lucataco / singing_voice_conversion

Amphion Singing Voice Conversion: DiffWaveNetSVC

Cold

Public
822 runs
L40S
GitHub
Paper
License

Run with an API

Playground API Examples README Versions

Input

Video Player is loading.

Current Time 00:00:000

Duration 00:00:000

Loaded: 0%

Stream Type LIVE

Remaining Time 00:00:000

source_audio

*file

Input source audio file

target_singer

string

Target singer to convert audio to

Default: "Taylor Swift"

pitch_shift_control

string

Pitch shift control

Default: "Auto Shift"

key_shift_mode

integer

(minimum: -6, maximum: 6)

Key shift values

Default: 0

diffusion_inference_steps

integer

(minimum: 0, maximum: 1000)

Diffusion inference steps

Default: 1000

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

npx create-replicate --model=lucataco/singing_voice_conversion

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

const output = await replicate.run(
  "lucataco/singing_voice_conversion:f29872ee3557e0186735048f1d6de98a52518ae5c49e19453b3fdaad710bdc2b",
  {
    input: {
      source_audio: "https://replicate.delivery/pbxt/K5coMzCs7mnhljhRVhdhN29I3RlHPkneVxrbPtyArzxvAVtI/adele.wav",
      target_singer: "Taylor Swift",
      key_shift_mode: 0,
      pitch_shift_control: "Auto Shift",
      diffusion_inference_steps: 1000
    }
  }
);
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 lucataco/singing_voice_conversion using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "lucataco/singing_voice_conversion:f29872ee3557e0186735048f1d6de98a52518ae5c49e19453b3fdaad710bdc2b",
    input={
        "source_audio": "https://replicate.delivery/pbxt/K5coMzCs7mnhljhRVhdhN29I3RlHPkneVxrbPtyArzxvAVtI/adele.wav",
        "target_singer": "Taylor Swift",
        "key_shift_mode": 0,
        "pitch_shift_control": "Auto Shift",
        "diffusion_inference_steps": 1000
    }
)
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 lucataco/singing_voice_conversion 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": "f29872ee3557e0186735048f1d6de98a52518ae5c49e19453b3fdaad710bdc2b",
    "input": {
      "source_audio": "https://replicate.delivery/pbxt/K5coMzCs7mnhljhRVhdhN29I3RlHPkneVxrbPtyArzxvAVtI/adele.wav",
      "target_singer": "Taylor Swift",
      "key_shift_mode": 0,
      "pitch_shift_control": "Auto Shift",
      "diffusion_inference_steps": 1000
    }
  }' \
  https://api.replicate.com/v1/predictions

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

Install Cog

brew install cog

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

Pull and run lucataco/singing_voice_conversion using Cog (this will download the full model and run it in your local environment):

cog predict r8.im/lucataco/singing_voice_conversion@sha256:f29872ee3557e0186735048f1d6de98a52518ae5c49e19453b3fdaad710bdc2b \
  -i 'source_audio="https://replicate.delivery/pbxt/K5coMzCs7mnhljhRVhdhN29I3RlHPkneVxrbPtyArzxvAVtI/adele.wav"' \
  -i 'target_singer="Taylor Swift"' \
  -i 'key_shift_mode=0' \
  -i 'pitch_shift_control="Auto Shift"' \
  -i 'diffusion_inference_steps=1000'

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

Pull and run lucataco/singing_voice_conversion using Docker (this will download the full model and run it in your local environment):

docker run -d -p 5000:5000 --gpus=all r8.im/lucataco/singing_voice_conversion@sha256:f29872ee3557e0186735048f1d6de98a52518ae5c49e19453b3fdaad710bdc2b
curl -s -X POST \
  -H "Content-Type: application/json" \
  -d $'{
    "input": {
      "source_audio": "https://replicate.delivery/pbxt/K5coMzCs7mnhljhRVhdhN29I3RlHPkneVxrbPtyArzxvAVtI/adele.wav",
      "target_singer": "Taylor Swift",
      "key_shift_mode": 0,
      "pitch_shift_control": "Auto Shift",
      "diffusion_inference_steps": 1000
    }
  }' \
  http://localhost:5000/predictions

Output

Video Player is loading.

