Capability

Audio to Prompt

Turn a song into a reusable style prompt, an instrumental prompt and its lyrics tagged by section, ready for the song and music generators.

audio_to_prompt — native payload
What it does
Response
Pricing

Native only — new capability

No compat route exists for Audio to Prompt. Call POST /v1/generations with the tool_id for model="v1", operation="audio_to_prompt" (see the Tool Catalog).

audio_to_prompt — native payload

POST/v1/generationstool_id for (model="v1", "audio_to_prompt")

Field	Type	Required	Default	Description
song	file input	Yes	—	The song to describe, up to 10 minutes. A file over 24MB is read from a compressed copy.
language	string	No	""	Optional ISO 639-1 code for the lyrics, e.g. 'en', 'hi', 'es'. Leave blank to detect it. The prompts are always in English.

curl -X POST https://apiv2.soundverse.ai/v1/generations \
  -H "Authorization: Bearer sksoundverse_..." \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: audio-to-prompt-req-001" \
  -d '{
    "tool_id": "<audio_to_prompt tool_id>",
    "payload_json": "{\"song\": {\"url\": \"https://example.com/track.mp3\"}}"
  }'

What it does

Listens to the song for its genre, mood, energy, instruments, production and vocals.
Measures its tempo and key, and finds its sections (intro, verse, chorus…).
Reads the lyrics, from the file’s saved timed lyrics when it has them, and tags them by section. The words are kept exactly as sung.
Writes a style prompt and an instrumental prompt. Prompts never name an artist or a song.

Listening and writing must succeed. If tempo and key, the sections or the lyrics can’t be read, that step is skipped and the prompts are written without it.

Response

Once the task completes, its output has this shape:

{
  "text": "A high-energy Latin pop anthem...\n\n**Style prompt**\n```\n...\n```\n...",
  "vocal_gender": "male",
  "has_vocals": true,
  "bpm": 100.5,
  "key": "Eb major",
  "genres": ["Latin Pop", "Cumbia Pop"],
  "mood": ["festive", "triumphant"],
  "instruments": ["brass section", "drum kit", "synth bass"],
  "sections": [{"label": "intro", "start": 0.0, "end": 8.2}, ...],
  "summary": "A high-energy Latin pop anthem at 100 BPM in Eb major.",
  "style_prompt": "Latin Pop, Cumbia Pop, festive, triumphant, 100 BPM, Eb major, bright male tenor lead, ...",
  "instrumental_prompt": "Latin Pop, Cumbia Pop, festive, triumphant, 100 BPM, Eb major, punchy brass hooks, ...",
  "lyrics": "[Verse]\nEn el potrero suena la pelota...\n\n[Chorus]\n...",
  "steps": [...]
}

text is the whole result as Markdown: the summary, then each prompt and the lyrics in their own code block.
style_prompt and instrumental_prompt are at most 1,024 characters each; lyrics is at most 3,000, cut at the last whole section that fits.
lyrics is empty for an instrumental. bpm is 0 and key is empty when they can’t be measured. vocal_gender is female, male, or empty when unclear.

Pricing

Version	Price
v1	$0.25

USD flat price per run, regardless of license tier.

USD per song, the same on every license tier. Rate limit: 30 calls an hour, 150 a day.

← TranscriptionSound Effects →