riffusion/riffusion – Run with an API on Replicate

riffusion / riffusion

Stable diffusion for real-time music generation

Cold

Public
1M runs
T4
GitHub
Paper
License

Run with an API

Playground API Examples README Versions

Input

prompt_a

string

Shift + Return to add a new line

The prompt for your audio

Default: "funky synth solo"

denoising

number

(minimum: 0, maximum: 1)

How much to transform input spectrogram

Default: 0.75

prompt_b

string

Shift + Return to add a new line

The second prompt to interpolate with the first, leave blank if no interpolation

alpha

number

(minimum: 0, maximum: 1)

Interpolation alpha if using two prompts. A value of 0 uses prompt_a fully, a value of 1 uses prompt_b fully

Default: 0.5

num_inference_steps

integer

(minimum: 1)

Number of steps to run the diffusion model

Default: 50

seed_image_id

string

Seed spectrogram to use

Default: "vibes"

Run this model in Node.js with one line of code:

npx create-replicate --model=riffusion/riffusion

or set up a project from scratch

Install Replicate’s Node.js client library:

npm install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import and set up the client:

import Replicate from "replicate";

const replicate = new Replicate({
  auth: process.env.REPLICATE_API_TOKEN,
});

Run riffusion/riffusion using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const output = await replicate.run(
  "riffusion/riffusion:8cf61ea6c56afd61d8f5b9ffd14d7c216c0a93844ce2d82ac1c9ecc9c7f24e05",
  {
    input: {
      alpha: 0.5,
      prompt_a: "funky synth solo",
      prompt_b: "90's rap",
      denoising: 0.75,
      seed_image_id: "vibes",
      num_inference_steps: 50
    }
  }
);

console.log(output);

To learn more, take a look at the guide on getting started with Node.js.

Install Replicate’s Python client library:

pip install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import the client:

import replicate

Run riffusion/riffusion using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "riffusion/riffusion:8cf61ea6c56afd61d8f5b9ffd14d7c216c0a93844ce2d82ac1c9ecc9c7f24e05",
    input={
        "alpha": 0.5,
        "prompt_a": "funky synth solo",
        "prompt_b": "90's rap",
        "denoising": 0.75,
        "seed_image_id": "vibes",
        "num_inference_steps": 50
    }
)
print(output)

To learn more, take a look at the guide on getting started with Python.

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Run riffusion/riffusion using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

curl -s -X POST \
  -H "Authorization: Bearer $REPLICATE_API_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Prefer: wait" \
  -d $'{
    "version": "riffusion/riffusion:8cf61ea6c56afd61d8f5b9ffd14d7c216c0a93844ce2d82ac1c9ecc9c7f24e05",
    "input": {
      "alpha": 0.5,
      "prompt_a": "funky synth solo",
      "prompt_b": "90\'s rap",
      "denoising": 0.75,
      "seed_image_id": "vibes",
      "num_inference_steps": 50
    }
  }' \
  https://api.replicate.com/v1/predictions

To learn more, take a look at Replicate’s HTTP API reference docs.

Output

audio

Video Player is loading.