Current Time 00:00:000

Duration 00:00:000

Loaded: 0%

Stream Type LIVE

Remaining Time 00:00:000

{
  "completed_at": "2023-12-21T22:37:49.201677Z",
  "created_at": "2023-12-21T22:35:12.670504Z",
  "data_removed": false,
  "error": null,
  "id": "h37sr5dbojsspt56c34pvjanoe",
  "input": {
    "source_audio": "https://replicate.delivery/pbxt/K5coMzCs7mnhljhRVhdhN29I3RlHPkneVxrbPtyArzxvAVtI/adele.wav",
    "target_singer": "Taylor Swift",
    "key_shift_mode": 0,
    "pitch_shift_control": "Auto Shift",
    "diffusion_inference_steps": 1000
  },
  "logs": "/tmp/input_audio\nvocalist_l1_TaylorSwift\nautoshift\ngetopt: unrecognized option '--diffusion_inference_steps'\nExprimental Configuration File: ckpts/svc/vocalist_l1_contentvec+whisper/args.json\nThe following values were not passed to `accelerate launch` and had defaults used instead:\n`--num_processes` was set to a value of `1`\n`--num_machines` was set to a value of `1`\n`--mixed_precision` was set to a value of `'no'`\n`--dynamo_backend` was set to a value of `'no'`\nTo avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.\nMonotonic align not found. Please make sure you have compiled it.\nThere are 1 source audios:\n**********\nConversion for source...\nPrepare for meta eval data: 0.0s\n  0%|          | 0/1 [00:00<?, ?it/s]\n  0%|          | 0/1 [00:00<?, ?it/s]\u001b[A\n100%|██████████| 1/1 [00:01<00:00,  1.98s/it]\u001b[A\n100%|██████████| 1/1 [00:01<00:00,  1.98s/it]\nPrepare for acoustic features: 2.0s\nPrepare for content features: 0.0s\n2023-12-21 22:37:31 | INFO | inference | ========================================================\n2023-12-21 22:37:31 | INFO | inference | ||\t\tNew inference process started.\t\t||\n2023-12-21 22:37:31 | INFO | inference | ========================================================\n2023-12-21 22:37:31 | INFO | inference |\n2023-12-21 22:37:31 | DEBUG | inference | Using DEBUG logging level.\n2023-12-21 22:37:31 | DEBUG | inference | Acoustic dir: ckpts/svc/vocalist_l1_contentvec+whisper\n2023-12-21 22:37:31 | DEBUG | inference | Vocoder dir: pretrained/bigvgan\n2023-12-21 22:37:31 | DEBUG | inference | Setting random seed done in 0.83ms\n2023-12-21 22:37:31 | DEBUG | inference | Random seed: 10086\n2023-12-21 22:37:31 | INFO | inference | Building dataset...\n2023-12-21 22:37:31 | INFO | inference | Building dataset done in 4.60ms\n2023-12-21 22:37:31 | INFO | inference | Building model...\n2023-12-21 22:37:31 | INFO | inference | Building model done in 276.183ms\n2023-12-21 22:37:31 | INFO | inference | Initializing accelerate...\n2023-12-21 22:37:32 | INFO | inference | Initializing accelerate done in 1057.268ms\n2023-12-21 22:37:32 | INFO | inference | Loading checkpoint...\n2023-12-21 22:37:32 | INFO | accelerate.accelerator | Loading states from ckpts/svc/vocalist_l1_contentvec+whisper/checkpoint/epoch-6852_step-0678447_loss-1.946773\n2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All model weights loaded successfully\n2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All optimizer states loaded successfully\n2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All scheduler states loaded successfully\n2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All dataloader sampler states loaded successfully\n2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All random states loaded successfully\n2023-12-21 22:37:32 | INFO | accelerate.accelerator | Loading in 0 custom states\n2023-12-21 22:37:32 | INFO | inference | Loading checkpoint done in 106.015ms\n2023-12-21 22:37:32 | INFO | inference | Using PNDM scheduler.