ryan5453 / demucs

Demucs is an audio source separator created by Facebook Research.

Cold

Public
363.5K runs
A100 (40GB)
GitHub
Paper
License

Run with an API

Playground API Examples README Versions

Input

audio

*file

Upload the file to be processed here.

model

string

Choose the demucs audio that proccesses your audio. The readme has more information on what to choose.

Default: "htdemucs"

stem

string

If you just want to isolate one stem, you can choose it here.

Default: "none"

Including output_format and 9 more...

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

npx create-replicate --model=ryan5453/demucs

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 ryan5453/demucs using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const output = await replicate.run(
  "ryan5453/demucs:5a7041cc9b82e5a558fea6b3d7b12dea89625e89da33f0447bd727c2d0ab9e77",
  {
    input: {
      jobs: 0,
      stem: "none",
      audio: "https://replicate.delivery/pbxt/KqCC8G0RySCjKBQinde7gog9XwsfUR8uF6IyD2h6HPirewAg/Josh%20Woodward%20-%20Are%20You%20Having%20Fun",
      model: "htdemucs",
      split: true,
      shifts: 1,
      overlap: 0.25,
      clip_mode: "rescale",
      mp3_preset: 2,
      wav_format: "int24",
      mp3_bitrate: 320,
      output_format: "mp3"
    }
  }
);
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 ryan5453/demucs using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "ryan5453/demucs:5a7041cc9b82e5a558fea6b3d7b12dea89625e89da33f0447bd727c2d0ab9e77",
    input={
        "jobs": 0,
        "stem": "none",
        "audio": "https://replicate.delivery/pbxt/KqCC8G0RySCjKBQinde7gog9XwsfUR8uF6IyD2h6HPirewAg/Josh%20Woodward%20-%20Are%20You%20Having%20Fun",
        "model": "htdemucs",
        "split": True,
        "shifts": 1,
        "overlap": 0.25,
        "clip_mode": "rescale",
        "mp3_preset": 2,
        "wav_format": "int24",
        "mp3_bitrate": 320,
        "output_format": "mp3"
    }
)
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 ryan5453/demucs 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": "5a7041cc9b82e5a558fea6b3d7b12dea89625e89da33f0447bd727c2d0ab9e77",
    "input": {
      "jobs": 0,
      "stem": "none",
      "audio": "https://replicate.delivery/pbxt/KqCC8G0RySCjKBQinde7gog9XwsfUR8uF6IyD2h6HPirewAg/Josh%20Woodward%20-%20Are%20You%20Having%20Fun",
      "model": "htdemucs",
      "split": true,
      "shifts": 1,
      "overlap": 0.25,
      "clip_mode": "rescale",
      "mp3_preset": 2,
      "wav_format": "int24",
      "mp3_bitrate": 320,
      "output_format": "mp3"
    }
  }' \
  https://api.replicate.com/v1/predictions

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

You can run this model locally using Cog. First, install Cog:

brew install cog

If you don’t have Homebrew, there are other installation options available.

Run this to download the model and run it in your local environment:

cog predict r8.im/ryan5453/demucs@sha256:5a7041cc9b82e5a558fea6b3d7b12dea89625e89da33f0447bd727c2d0ab9e77 \
  -i 'jobs=0' \
  -i 'stem="none"' \
  -i 'audio="https://replicate.delivery/pbxt/KqCC8G0RySCjKBQinde7gog9XwsfUR8uF6IyD2h6HPirewAg/Josh%20Woodward%20-%20Are%20You%20Having%20Fun"' \
  -i 'model="htdemucs"' \
  -i 'split=true' \
  -i 'shifts=1' \
  -i 'overlap=0.25' \
  -i 'clip_mode="rescale"' \
  -i 'mp3_preset=2' \
  -i 'wav_format="int24"' \
  -i 'mp3_bitrate=320' \
  -i 'output_format="mp3"'

To learn more, take a look at the Cog documentation.

