Text-to-Music API: Generate Licensed Music from a Prompt

The Text-to-Music API generates original, licensed background music from a single text prompt — no source video required. Here's the full request format, response modes, and working code examples.

Quick answer

Sonilo's Text-to-Music API (POST /v1/text-to-music) generates original background music from a text prompt alone, with no source video required, unlike Sonilo's video-to-music endpoint. Send a prompt describing the mood, instrumentation, and tempo you want, for example "warm lo-fi piano with soft rain ambience, 80 BPM", along with a target duration, and the API returns fully licensed, commercial-safe music either as a live stream of audio chunks or as an async task you poll for a downloadable file. Output is licensed through Shutterstock, and in independent benchmarking on the open-source audio-eval toolkit, Sonilo Music 1.1 outperformed Suno v5.5 on 10 of 13 MusicCaps metrics.

Example request

The following example generates 30 seconds of upbeat corporate background music asynchronously and polls for the result:

curl -X POST https://api.sonilo.com/v1/text-to-music -H "Authorization: Bearer $SONILO_API_KEY" -F prompt="upbeat corporate background track, 90 BPM" -F duration=30 -F mode=async

Async mode returns 202 Accepted immediately with {"task_id": "...", "status": "processing"}. Poll GET /v1/tasks/{task_id} until the response status is "succeeded", at which point the result includes a music object with url, content_type, and file_size, or until it's "failed".

For production use, follow the async music API polling guide to budget status reads, cap retries, and prevent a failed status read from creating a second task.

For async jobs, treat the returned music URL as temporary. Follow the durable storage workflow for generated music before making the file available to users.

Omit mode (or set it to "stream", the default) to receive a 200 NDJSON event stream of audio chunks as they're generated, which suits low-latency playback. Async mode is the better fit for batch jobs, WAV output, or any workflow where polling for a finished file is preferable to handling a live stream. For the byte-to-file implementation, adapt the tested Node.js NDJSON stream parser; its runnable request uses Video-to-Music, so keep your Text-to-Music request fields.

curl -X POST https://api.sonilo.com/v1/text-to-music -H "Authorization: Bearer $SONILO_API_KEY" -F prompt="warm lo-fi piano with soft rain ambience, 80 BPM" -F duration=60

FAQ

What's the maximum duration for text-to-music?

360 seconds (6 minutes) per request, with a 5-second minimum. Use the optional segments field to script per-segment prompts if you need the mood or instrumentation to change over the course of a longer track.

Can I get WAV output instead of M4A?

Yes. Set output_format to "wav". WAV output is only available with mode=async; stream mode always returns M4A chunks.

Is the output licensed for commercial use?

Yes. All Text-to-Music output is fully licensed through Shutterstock and cleared for commercial use, so tracks can go into client work, ads, and monetized video without additional licensing.

Is there a Python or JavaScript SDK?

Yes. Install the official SDK with pip install sonilo (Python) or npm install sonilo (JavaScript/TypeScript) to call the endpoint without hand-building multipart requests.

How is this different from the video-to-music endpoint?

Text-to-Music generates standalone background music from a prompt alone, with no video input. Sonilo's video-to-music endpoint is video-aware: it analyzes an uploaded video and generates a soundtrack timed to its content. Use Text-to-Music when you need a track that isn't tied to a specific video.

Related Sonilo pages

• Full API reference: https://platform.sonilo.com/docs/api/text-to-music
• Pricing: https://sonilo.com/pricing
• Open-source benchmark toolkit (audio-eval): https://github.com/shwj114114/audio_eval