\nModel Init: 1.5s\nAuto transposing: source f0 median = 372.9, target f0 median = 286.9, factor = 0.77\n  0%|          | 0/1009 [00:00<?, ?it/s]\u001b[A\n  0%|          | 1/1009 [00:02<39:02,  2.32s/it]\u001b[A\n  2%|▏         | 20/1009 [00:02<01:26, 11.39it/s]\u001b[A\n  4%|▍         | 39/1009 [00:02<00:38, 25.02it/s]\u001b[A\n  6%|▌         | 58/1009 [00:02<00:23, 41.23it/s]\u001b[A\n  8%|▊         | 77/1009 [00:02<00:15, 59.36it/s]\u001b[A\n 10%|▉         | 96/1009 [00:02<00:11, 78.69it/s]\u001b[A\n 11%|█▏        | 115/1009 [00:02<00:09, 98.15it/s]\u001b[A\n 13%|█▎        | 134/1009 [00:03<00:07, 116.43it/s]\u001b[A\n 15%|█▌        | 153/1009 [00:03<00:06, 132.68it/s]\u001b[A\n 17%|█▋        | 172/1009 [00:03<00:05, 146.29it/s]\u001b[A\n 19%|█▉        | 191/1009 [00:03<00:05, 157.09it/s]\u001b[A\n 21%|██        | 210/1009 [00:03<00:04, 164.64it/s]\u001b[A\n 23%|██▎       | 229/1009 [00:03<00:04, 170.24it/s]\u001b[A\n 25%|██▍       | 248/1009 [00:03<00:04, 174.54it/s]\u001b[A\n 26%|██▋       | 267/1009 [00:03<00:04, 176.83it/s]\u001b[A\n 28%|██▊       | 286/1009 [00:03<00:04, 178.76it/s]\u001b[A\n 30%|███       | 305/1009 [00:03<00:03, 180.21it/s]\u001b[A\n 32%|███▏      | 324/1009 [00:04<00:03, 179.83it/s]\u001b[A\n 34%|███▍      | 343/1009 [00:04<00:03, 179.98it/s]\u001b[A\n 36%|███▌      | 362/1009 [00:04<00:03, 181.63it/s]\u001b[A\n 38%|███▊      | 381/1009 [00:04<00:03, 181.04it/s]\u001b[A\n 40%|███▉      | 400/1009 [00:04<00:03, 182.13it/s]\u001b[A\n 42%|████▏     | 419/1009 [00:04<00:03, 182.37it/s]\u001b[A\n 43%|████▎     | 438/1009 [00:04<00:03, 183.85it/s]\u001b[A\n 45%|████▌     | 457/1009 [00:04<00:02, 184.98it/s]\u001b[A\n 47%|████▋     | 476/1009 [00:04<00:02, 185.91it/s]\u001b[A\n 49%|████▉     | 495/1009 [00:04<00:02, 186.25it/s]\u001b[A\n 51%|█████     | 514/1009 [00:05<00:02, 187.04it/s]\u001b[A\n 53%|█████▎    | 533/1009 [00:05<00:02, 187.69it/s]\u001b[A\n 55%|█████▍    | 552/1009 [00:05<00:02, 188.32it/s]\u001b[A\n 57%|█████▋    | 571/1009 [00:05<00:02, 188.13it/s]\u001b[A\n 58%|█████▊    | 590/1009 [00:05<00:02, 188.37it/s]\u001b[A\n 60%|██████    | 609/1009 [00:05<00:02, 188.66it/s]\u001b[A\n 62%|██████▏   | 628/1009 [00:05<00:02, 188.88it/s]\u001b[A\n 64%|██████▍   | 647/1009 [00:05<00:01, 188.97it/s]\u001b[A\n 66%|██████▌   | 666/1009 [00:05<00:01, 188.77it/s]\u001b[A\n 68%|██████▊   | 685/1009 [00:05<00:01, 188.38it/s]\u001b[A\n 70%|██████▉   | 704/1009 [00:06<00:01, 188.63it/s]\u001b[A\n 72%|███████▏  | 723/1009 [00:06<00:01, 188.84it/s]\u001b[A\n 74%|███████▎  | 742/1009 [00:06<00:01, 189.15it/s]\u001b[A\n 75%|███████▌  | 761/1009 [00:06<00:01, 188.98it/s]\u001b[A\n 77%|███████▋  | 780/1009 [00:06<00:01, 189.16it/s]\u001b[A\n 79%|███████▉  | 799/1009 [00:06<00:01, 186.62it/s]\u001b[A\n 81%|████████  | 819/1009 [00:06<00:01, 188.31it/s]\u001b[A\n 83%|████████▎ | 838/1009 [00:06<00:00, 185.22it/s]\u001b[A\n 85%|████████▍ | 857/1009 [00:06<00:00, 186.46it/s]\u001b[A\n 87%|████████▋ | 877/1009 [00:07<00:00, 188.15it/s]\u001b[A\n 89%|████████▉ | 897/1009 [00:07<00:00, 188.85it/s]\u001b[A\n 91%|█████████ | 917/1009 [00:07<00:00, 189.80it/s]\u001b[A\n 93%|█████████▎| 937/1009 [00:07<00:00, 190.56it/s]\u001b[A\n 95%|█████████▍| 957/1009 [00:07<00:00, 190.87it/s]\u001b[A\n 97%|█████████▋| 977/1009 [00:07<00:00, 191.29it/s]\u001b[A\n 99%|█████████▉| 997/1009 [00:07<00:00, 190.46it/s]\u001b[A\n100%|██████████| 1009/1009 [00:07<00:00, 130.99it/s]\nSynthesis audios using bigvgan vocoder...\nLoading Vocoder from Weights file: /src/Amphion/pretrained/bigvgan/400000.pt\nFor predicted mels, #sample = 1...\nModel inference: 14.1s\n100%|██████████| 1/1 [00:17<00:00, 17.56s/it]\n100%|██████████| 1/1 [00:17<00:00, 17.56s/it]\n/src/Amphion/result/source/source_vocalist_l1_TaylorSwift.wav",
  "metrics": {
    "predict_time": 31.512607,
    "total_time": 156.531173
  },
  "output": "https://replicate.delivery/pbxt/FoHobqVw0mrPOluLgRQGEW01GwDo5tvSefKNDyc79hY8AnESA/source_vocalist_l1_TaylorSwift.wav",
  "started_at": "2023-12-21T22:37:17.689070Z",
  "status": "succeeded",
  "urls": {
    "get": "https://api.replicate.com/v1/predictions/h37sr5dbojsspt56c34pvjanoe",
    "cancel": "https://api.replicate.com/v1/predictions/h37sr5dbojsspt56c34pvjanoe/cancel"
  },
  "version": "f29872ee3557e0186735048f1d6de98a52518ae5c49e19453b3fdaad710bdc2b"
}