Run this to download the model and run it in your local environment:

docker run -d -p 5000:5000 --gpus=all r8.im/ryan5453/demucs@sha256:5a7041cc9b82e5a558fea6b3d7b12dea89625e89da33f0447bd727c2d0ab9e77
curl -s -X POST \
  -H "Content-Type: application/json" \
  -d $'{
    "input": {
      "jobs": 0,
      "stem": "none",
      "audio": "https://replicate.delivery/pbxt/KqCC8G0RySCjKBQinde7gog9XwsfUR8uF6IyD2h6HPirewAg/Josh%20Woodward%20-%20Are%20You%20Having%20Fun",
      "model": "htdemucs",
      "split": true,
      "shifts": 1,
      "overlap": 0.25,
      "clip_mode": "rescale",
      "mp3_preset": 2,
      "wav_format": "int24",
      "mp3_bitrate": 320,
      "output_format": "mp3"
    }
  }' \
  http://localhost:5000/predictions

To learn more, take a look at the Cog documentation.

Output

{ "bass": "https://storage.googleapis.com/replicate-files/DGKrtyY9DLJpDNwK8wyx1YJTwffy136G7ST6XM2Mwzp4i2vSA/file", "drums": "https://storage.googleapis.com/replicate-files/tyAr6QMjIl7mP5YPNZWf29VzlG4eRQZScDf64DdxMfMdLafVC/file", "other": "https://storage.googleapis.com/replicate-files/evRBGYcxmfpCJ0qwn8ELHW0BzfsytARPYUOqeK728qcnLafVC/file", "vocals": "https://storage.googleapis.com/replicate-files/fWw45bCZ5ZWILSET8mslNXu2gI4DiBD3NDaCixz2WnvcR7XJA/file" }

