If your product lets users upload songs, they can try different arrangements and vocal styles with the same audio. Your server stores the audio ID, submits a task for each remix direction, then checks the task and sends the finished versions to the player.
Integration
Field | Details |
|---|---|
Endpoint |
|
Model parameter |
|
Audio source | Upload audio with purpose=remix to obtain upload_audio_id, or provide the song_id of a Mureka song generated in the past month. |
Duration | 10–480 s |
Outputs per request | n defaults to 2, maximum 3 |
Price | New version: $0.5 per track (V8 stays at $0.2). |
Documentation |
Results reinterpret the original song, so they do not retain the original singer’s timbre. Specifying a particular singer’s timbre or uploading additional audio as a style reference is currently unsupported.
Model selection: Set model to mureka-9.5 to use Remix V9.5, or to mureka-8 to use V8.
Use cases
- Cover and remix apps: let users turn a dry vocal recording into fully arranged versions in several styles.
- Short video and content platforms: make rock, electronic, or jazz versions of one backing track to match different scenes.
- Musicians and production tools: quickly turn a sung, played, or hummed demo into a full arrangement and explore different styles.
- Games and advertising: create variations of one theme at different intensity levels for different scenes, as in Demo B.
API workflow
Prepare the source audio
This example uses a song with vocals. Call POST /v1/files/upload with purpose set to remix and save the returned file ID. V9.5 accepts MP3 and M4A files from 10 to 480 seconds. Keep the API key on your server and use Bearer authentication.
curl --fail-with-body https://api.mureka.ai/v1/files/upload \
-H "Authorization: Bearer ${MUREKA_API_KEY}" \
-F "purpose=remix" \
-F "file=@./original.mp3"
For a track created through Mureka’s song generation API, you can also use an eligible song_id. Use upload_audio_id for an uploaded file or song_id for an existing song. Choose one input per request and consult the API reference for each path’s limits.
Submit a task for one remix direction
Send the upload ID and a prompt describing the target style to POST /v1/song/remix. The example specifies mureka-9.5 and requests a rock remix with vocals. Replace the placeholders with the returned upload ID and the lyrics to be sung.
curl --fail-with-body https://api.mureka.ai/v1/song/remix \
-H "Authorization: Bearer ${MUREKA_API_KEY}" \
-H "Content-Type: application/json" \
-d '{
"upload_audio_id": "<UPLOAD_AUDIO_ID>",
"model": "mureka-9.5",
"lyrics": "<LYRICS_FOR_THIS_VERSION>",
"prompt": "Keep the chorus melody. Use distorted guitars and live rock drums, with a restrained verse and a bigger chorus.",
"n": 1
}'
Use n to request several candidates with the same prompt. For rock, EDM, and female-lead versions, reuse the upload ID in three requests with different prompts. In your app, associate each task with the original audio ID, the version label, and the task ID so you can match results when they arrive.
Poll the task and present the results
Save the task ID from the submission response and call GET /v1/song/query/{task_id}. While generation is in progress, show a waiting state and keep polling. On success, use the returned audio URL for playback. On failure, log the task ID and error for troubleshooting.
curl --fail-with-body \
"https://api.mureka.ai/v1/song/query/${TASK_ID}" \
-H "Authorization: Bearer ${MUREKA_API_KEY}"
In the player, use the same song title for the original and the remixes and distinguish them by version label. Users can return to the original and always see which interpretation is playing.
API reference