doseedo.com"
H = {"X-API-Key": os.environ["DOO_API_KEY"]}
def run(kind, input, key=None):
"""Submit a job and long-poll it to the end. Returns the finished job."""
r = requests.post(f"{API}/api/jobs", json={"kind": kind, "input": input},
headers={**H, **({"Idempotency-Key": key} if key else {})})
job = r.json()
if r.status_code >= 400:
raise SystemExit(job["error"])
while job["status"] in ("queued", "running"):
job = requests.get(f"{API}/api/jobs/{job['id']}", headers=H, params={"wait": 50}).json()
if job["status"] != "succeeded":
raise SystemExit(job["error"])
return job
VIDEO = "https://example.com/cut.mp4" # any public https URL (or upload it: see below)
# 1. Watch the video: its cuts, scenes, dialogue and the cues it proposes. No credits.
sheet = run("spot_video", {"video_url": VIDEO})["output"]["cue_sheet"]
cue = sheet["cues"][0]
# 2. The music for that cue. Here a band generator, directed in seconds: it lands the
# cue's hits together. Send your own notes as "parts" instead and they are played as written.
score = run("render_score", {
"bpm": cue["suggested_bpm"],
"generator": {
"id": "jazz", "lead": "none", # the rhythm section alone: a bed under the voices
"seconds": cue["duration_s"],
"hits": [{"at_seconds": h["at_s"], "kind": "medium"} for h in cue["hits"]],
},
})
# 3. Mix the parts. 4. Lay the mix on the picture at the cue's timecode, ducked under the dialogue.
mix = run("mix", {"from_job": score["id"]})["output"]["mix_url"]
film = run("score_to_picture", {
"video_url": VIDEO,
"tracks": [{"audio_url": mix, "start_timecode": cue["start_timecode"], "name": cue["id"]}],
"loudness": "web",
})["output"]
print(film["video"]) # the scored video
print(film["printmaster"]) # the whole mix, 24-bit 48 kHz
Run on November 16, 2026 against a 60-second 1960 television commercial in the public domain, the script finished in 76 seconds. The generator wrote a drums, bass and piano bed for the first cue, each part passed its check against its notes, and the mix was laid at the cue's timecode and dipped 10 dB under the film's seven spoken lines. It ended with the scored video, a printmaster, an M&E and the music alone. Watching the video and laying the score cost no credits; the render was 10 credits.
What the cue sheet holds
spot_video reads the video and returns output.cue_sheet:
cuts,shotsandscenes: every cut to the frame, and where a scene changes.dialogue: the spans where people talk, in seconds and timecode.cues: where music could start and stop. Each has an id (1M1, 1M2 …),start_timecodeandend_timecode,duration_s, asuggested_bpm, thehitsworth landing inside it, and the dialogue it runs under.visual_hitsandaudio_hits: the moments that stand out in the picture and in the film's own sound.
It also returns keyframes and a storyboard image of the whole video, for a model that will write the music. The video is a public https URL, or a file you upload first.
The music is yours to direct
The API does not decide what the music is. For each cue you choose one of two things:
- A band generator. A genre by id, a length in seconds, and direction in seconds from the start of the piece:
dynamics(how loud the band is at each moment, and whether it gets there gradually),hits(a moment the band lands together: light, medium or full) andlay_out(who rests where, as under a line of dialogue). The script above lands the cue's hits. - Your own notes. Send
partswritten by you or by your model, and they are played as written, each part its own stem. This is how an AI composer scores a film through this API: it reads the cue sheet and the keyframes, writes every part, and renders them. See scoring a film with an AI agent.
If you want the composing done for you, with no code, that is the scoring studio in the browser: upload a video and an AI composer writes, plays and mixes the score.
Laying it on the picture
score_to_picture takes the video and a list of tracks, each a mixed cue with the timecode it starts at:
{
"kind": "score_to_picture",
"input": {
"video_url": "https://example.com/cut.mp4",
"loudness": "web",
"tracks": [
{ "audio_url": "<the mix_url>", "start_timecode": "00:00:00:01", "name": "1M1" }
]
}
}
- Ducked under the dialogue. Music tracks dip under each spoken line; set the depth and the timing in
duck, or turn it off for one track. - The film's own sound is kept under the score by default, or dropped.