{
  "completed_at": "2024-05-01T03:27:54.155224Z",
  "created_at": "2024-05-01T03:24:58.620000Z",
  "data_removed": false,
  "error": null,
  "id": "t2fdpqgsfhrgp0cf68qrcjnrtw",
  "input": {
    "jobs": 0,
    "stem": "none",
    "audio": "https://replicate.delivery/pbxt/KqCC8G0RySCjKBQinde7gog9XwsfUR8uF6IyD2h6HPirewAg/Josh%20Woodward%20-%20Are%20You%20Having%20Fun",
    "model": "htdemucs",
    "split": true,
    "shifts": 1,
    "overlap": 0.25,
    "clip_mode": "rescale",
    "mp3_preset": 2,
    "wav_format": "int24",
    "mp3_bitrate": 320,
    "output_format": "mp3"
  },
  "logs": "0%|                                                                     | 0.0/163.79999999999998 [00:00<?, ?seconds/s]\n  4%|██▏                                                         | 5.85/163.79999999999998 [00:00<00:14, 10.75seconds/s]\n  7%|████▎                                                       | 11.7/163.79999999999998 [00:00<00:07, 19.54seconds/s]\n 11%|████▉                                         | 17.549999999999997/163.79999999999998 [00:00<00:05, 26.84seconds/s]\n 14%|████████▌                                                   | 23.4/163.79999999999998 [00:00<00:04, 32.60seconds/s]\n 18%|██████████▌                                                | 29.25/163.79999999999998 [00:01<00:03, 36.95seconds/s]\n 21%|█████████▊                                    | 35.099999999999994/163.79999999999998 [00:01<00:03, 39.81seconds/s]\n 25%|███████████▌                                  | 40.949999999999996/163.79999999999998 [00:01<00:02, 42.02seconds/s]\n 29%|█████████████████▏                                          | 46.8/163.79999999999998 [00:01<00:02, 43.57seconds/s]\n 32%|██████████████████▉                                        | 52.65/163.79999999999998 [00:01<00:02, 45.10seconds/s]\n 36%|█████████████████████▍                                      | 58.5/163.79999999999998 [00:01<00:02, 46.14seconds/s]\n 39%|███████████████████████▏                                   | 64.35/163.79999999999998 [00:01<00:02, 46.93seconds/s]\n 43%|████████████████████▏                          | 70.19999999999999/163.79999999999998 [00:01<00:01, 47.49seconds/s]\n 46%|███████████████████████████▍                               | 76.05/163.79999999999998 [00:02<00:01, 47.86seconds/s]\n 50%|███████████████████████▌                       | 81.89999999999999/163.79999999999998 [00:02<00:01, 46.86seconds/s]\n 54%|███████████████████████████████▌                           | 87.75/163.79999999999998 [00:02<00:01, 46.92seconds/s]\n 57%|██████████████████████████████████▎                         | 93.6/163.79999999999998 [00:02<00:01, 47.38seconds/s]\n 61%|████████████████████████████▌                  | 99.44999999999999/163.79999999999998 [00:02<00:01, 47.66seconds/s]\n 64%|█████████████████████████████████████▉                     | 105.3/163.79999999999998 [00:02<00:01, 47.91seconds/s]\n 68%|███████████████████████████████▏              | 111.14999999999999/163.79999999999998 [00:02<00:01, 48.06seconds/s]\n 71%|██████████████████████████████████████████▏                | 117.0/163.79999999999998 [00:02<00:00, 48.23seconds/s]\n 75%|███████████████████████████████████████████▌              | 122.85/163.79999999999998 [00:02<00:00, 48.37seconds/s]\n 79%|██████████████████████████████████████████████▎            | 128.7/163.79999999999998 [00:03<00:00, 48.30seconds/s]\n 82%|█████████████████████████████████████▊        | 134.54999999999998/163.79999999999998 [00:03<00:00, 48.35seconds/s]\n 86%|███████████████████████████████████████▍      | 140.39999999999998/163.79999999999998 [00:03<00:00, 48.19seconds/s]\n 89%|███████████████████████████████████████████████████▊      | 146.25/163.79999999999998 [00:03<00:00, 48.29seconds/s]\n 93%|██████████████████████████████████████████████████████▊    | 152.1/163.79999999999998 [00:03<00:00, 48.38seconds/s]\n 96%|███████████████████████████████████████████████████████▉  | 157.95/163.79999999999998 [00:03<00:00, 48.53seconds/s]\n100%|██████████████████████████████████████████████| 163.79999999999998/163.79999999999998 [00:03<00:00, 48.72seconds/s]\n100%|██████████████████████████████████████████████| 163.79999999999998/163.79999999999998 [00:03<00:00, 42.77seconds/s]",
  "metrics": {
    "predict_time": 25.031137,
    "total_time": 175.535224
  },
  "output": {
    "bass": "https://storage.googleapis.com/replicate-files/DGKrtyY9DLJpDNwK8wyx1YJTwffy136G7ST6XM2Mwzp4i2vSA/file",
    "drums": "https://storage.googleapis.com/replicate-files/tyAr6QMjIl7mP5YPNZWf29VzlG4eRQZScDf64DdxMfMdLafVC/file",
    "other": "https://storage.googleapis.com/replicate-files/evRBGYcxmfpCJ0qwn8ELHW0BzfsytARPYUOqeK728qcnLafVC/file",
    "vocals": "https://storage.googleapis.com/replicate-files/fWw45bCZ5ZWILSET8mslNXu2gI4DiBD3NDaCixz2WnvcR7XJA/file"
  },
  "started_at": "2024-05-01T03:27:29.124087Z",
  "status": "succeeded",
  "urls": {
    "get": "https://api.replicate.com/v1/predictions/t2fdpqgsfhrgp0cf68qrcjnrtw",
    "cancel": "https://api.replicate.com/v1/predictions/t2fdpqgsfhrgp0cf68qrcjnrtw/cancel"
  },
  "version": "53191dee0efbfc3cbfdbab276b0dcce930705e9a4d1bb9fe1e2a7cdd33d9ca82"
}