Generated in

31.5 seconds

Tweak itReport

/tmp/input_audio
vocalist_l1_TaylorSwift
autoshift
getopt: unrecognized option '--diffusion_inference_steps'
Exprimental Configuration File: ckpts/svc/vocalist_l1_contentvec+whisper/args.json
The following values were not passed to `accelerate launch` and had defaults used instead:
`--num_processes` was set to a value of `1`
`--num_machines` was set to a value of `1`
`--mixed_precision` was set to a value of `'no'`
`--dynamo_backend` was set to a value of `'no'`
To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
Monotonic align not found. Please make sure you have compiled it.
There are 1 source audios:
**********
Conversion for source...
Prepare for meta eval data: 0.0s
  0%|          | 0/1 [00:00<?, ?it/s]
  0%|          | 0/1 [00:00<?, ?it/s]
100%|██████████| 1/1 [00:01<00:00,  1.98s/it]
100%|██████████| 1/1 [00:01<00:00,  1.98s/it]
Prepare for acoustic features: 2.0s
Prepare for content features: 0.0s
2023-12-21 22:37:31 | INFO | inference | ========================================================
2023-12-21 22:37:31 | INFO | inference | ||		New inference process started.		||
2023-12-21 22:37:31 | INFO | inference | ========================================================
2023-12-21 22:37:31 | INFO | inference |
2023-12-21 22:37:31 | DEBUG | inference | Using DEBUG logging level.
2023-12-21 22:37:31 | DEBUG | inference | Acoustic dir: ckpts/svc/vocalist_l1_contentvec+whisper
2023-12-21 22:37:31 | DEBUG | inference | Vocoder dir: pretrained/bigvgan
2023-12-21 22:37:31 | DEBUG | inference | Setting random seed done in 0.83ms
2023-12-21 22:37:31 | DEBUG | inference | Random seed: 10086
2023-12-21 22:37:31 | INFO | inference | Building dataset...
2023-12-21 22:37:31 | INFO | inference | Building dataset done in 4.60ms
2023-12-21 22:37:31 | INFO | inference | Building model...
2023-12-21 22:37:31 | INFO | inference | Building model done in 276.183ms
2023-12-21 22:37:31 | INFO | inference | Initializing accelerate...
2023-12-21 22:37:32 | INFO | inference | Initializing accelerate done in 1057.268ms
2023-12-21 22:37:32 | INFO | inference | Loading checkpoint...
2023-12-21 22:37:32 | INFO | accelerate.accelerator | Loading states from ckpts/svc/vocalist_l1_contentvec+whisper/checkpoint/epoch-6852_step-0678447_loss-1.946773
2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All model weights loaded successfully
2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All optimizer states loaded successfully
2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All scheduler states loaded successfully
2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All dataloader sampler states loaded successfully
2023-12-21 22:37:32 | INFO | accelerate.checkpointing | All random states loaded successfully
2023-12-21 22:37:32 | INFO | accelerate.accelerator | Loading in 0 custom states
2023-12-21 22:37:32 | INFO | inference | Loading checkpoint done in 106.015ms
2023-12-21 22:37:32 | INFO | inference | Using PNDM scheduler.
Model Init: 1.5s
Auto transposing: source f0 median = 372.9, target f0 median = 286.9, factor = 0.77
  0%|          | 0/1009 [00:00<?, ?it/s]
