Introduction
The Song and Background Music endpoint accepts a style prompt of up to 1,024 characters and lyrics of up to 5,000 characters. Use instrumental without lyrics for backing music, choose the output file format, and set how many versions to receive.
| Field | Value |
|---|---|
| Input | Text |
| Output | Audio |
| Style input | prompt: up to 1,024 characters |
| Lyric input | lyrics: up to 5,000 characters, including section tags |
| Version count |
|
| File choices | mp3, wav, flac |
| Publisher-stated lengths | Mureka says about 5:30 for a song and about 4:30 for background music |
| Observed folk result | Approximately 2:31, acoustic guitar and warm male vocal |
| Billing quantity | Each V9 version is one billable song; instrumental and vocal modes share the rate |
Key features & highlights
Warm male vocals on a slow R&B track
The supplied verse and chorus became a slow R&B song about wanting a late night to last. The returned MP3 carries a warm male voice singing those words; audition it as a concrete example of this musical direction.
A folk song, 2 minutes 31 seconds
We supplied verses and a chorus with an acoustic-guitar brief. Mureka V9 returned a 2:31 track with a warm male voice singing the given lines.
A lo-fi backing track for study
With instrumental switched on and no lyrics, the same model returned a slow lo-fi beat with piano for background use.
Request parameters
| Parameter | Required | Description |
|---|---|---|
prompt | Yes | Give musical direction in up to 1,024 characters: genre, instrumentation, mood, tempo and voice; the description does not change the per-song rate. at least 1 characters · at most 1024 characters |
lyrics | No | Supply up to 5,000 characters of sung words, using labels such as [Verse] and [Chorus]; omit lyrics for instrumental music, with no separate lyric charge. at least 1 characters · at most 5000 characters |
instrumental | No | Turn on to create background music from prompt without lyrics or vocals; songs and background music use the same V9 rate. Default: |
num_outputs | No | Set the number of V9 versions to 1–3; each version is charged, so requesting more versions increases the bill. |
output_format | No | Choose mp3, wav or flac as the returned audio format; V9 billing counts songs rather than format choices. |
Pricing
$0.084 / song
V9 bills each version requested through num_outputs at the same song or background-music rate; failed requests are free, and V9 costs less per song than V9.5.
Use cases
Hear a lyric draft performed
Turn a written verse and chorus into a vocal demo before revising the words. Our acoustic folk test returned a track of about 2:31 with a warm male voice singing the supplied lines.
Explore an intimate R&B mood
Combine a relaxed groove with vocal direction and your own late-night lyric. The new example returned a slow R&B MP3 with a warm male voice performing the supplied verse and chorus.
Build a quiet study soundtrack
Use instrumental mode for a backing track that does not need sung words. Our study-music example produced a slow lo-fi beat with piano and vinyl crackle; listen to your result in its intended setting.
Compare takes of the same brief
Request two or three versions when you want candidates to audition together. Keep the lyric and musical direction unchanged across that request, then judge the performances against the same requirements; every version is billable.
Prepare audio for a project handoff
Select MP3, WAV or FLAC to match the file format requested by your collaborator. Keep the downloaded track with its source lyric and style brief so later revisions have a documented starting point.
Prompt guide & pro tips
Describe the V9 arrangement
Write a musical brief in prompt, such as indie folk with acoustic guitar and a campfire mood.
Choose sung words or backing music
For a song, place tagged lyrics in lyrics; for a backing track, enable instrumental and omit the lyric text.
Set version count and export
Select one to three outputs and choose the mp3, wav or flac format before submitting the music request.
From a musical brief to a song or backing track
Mureka V9 combines style direction with supplied lyrics to create a vocal track. The same Song and Background Music endpoint also accepts an instrumental brief: enable instrumental and leave the lyrics field empty. Choose this workflow when you have words to audition as a song, or need a backing track whose mood and instrumentation you can describe.
V9 also offers a version-count control. A request can produce one, two or three candidates for the same musical idea, with each version billed separately. Use that choice to plan a listening comparison rather than treating the additional tracks as free alternatives.