Generated in

25.0 seconds

Tweak it Share Report

0%|                                                                     | 0.0/163.79999999999998 [00:00<?, ?seconds/s]
  4%|██▏                                                         | 5.85/163.79999999999998 [00:00<00:14, 10.75seconds/s]
  7%|████▎                                                       | 11.7/163.79999999999998 [00:00<00:07, 19.54seconds/s]
 11%|████▉                                         | 17.549999999999997/163.79999999999998 [00:00<00:05, 26.84seconds/s]
 14%|████████▌                                                   | 23.4/163.79999999999998 [00:00<00:04, 32.60seconds/s]
 18%|██████████▌                                                | 29.25/163.79999999999998 [00:01<00:03, 36.95seconds/s]
 21%|█████████▊                                    | 35.099999999999994/163.79999999999998 [00:01<00:03, 39.81seconds/s]
 25%|███████████▌                                  | 40.949999999999996/163.79999999999998 [00:01<00:02, 42.02seconds/s]
 29%|█████████████████▏                                          | 46.8/163.79999999999998 [00:01<00:02, 43.57seconds/s]
 32%|██████████████████▉                                        | 52.65/163.79999999999998 [00:01<00:02, 45.10seconds/s]
 36%|█████████████████████▍                                      | 58.5/163.79999999999998 [00:01<00:02, 46.14seconds/s]
 39%|███████████████████████▏                                   | 64.35/163.79999999999998 [00:01<00:02, 46.93seconds/s]
 43%|████████████████████▏                          | 70.19999999999999/163.79999999999998 [00:01<00:01, 47.49seconds/s]
 46%|███████████████████████████▍                               | 76.05/163.79999999999998 [00:02<00:01, 47.86seconds/s]
 50%|███████████████████████▌                       | 81.89999999999999/163.79999999999998 [00:02<00:01, 46.86seconds/s]
 54%|███████████████████████████████▌                           | 87.75/163.79999999999998 [00:02<00:01, 46.92seconds/s]
 57%|██████████████████████████████████▎                         | 93.6/163.79999999999998 [00:02<00:01, 47.38seconds/s]
 61%|████████████████████████████▌                  | 99.44999999999999/163.79999999999998 [00:02<00:01, 47.66seconds/s]
 64%|█████████████████████████████████████▉                     | 105.3/163.79999999999998 [00:02<00:01, 47.91seconds/s]
 68%|███████████████████████████████▏              | 111.14999999999999/163.79999999999998 [00:02<00:01, 48.06seconds/s]
 71%|██████████████████████████████████████████▏                | 117.0/163.79999999999998 [00:02<00:00, 48.23seconds/s]
 75%|███████████████████████████████████████████▌              | 122.85/163.79999999999998 [00:02<00:00, 48.37seconds/s]
 79%|██████████████████████████████████████████████▎            | 128.7/163.79999999999998 [00:03<00:00, 48.30seconds/s]
 82%|█████████████████████████████████████▊        | 134.54999999999998/163.79999999999998 [00:03<00:00, 48.35seconds/s]
 86%|███████████████████████████████████████▍      | 140.39999999999998/163.79999999999998 [00:03<00:00, 48.19seconds/s]
 89%|███████████████████████████████████████████████████▊      | 146.25/163.79999999999998 [00:03<00:00, 48.29seconds/s]
 93%|██████████████████████████████████████████████████████▊    | 152.1/163.79999999999998 [00:03<00:00, 48.38seconds/s]
 96%|███████████████████████████████████████████████████████▉  | 157.95/163.79999999999998 [00:03<00:00, 48.53seconds/s]
100%|██████████████████████████████████████████████| 163.79999999999998/163.79999999999998 [00:03<00:00, 48.72seconds/s]
100%|██████████████████████████████████████████████| 163.79999999999998/163.79999999999998 [00:03<00:00, 42.77seconds/s]

This example was created by a different version, ryan5453/demucs:53191dee.

Run time and cost

This model costs approximately $0.068 to run on Replicate, or 14 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 A100 (40GB) GPU hardware. Predictions typically complete within 60 seconds. The predict time for this model varies significantly based on the inputs.

Readme

About

Demucs is an audio source separator model created by Facebook Research. The model ingests an audio file, and returns stems (the separated parts such as the vocals, drums, and bass.)

The original author of Demucs left Meta and therefore the original repository is not maintained, so this Cog uses a fork by the original author. You can find the fork here.

Model Versions

Demucs v4 (Hybrid Transformers Demucs)

htdemucs - base model
htdemucs_ft - fine-tuned version (this has better results but will take 4x as long)
htdemucs_6s - adds piano and guitar stems (piano does not work very well)

Demucs v3 (Hybrid Demucs)

hdemucs_mmi - base model

Demucs v2

mdx_q - quantized of the base model (trained on only MusDB)
mdx_extra_q - quantized version of the base model with extra data