Mira Music

Mira Music composes original songs from a text prompt. Describe genre, mood, and instruments — the model writes the arrangement and (optionally) generates and sings lyrics. An instrumental-only mode is also available.

Capabilities

• Text-to-song — describe the vibe, get a full vocal-led track
• Custom lyrics — pass your own text in lyrics — the model arranges around it
• Instrumental mode — instrumental=true — no vocals, arrangement only
• Style / mood / genre — independent style, mood, genre fields — precise control without long prompts
• Duration — up to 4 minutes per track

Async flow (job + polling)

Generating a song takes 30–120 seconds, so the endpoint returns a job id, not audio. Poll /v1/audio/music/{id} until status="succeeded" then download from audio_url.

Request: create job

curl https://api.vmira.ai/v1/audio/music \
  -H "Authorization: Bearer $MIRA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Melancholic synthwave track with vintage drums",
    "genre": "synthwave",
    "mood": "nostalgic",
    "style": "80s, analog",
    "duration_seconds": 120,
    "instrumental": false,
    "lyrics": null
  }'

Response: job created

{
  "id": "music_job_a1b2c3",
  "status": "queued",
  "created_at": 1714000000
}

Poll status

curl https://api.vmira.ai/v1/audio/music/music_job_a1b2c3 \
  -H "Authorization: Bearer $MIRA_API_KEY"

{
  "id": "music_job_a1b2c3",
  "status": "succeeded",
  "audio_url": "https://cdn.vmira.ai/music/a1b2c3.mp3",
  "duration_seconds": 120,
  "lyrics": "[Verse 1]\nThe city sleeps tonight…"
}

Invoking from chat (tool call)

Mira Music is also exposed as a tool inside chat-completions. The model decides when to invoke it and returns audio_url in the response content.

{
  "type": "tool_call",
  "name": "mira_music_generate",
  "arguments": {
    "prompt": "Epic orchestral overture, 90 seconds",
    "genre": "orchestral",
    "mood": "epic",
    "duration_seconds": 90,
    "instrumental": true
  }
}

Parameters

• prompt — required track description
• genre — optional genre (pop, rock, synthwave, jazz, orchestral…)
• mood — optional mood
• style — optional style hints
• lyrics — optional custom lyrics (null = model writes them)
• instrumental — true for vocals-free instrumental
• duration_seconds — duration, 15–240 seconds