Retrieval-based Voice Conversion

Start the API Server

Run the server using:

rvc-api

Inference Endpoints

All inference requests are POST requests to /inference using multipart/form-data. You can specify the response type via the res_type query parameter.

Get response as blob

Use res_type=blob to receive the raw audio data.

Get response as JSON

Use res_type=json to receive a JSON object containing the audio and timing information.

# Get as blob
curl -X 'POST' \
      'http://127.0.0.1:8000/inference?res_type=blob' \
      -H 'accept: application/json' \
      -H 'Content-Type: multipart/form-data' \
      -F 'modelpath={model.pth}' \
      -F 'input={input audio path}'

# Get as json(include time)
curl -X 'POST' \
      'http://127.0.0.1:8000/inference?res_type=json' \
      -H 'accept: application/json' \
      -H 'Content-Type: multipart/form-data' \
      -F 'modelpath={model.pth}' \
      -F 'input={input audio path}'

The following options are available for the rvc infer command:

| option        | flag | type         | default value | description |
|---------------|------------|--------------|---------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| modelPath     | -m         | Path         | *required     | Model path or filename (reads in the directory set in env) |
| inputPath     | -i         | Path         | *required     | Input audio path or folder |
| outputPath    | -o         | Path         | *required     | Output audio path or folder |
| sid           | -s         | int          | 0             | Speaker/Singer ID |
| f0_up_key     | -fu        | int          | 0             | Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12) |
| f0_method     | -fm        | str          | rmvpe         | pitch extraction algorithm (pm, harvest, crepe, rmvpe) |
| f0_file       | -ff        | Path \| None | None          | F0 curve file (optional). One pitch per line. Replaces the default F0 and pitch modulation |
| index_file    | -if        | Path \| None | None          | Path to the feature index file |
| index_rate    | -if        | float        | 0.75          | Search feature ratio (controls accent strength, too high has artifacting) |
| filter_radius | -fr        | int          | 3             | If >=3: apply median filtering to the harvested pitch results. The value represents the filter radius and can reduce breathiness |
| resample_sr   | -rsr       | int          | 0             | Resample the output audio in post-processing to the final sample rate. Set to 0 for no resampling |
| rms_mix_rate  | -rmr       | float        | 0.25          | Adjust the volume envelope scaling. Closer to 0, the more it mimicks the volume of the original vocals. Closer to 1 will be a more of a consistently loud volume |
| protect       | -p         | float        | 0.33          | Protect voiceless consonants and breath sounds to prevent artifacts such as tearing in electronic music. Set to 0.5 to disable. Decrease the value to increase protection, but it may reduce indexing accuracy |

Full list of command-line options available for the infer command.

--modelPath, -m (str, required): Model path or filename (reads in the directory set in env)
--inputPath, -i (Path, required): input audio path or folder
--outputPath, -o (Path, required): output audio path or folder
--sid, -s (int): Speaker/Singer id (default: 0)
--f0upkey, -fu (int): Transpose (default: 0)
--f0method, -fm (str): Pitch extraction algorith (default: "rmvpe")
--f0file, -ff (Path): F0 curve file (optional)
--indexFile, -if (Path): Feature index file
--indexRate, -ir (float): Search feature ratio (default: 0.75)
--filterRadius, -fr (int): Apply median filtering (default: 3)
--resamplesr, -rsr (int): Resample the output audio (default: 0)
--rmsmixrate, -rmr (float): Adjust the volume envelope scaling (default: 0.25)
--protect, -p (float): Protect voiceless consonants and breath sounds (default: 0.33)