Current Time 00:00:000

Duration 00:00:000

Loaded: 0%

Stream Type LIVE

Remaining Time 00:00:000

spectrogram

{
  "completed_at": "2022-12-16T08:17:18.378786Z",
  "created_at": "2022-12-16T08:17:09.312030Z",
  "data_removed": false,
  "error": null,
  "id": "jvlsnvme3jbhvdyoz62barakza",
  "input": {
    "alpha": 0.5,
    "prompt_a": "funky synth solo",
    "prompt_b": "90's rap",
    "denoising": 0.75,
    "seed_image_id": "vibes",
    "num_inference_steps": 50
  },
  "logs": "tokens: [[280, 271, 262, 338, 7348]]\nweights: [[1.0, 1.0, 1.0, 1.0, 1.0]]\n  0%|          | 0/38 [00:00<?, ?it/s]\n  3%|▎         | 1/38 [00:00<00:06,  5.69it/s]\n  5%|▌         | 2/38 [00:00<00:06,  5.68it/s]\n  8%|▊         | 3/38 [00:00<00:06,  5.65it/s]\n 11%|█         | 4/38 [00:00<00:06,  5.61it/s]\n 13%|█▎        | 5/38 [00:00<00:06,  5.27it/s]\n 16%|█▌        | 6/38 [00:01<00:06,  5.28it/s]\n 18%|█▊        | 7/38 [00:01<00:05,  5.35it/s]\n 21%|██        | 8/38 [00:01<00:05,  5.40it/s]\n 24%|██▎       | 9/38 [00:01<00:05,  5.42it/s]\n 26%|██▋       | 10/38 [00:01<00:05,  5.30it/s]\n 29%|██▉       | 11/38 [00:02<00:05,  5.21it/s]\n 32%|███▏      | 12/38 [00:02<00:04,  5.28it/s]\n 34%|███▍      | 13/38 [00:02<00:04,  5.36it/s]\n 37%|███▋      | 14/38 [00:02<00:04,  5.42it/s]\n 39%|███▉      | 15/38 [00:02<00:04,  5.36it/s]\n 42%|████▏     | 16/38 [00:02<00:04,  5.26it/s]\n 45%|████▍     | 17/38 [00:03<00:04,  5.24it/s]\n 47%|████▋     | 18/38 [00:03<00:03,  5.28it/s]\n 50%|█████     | 19/38 [00:03<00:03,  5.34it/s]\n 53%|█████▎    | 20/38 [00:03<00:03,  5.32it/s]\n 55%|█████▌    | 21/38 [00:03<00:03,  5.25it/s]\n 58%|█████▊    | 22/38 [00:04<00:03,  5.26it/s]\n 61%|██████    | 23/38 [00:04<00:02,  5.26it/s]\n 63%|██████▎   | 24/38 [00:04<00:02,  5.27it/s]\n 66%|██████▌   | 25/38 [00:04<00:02,  5.29it/s]\n 68%|██████▊   | 26/38 [00:04<00:02,  5.28it/s]\n 71%|███████   | 27/38 [00:05<00:02,  5.19it/s]\n 74%|███████▎  | 28/38 [00:05<00:01,  5.19it/s]\n 76%|███████▋  | 29/38 [00:05<00:01,  5.24it/s]\n 79%|███████▉  | 30/38 [00:05<00:01,  5.24it/s]\n 82%|████████▏ | 31/38 [00:05<00:01,  5.21it/s]\n 84%|████████▍ | 32/38 [00:06<00:01,  5.17it/s]\n 87%|████████▋ | 33/38 [00:06<00:00,  5.16it/s]\n 89%|████████▉ | 34/38 [00:06<00:00,  5.16it/s]\n 92%|█████████▏| 35/38 [00:06<00:00,  5.22it/s]\n 95%|█████████▍| 36/38 [00:06<00:00,  5.18it/s]\n 97%|█████████▋| 37/38 [00:07<00:00,  5.16it/s]\n100%|██████████| 38/38 [00:07<00:00,  5.18it/s]\n100%|██████████| 38/38 [00:07<00:00,  5.28it/s]",
  "metrics": {
    "predict_time": 9.069519,
    "total_time": 9.066756
  },
  "output": {
    "audio": "https://replicate.delivery/pbxt/SCiO1SBkqj7gL5cTsq8AXz5pIwPajeiWbb9s17KtyQ2G3OFIA/gen_sound.wav",
    "spectrogram": "https://replicate.delivery/pbxt/Xkr2iGJhHkJ3K9j5B7Xa1sSbRQmPZ7d8R17WXsalRbnjbnCE/spectrogram.jpg"
  },
  "started_at": "2022-12-16T08:17:09.309267Z",
  "status": "succeeded",
  "urls": {
    "get": "https://api.replicate.com/v1/predictions/jvlsnvme3jbhvdyoz62barakza",
    "cancel": "https://api.replicate.com/v1/predictions/jvlsnvme3jbhvdyoz62barakza/cancel"
  },
  "version": "8cf61ea6c56afd61d8f5b9ffd14d7c216c0a93844ce2d82ac1c9ecc9c7f24e05"
}

