MetaMediumStandard

MusicGen

Meta's MusicGen via Replicate. Generates music from a text prompt or a reference melody (stereo / melody / large variants), can continue an input clip, and outputs MP3 or WAV. Billed per second of requested duration.

94 credits

per 8s — 30.5 + 7.9 × seconds (measured), 30 s about 270

Text-to-music and melody-conditioned generation

Continuation of an input audio clip

Four model variants (stereo-melody-large, stereo-large, melody-large, large)

Up to 30 seconds per generation

Sampling controls (top_k, top_p, temperature, CFG)

Run it right now

Test this model instantly in the Console Playground — no code required

Use with AI Assistant

Copy usage instructions for Claude, ChatGPT, or other AI

llms.txtllms.txt ↗

Quick Start

curl -X POST "https://api.core.today/v1/predictions" \
  -H "Content-Type: application/json" \
  -H "X-API-Key: cdt_your_api_key" \
  -d '{
  "model": "meta/musicgen",
  "input": {
    "prompt": "Upbeat K-pop electro-pop company jingle instrumental, 128 BPM, synth brass stabs, punchy drums, bright synth lead, catchy hook, fun and playful",
    "model_version": "stereo-melody-large",
    "duration": 30,
    "output_format": "mp3"
  }
}'

Parameters

Parameter	Type	Required	Default	Description
`prompt`	string	Yes	-	A description of the music you want to generate
`model_version`	string	No	stereo-melody-large	Model variant. stereo-* produce stereo; melody-* accept input_audio as a melody reference stereo-melody-largestereo-largemelody-largelarge
`duration`	integer	No	8	Duration of the generated audio in seconds (1-30). Billed per second
`input_audio`	string	No	-	Audio file that influences the output: continued when continuation is true, otherwise its melody is mimicked (melody models)
`continuation`	boolean	No	false	If true, generated music continues from input_audio
`continuation_start`	integer	No	0	Start time (s) of input_audio for continuation
`continuation_end`	integer	No	-	End time (s) of input_audio for continuation. Omit for the end of the clip
`multi_band_diffusion`	boolean	No	false	Decode EnCodec tokens with MultiBand Diffusion (non-stereo models only)
`normalization_strategy`	string	No	loudness	Strategy for normalizing audio loudnessclippeakrms
`top_k`	integer	No	250	Reduces sampling to the k most likely tokens
`top_p`	number	No	0	Nucleus sampling probability. 0 uses top_k sampling
`temperature`	number	No	1	Sampling temperature. Higher means more diversity
`classifier_free_guidance`	integer	No	3	Influence of inputs on the output. Higher adheres more to the prompt
`output_format`	string	No	wav	Output format for generated audio wavmp3
`seed`	integer	No	-	Seed for the random number generator. Omit or -1 for random

How to Provide File Input

There are 3 ways to provide files for the input_audio parameter:

Recommended

Image URL

Pass a publicly accessible URL directly. With the Storage API (POST /v1/files/upload-url) the file uploads straight to S3 (50MB per file) and you pass the returned file_url.

{
  "model": "meta/musicgen",
  "input": {
    "prompt": "your prompt here",
    "input_audio": "https://example.com/image.jpg"
  }
}

Direct Upload (Multipart)

Attach files directly to POST /v1/predictions/upload. No separate upload step, but the bytes pass through the API server so the whole request is capped at 10MB.

curl -X POST "https://api.core.today/v1/predictions/upload" \
  -H "X-API-Key: cdt_your_api_key" \
  -F "model=meta/musicgen" \
  -F 'input={"prompt":"your prompt here"}' \
  -F "file:input_audio=@your_file.png"

See the File Upload docs for more upload methods including Presigned URLs.

Common Parameters

Common parameters used when calling POST /v1/predictions.

Parameter	Type	Required	Default	Description
`model`	string	Yes	-	Model identifier
`input`	object	Yes	-	Object containing the model-specific parameters from the table above
`output_folder`	string	No	-	Folder path for output files (max 256 chars, '..' not allowed)
`webhook_url`	string	No	-	Webhook URL to call on completion
`is_public`	boolean	No	false	If true, output files are also available via permanent public URLs

Examples

Core.Today theme (instrumental)

The same company-theme prompt given to every instrumental model — compare the results side by side

curl -X POST "https://api.core.today/v1/predictions" \
  -H "Content-Type: application/json" \
  -H "X-API-Key: cdt_your_api_key" \
  -d '{
  "model": "meta/musicgen",
  "input": {
    "prompt": "Upbeat K-pop electro-pop company jingle instrumental, 128 BPM, synth brass stabs, punchy drums, bright synth lead, catchy hook, fun and playful",
    "model_version": "stereo-melody-large",
    "duration": 30,
    "output_format": "mp3"
  }
}'

Cinematic 8-second cue

Default duration with the stereo-large variant and MP3 output

curl -X POST "https://api.core.today/v1/predictions" \
  -H "Content-Type: application/json" \
  -H "X-API-Key: cdt_your_api_key" \
  -d '{
  "model": "meta/musicgen",
  "input": {
    "prompt": "Triumphant cinematic orchestral melody in G major leading to a crescendo",
    "model_version": "stereo-large",
    "duration": 8,
    "output_format": "mp3"
  }
}'

30-second folk melody (large)

Maximum duration with the mono large variant — about 270 credits

curl -X POST "https://api.core.today/v1/predictions" \
  -H "Content-Type: application/json" \
  -H "X-API-Key: cdt_your_api_key" \
  -d '{
  "model": "meta/musicgen",
  "input": {
    "prompt": "Acoustic guitar folk melody, gentle and warm",
    "model_version": "large",
    "duration": 30,
    "output_format": "mp3"
  }
}'