Parameters and input planning
| Field | Values | Notes |
|---|---|---|
| prompt | Up to 1,024 characters | Describe genre, instruments, mood, tempo and vocal character. |
| lyrics | Up to 5,000 characters | Supply the sung lines with section labels such as [Verse] and [Chorus]. |
| instrumental | Enable for background music | Leave lyrics empty and put the musical direction in prompt. |
output_format | mp3, wav, flac | Select the audio file format for your project. |
num_outputs | 1, 2 or 3 | Choose the number of billable versions to receive. |
For a vocal song, keep arrangement notes in prompt and the words to perform in lyrics. A genre label gives an overall direction; naming the lead instrument and supporting parts makes the intended arrangement easier to assess afterward. For a backing track, use instrumental mode rather than placing a request for no singing inside a lyric draft.
Writing a brief you can evaluate
Give the instruments a purpose
Try a direction such as "restrained acoustic folk, fingerpicked guitar leading the arrangement, soft percussion underneath, a warm adult male voice and a reflective evening mood." This is an illustrative brief, not an additional test result. It identifies several audible choices you can listen for without filling the prompt with unrelated adjectives.
For lo-fi background music, describe the piano or other lead sound, the pace of the beat and how busy the arrangement should feel. Keep the intended use in mind: a quiet study track and a foreground song need different amounts of musical activity. Instrumental mode changes the workflow; the descriptive prompt still supplies its character.
Keep the lyric draft readable
Place [Verse] and [Chorus] on their own lines, followed by the words for each section. Separate line breaks make a draft easier to compare with the performance. Write out the lines you want to audition instead of mixing production instructions into the sung text. Review phrasing and section changes by listening; tags do not establish exact timing for every line.
Compare candidates against one brief
Keep the words and musical direction together when saving a result. If you request several outputs, audition each for the same requirements: the intended voice character, intelligible words, the desired instruments and a convincing transition into the chorus. For a later request, change one musical instruction at a time so the comparison has a clear purpose.
What our examples returned
- Late-night R&B: the new MP3 sample is a slow R&B song with a warm male voice singing the supplied verse and chorus about an evening that should continue. Its source brief named soft electric piano and a late-night bass groove; the recorded observation establishes the song style and vocal delivery, without a separate measurement of each instrument.
- Road-trip folk: our supplied lyric became an acoustic-guitar song with warm male vocals, running about 2 minutes 31 seconds. That duration belongs to this particular recording.
- Lo-fi study music: with instrumental enabled, the study brief returned a slow lo-fi beat with piano and vinyl crackle, without vocals.
These samples demonstrate specific outcomes from the listed inputs. They do not establish a completion time, a fixed track length or identical results for later requests. Listen through your own download before placing it into a project, including the transition out of the chorus and the ending.
Pricing and version count
Billing follows the number of generated V9 versions. Vocal songs and instrumental background tracks share the same per-song rate, while failed requests are not charged. Selecting two or three outputs increases the billable quantity accordingly. Consult the Mureka V9 model page for the current price before submitting; a currency amount is not fixed in this guide.
Length, retention and model choice
Mureka says a song can reach about 5 minutes 30 seconds and background music about 4 minutes 30 seconds. Treat those as publisher-stated limits, not target durations for each track. The controls listed here do not include a duration setting, so plan to review the actual length when your project has a timing requirement.
Generated audio is deleted after 14 days. Download the chosen takes within that period and keep their prompt, lyrics, format and version count with them.
Choose Mureka V9.5 when you want its optional female or male vocal_gender selector and accept its higher per-song rate. V9 instead gives you num_outputs and a lower rate per song. MiniMax Music 3.0 is another song-generation model to audition with your material; make the comparison using the resulting music and the controls your brief requires.
Mureka V9 charges by generated version; unsuccessful music requests do not incur a charge.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "mureka/v9/text-to-audio",
"input": {
"prompt": "Calm ambient pads with a slow rising motif",
"num_outputs": 1,
"output_format": "mp3"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.