Generated in

9.1 seconds

Tweak itReport

tokens: [[280, 271, 262, 338, 7348]]
weights: [[1.0, 1.0, 1.0, 1.0, 1.0]]
  0%|          | 0/38 [00:00<?, ?it/s]
  3%|▎         | 1/38 [00:00<00:06,  5.69it/s]
  5%|▌         | 2/38 [00:00<00:06,  5.68it/s]
  8%|▊         | 3/38 [00:00<00:06,  5.65it/s]
 11%|█         | 4/38 [00:00<00:06,  5.61it/s]
 13%|█▎        | 5/38 [00:00<00:06,  5.27it/s]
 16%|█▌        | 6/38 [00:01<00:06,  5.28it/s]
 18%|█▊        | 7/38 [00:01<00:05,  5.35it/s]
 21%|██        | 8/38 [00:01<00:05,  5.40it/s]
 24%|██▎       | 9/38 [00:01<00:05,  5.42it/s]
 26%|██▋       | 10/38 [00:01<00:05,  5.30it/s]
 29%|██▉       | 11/38 [00:02<00:05,  5.21it/s]
 32%|███▏      | 12/38 [00:02<00:04,  5.28it/s]
 34%|███▍      | 13/38 [00:02<00:04,  5.36it/s]
 37%|███▋      | 14/38 [00:02<00:04,  5.42it/s]
 39%|███▉      | 15/38 [00:02<00:04,  5.36it/s]
 42%|████▏     | 16/38 [00:02<00:04,  5.26it/s]
 45%|████▍     | 17/38 [00:03<00:04,  5.24it/s]
 47%|████▋     | 18/38 [00:03<00:03,  5.28it/s]
 50%|█████     | 19/38 [00:03<00:03,  5.34it/s]
 53%|█████▎    | 20/38 [00:03<00:03,  5.32it/s]
 55%|█████▌    | 21/38 [00:03<00:03,  5.25it/s]
 58%|█████▊    | 22/38 [00:04<00:03,  5.26it/s]
 61%|██████    | 23/38 [00:04<00:02,  5.26it/s]
 63%|██████▎   | 24/38 [00:04<00:02,  5.27it/s]
 66%|██████▌   | 25/38 [00:04<00:02,  5.29it/s]
 68%|██████▊   | 26/38 [00:04<00:02,  5.28it/s]
 71%|███████   | 27/38 [00:05<00:02,  5.19it/s]
 74%|███████▎  | 28/38 [00:05<00:01,  5.19it/s]
 76%|███████▋  | 29/38 [00:05<00:01,  5.24it/s]
 79%|███████▉  | 30/38 [00:05<00:01,  5.24it/s]
 82%|████████▏ | 31/38 [00:05<00:01,  5.21it/s]
 84%|████████▍ | 32/38 [00:06<00:01,  5.17it/s]
 87%|████████▋ | 33/38 [00:06<00:00,  5.16it/s]
 89%|████████▉ | 34/38 [00:06<00:00,  5.16it/s]
 92%|█████████▏| 35/38 [00:06<00:00,  5.22it/s]
 95%|█████████▍| 36/38 [00:06<00:00,  5.18it/s]
 97%|█████████▋| 37/38 [00:07<00:00,  5.16it/s]
100%|██████████| 38/38 [00:07<00:00,  5.18it/s]
100%|██████████| 38/38 [00:07<00:00,  5.28it/s]

Examples

View more examples

Run time and cost

This model costs approximately $0.037 to run on Replicate, or 27 runs per $1, but this varies depending on your inputs. It is also open source and you can run it on your own computer with Docker.

This model runs on Nvidia T4 GPU hardware. Predictions typically complete within 3 minutes. The predict time for this model varies significantly based on the inputs.

Readme

You’ve heard of Stable Diffusion, the open-source AI model that generates images from text?

photograph of an astronaut riding a horse

Well, we fine-tuned the model to generate images of spectrograms, like this:

funk bassline with a jazzy saxophone solo

The magic is that this spectrogram can then be converted to an audio clip. (See demo above.)

This is the v1.5 stable diffusion model with no modifications, just fine-tuned on images of spectrograms paired with text. Audio processing happens downstream of the model.

It can generate infinite variations of a prompt by varying the seed. All the same web UIs and techniques like img2img, inpainting, negative prompts, and interpolation work out of the box.

Read more about it here: https://www.riffusion.com/about

Citation:

@software{Forsgren_Martiros_2022,
  author = {Forsgren, Seth* and Martiros, Hayk*},
  title = {{Riffusion - Stable diffusion for real-time music generation}},
  url = {https://riffusion.com/about},
  year = {2022}
}