You're looking at a specific version of this model. Jump to the model overview.

replicate /rvc-trainer:5eb9fc56

Input schema

The fields you can use to run this model with an API. If you don’t give a value for a field its default value will be used.

Field Type Default value Description
training_data
string
ZIP or tar archive containing the artist audio used for training
huggingface_dataset_id
string
Hugging Face dataset containing audio files or audio archives; use instead of training_data
dataset_revision
string
Optional branch, tag, or commit for huggingface_dataset_id
auto_demux
boolean
True
Automatically isolate vocals before training; disable only for already-isolated vocals
separation_method
None
pymss
Vocal separator used when auto_demux or validation_auto_demux is enabled
pymss_model
string
model_bs_roformer_ep_368_sdr_12.9628
PyMSS catalog model used for vocal separation
demucs_model
None
htdemucs
Demucs model used when separation_method is demucs
model_name
string
RVC Voice Model
Display name stored in the portable RVCPipeline
training_steps
integer
1000

Min: 10

Max: 20000

RVC optimizer steps
batch_size
None
4
Training batch size
learning_rate
number
0.0001

Min: 0.000001

Max: 0.001

RVC learning rate
max_seconds_per_file
number
180

Min: 5

Max: 600

Maximum duration consumed from each training file
build_index
boolean
True
Build the retrieval index included with the trained model
export_webui_model
boolean
True
Also include a classic RVC WebUI .pth, retrieval index, and feature array in the result
validation_data
string
Optional ZIP or tar of audio to convert with the trained model before returning it
validation_dataset_id
string
Optional Hugging Face dataset of validation audio; use instead of validation_data
validation_revision
string
Optional branch, tag, or commit for validation_dataset_id
validation_auto_demux
boolean
True
Isolate and replace vocals in complete validation songs; disable for isolated vocal files
validation_retrieval_strength
number
0.75

Max: 1

Feature-index retrieval strength used for validation conversion
validation_rms_mix_rate
number
1

Max: 1

Preserve the source validation volume envelope at 0 or use the converted envelope at 1
hf_token
string
Hugging Face token for private datasets or optional model publishing
hub_model_id
string
Optional Hugging Face model ID where the resulting RVCPipeline will be published
hub_private
boolean
True
Create the Hugging Face model repository as private when publishing

Output schema

The shape of the response you’ll get when you run this model with an API.

Schema
{'format': 'uri', 'title': 'Output', 'type': 'string'}