- Delivery loudness:
web(−16 LUFS),streaming(−14),broadcast(−23), or a number. - Loops and trims. A track can start part-way into its audio, play for a set length, fade, or repeat every so many seconds.
It returns video (the picture copied untouched, with the new soundtrack), printmaster (the whole mix), music_and_effects and score (the music alone), each a pre-signed link. Up to 25 minutes of film a call.
Uploading a video
A video that is not at a public URL goes up in two calls, and its key is passed as video_key:
# 1. Ask for an upload slot
curl -s -X POST https://api.doseedo.com/api/jobs/uploads \
-H "X-API-Key: $DOO_API_KEY" -H 'Content-Type: application/json' \
-d '{"filename":"cut.mp4"}'
# → { "key": "ingest/…", "upload_url": "https://…", "headers": { "Content-Type": "audio/mp4" }, "input_param": "video_key" }
# 2. PUT the bytes to upload_url, with that Content-Type
curl -s -X PUT -H 'Content-Type: audio/mp4' --data-binary @cut.mp4 "$UPLOAD_URL"
# 3. Pass the key to the job as input.video_key
curl -s -X POST https://api.doseedo.com/api/jobs \
-H "X-API-Key: $DOO_API_KEY" -H 'Content-Type: application/json' \
-d '{"kind":"spot_video","input":{"video_key":"ingest/…"}}'
MP4, MOV, M4V, WebM or MKV, up to 2 GB. The same key is passed to score_to_picture.
What it costs
Watching the video (spot_video), mixing (mix) and laying the score (score_to_picture) cost no credits. The performances do: 10 credits for each 30 seconds of music for each group of up to six parts.
One credit pool pays for every job. A free account has 120 credits a month; Pro is $12 a month with 2,000, plus one-time extra usage (500 credits for $5). The quote endpoint prices a call before you make it, a refused call spends nothing, and a failed job is refunded.
On a free account a render of several parts comes back as its mix: each part as its own stem comes with Pro. A one-part render, a separation and every analysis job are the same on every plan.
Limits
- Not one call. No single request turns a video into a finished soundtrack: the API gives your code the cue sheet, the players and the lay-back, and the music is your notes or a generator you direct.
- Asynchronous. Watching a video usually takes 20 s to 10 min, by the catalog, depending on its length.
- Up to 25 minutes of film a call; a longer film is delivered reel by reel.
- A performance, not a playback. A rendered part can miss or change a note, so each part of a score comes back with a check against its notes.
A video soundtrack API
Video soundtrack API, video to music API, music for video API and score to picture API name the same pipeline: a video is watched, music is rendered for its cues, and the mix is laid back on the picture.
Score a video from your code
Make a key and run the script above against your own cut. Watching the video and laying the score cost no credits.
Get an API key →FAQ
Is there an API that adds music to a video?
Yes. On Doseedo it is three jobs: spot_video watches the video and returns a cue sheet with timecodes, render_score renders the music you direct for each cue, and score_to_picture lays the mix on the picture under the dialogue and returns the scored video.
Does the API compose the music from the video by itself?
No. It tells your code what is in the video and where music could go, and it plays and lays the music. The music is either a band generator you direct in seconds or notes you (or your model) write. For a soundtrack composed for you with no code, use the scoring studio in the browser.
Does the music land on the cuts?
It can: the cue sheet gives every cut and each cue's hits in seconds, and a band generator takes hits at those seconds and lands them together. Notes you write land where you write them.
What comes back?
The scored video with the picture copied untouched, a printmaster of the whole mix, a music-and-effects mix without the dialogue, and the music alone. Each part of the score is also its own stem.
How much does it cost?
Watching the video, mixing and laying the score cost no credits. The performances cost 10 credits for each 30 seconds of music for each group of up to six parts. A free account has 120 credits a month.
The music API · music your code controls
- MIDI to audio APIyour notes, played
- Video soundtrack APIspot, render, lay on the picture
- Stem separation APIa mix in, stems and MIDI out
- Game music generatorseamless loops, in layers
- The MCP serverthe same jobs for an AI agent
- Every toolthe reference
- Planscredits a month