  0%|          | 1/1009 [00:02<39:02,  2.32s/it]
  2%|▏         | 20/1009 [00:02<01:26, 11.39it/s]
  4%|▍         | 39/1009 [00:02<00:38, 25.02it/s]
  6%|▌         | 58/1009 [00:02<00:23, 41.23it/s]
  8%|▊         | 77/1009 [00:02<00:15, 59.36it/s]
 10%|▉         | 96/1009 [00:02<00:11, 78.69it/s]
 11%|█▏        | 115/1009 [00:02<00:09, 98.15it/s]
 13%|█▎        | 134/1009 [00:03<00:07, 116.43it/s]
 15%|█▌        | 153/1009 [00:03<00:06, 132.68it/s]
 17%|█▋        | 172/1009 [00:03<00:05, 146.29it/s]
 19%|█▉        | 191/1009 [00:03<00:05, 157.09it/s]
 21%|██        | 210/1009 [00:03<00:04, 164.64it/s]
 23%|██▎       | 229/1009 [00:03<00:04, 170.24it/s]
 25%|██▍       | 248/1009 [00:03<00:04, 174.54it/s]
 26%|██▋       | 267/1009 [00:03<00:04, 176.83it/s]
 28%|██▊       | 286/1009 [00:03<00:04, 178.76it/s]
 30%|███       | 305/1009 [00:03<00:03, 180.21it/s]
 32%|███▏      | 324/1009 [00:04<00:03, 179.83it/s]
 34%|███▍      | 343/1009 [00:04<00:03, 179.98it/s]
 36%|███▌      | 362/1009 [00:04<00:03, 181.63it/s]
 38%|███▊      | 381/1009 [00:04<00:03, 181.04it/s]
 40%|███▉      | 400/1009 [00:04<00:03, 182.13it/s]
 42%|████▏     | 419/1009 [00:04<00:03, 182.37it/s]
 43%|████▎     | 438/1009 [00:04<00:03, 183.85it/s]
 45%|████▌     | 457/1009 [00:04<00:02, 184.98it/s]
 47%|████▋     | 476/1009 [00:04<00:02, 185.91it/s]
 49%|████▉     | 495/1009 [00:04<00:02, 186.25it/s]
 51%|█████     | 514/1009 [00:05<00:02, 187.04it/s]
 53%|█████▎    | 533/1009 [00:05<00:02, 187.69it/s]
 55%|█████▍    | 552/1009 [00:05<00:02, 188.32it/s]
 57%|█████▋    | 571/1009 [00:05<00:02, 188.13it/s]
 58%|█████▊    | 590/1009 [00:05<00:02, 188.37it/s]
 60%|██████    | 609/1009 [00:05<00:02, 188.66it/s]
 62%|██████▏   | 628/1009 [00:05<00:02, 188.88it/s]
 64%|██████▍   | 647/1009 [00:05<00:01, 188.97it/s]
 66%|██████▌   | 666/1009 [00:05<00:01, 188.77it/s]
 68%|██████▊   | 685/1009 [00:05<00:01, 188.38it/s]
 70%|██████▉   | 704/1009 [00:06<00:01, 188.63it/s]
 72%|███████▏  | 723/1009 [00:06<00:01, 188.84it/s]
 74%|███████▎  | 742/1009 [00:06<00:01, 189.15it/s]
 75%|███████▌  | 761/1009 [00:06<00:01, 188.98it/s]
 77%|███████▋  | 780/1009 [00:06<00:01, 189.16it/s]
 79%|███████▉  | 799/1009 [00:06<00:01, 186.62it/s]
 81%|████████  | 819/1009 [00:06<00:01, 188.31it/s]
 83%|████████▎ | 838/1009 [00:06<00:00, 185.22it/s]
 85%|████████▍ | 857/1009 [00:06<00:00, 186.46it/s]
 87%|████████▋ | 877/1009 [00:07<00:00, 188.15it/s]
 89%|████████▉ | 897/1009 [00:07<00:00, 188.85it/s]
 91%|█████████ | 917/1009 [00:07<00:00, 189.80it/s]
 93%|█████████▎| 937/1009 [00:07<00:00, 190.56it/s]
 95%|█████████▍| 957/1009 [00:07<00:00, 190.87it/s]
 97%|█████████▋| 977/1009 [00:07<00:00, 191.29it/s]
 99%|█████████▉| 997/1009 [00:07<00:00, 190.46it/s]
100%|██████████| 1009/1009 [00:07<00:00, 130.99it/s]
Synthesis audios using bigvgan vocoder...
Loading Vocoder from Weights file: /src/Amphion/pretrained/bigvgan/400000.pt
For predicted mels, #sample = 1...
Model inference: 14.1s
100%|██████████| 1/1 [00:17<00:00, 17.56s/it]
100%|██████████| 1/1 [00:17<00:00, 17.56s/it]
/src/Amphion/result/source/source_vocalist_l1_TaylorSwift.wav

Run time and cost

This model costs approximately $0.11 to run on Replicate, or 9 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 L40S GPU hardware. Predictions typically complete within 113 seconds. The predict time for this model varies significantly based on the inputs.

Readme

Implementation of the Hugginface Space: amphion/singing_voice_conversion

Amphion Singing Voice Conversion Pretrained Models

We provide a DiffWaveNetSVC pretrained checkpoint for you to play. Specially, it is trained under the real-world vocalist data (total duration: 6.16 hours), including the following 15 professional singers:

Singers:

Adele
John Mayer
Bruno Mars
Beyonce
Michael Jackson
Taylor Swift
David Tao 陶喆
Eason Chan 陈奕迅
Feng Wang 汪峰
Jian Li 李健
Ying Na 那英
Yijie Shi 石倚洁
Jacky Cheung 张学友
Faye Wong 王菲
Tsai Chin 蔡琴

@article{zhang2023leveraging,
  title={Leveraging Content-based Features from Multiple Acoustic Models for Singing Voice Conversion},
  author={Zhang, Xueyao and Gu, Yicheng and Chen, Haopeng and Fang, Zihao and Zou, Lexiao and Xue, Liumeng and Wu, Zhizheng},
  journal={Machine Learning for Audio Worshop, NeurIPS 2023},
  year={2023}
}