Wavenext-encodec
Model Details
Model Description
Wavenext is a modification of Vocos, where the last ISTFT layer is replaced with a a trainable linear layer that can directly predict speech waveform samples.
This version of Wavenext uses encodec tokens as input features, it's trained using the following bandwidths from encodec (1.5, 3.0, 6.0, 12.0) .
Intended Uses and limitations
The model is aimed to serve as a vocoder to synthesize audio waveforms from encodec discrete codes. Is trained to generate speech and if is used in other audio domain is possible that the model won't produce high quality samples.
Usage
Installation
To use Wavenext only in inference mode, install it using:
pip install git+https://github.com/langtech-bsc/wavenext_pytorch
Reconstruct audio from encodec tokens
You need to provide a bandwidth_id which corresponds to the embedding for bandwidth from the list: [1.5, 3.0, 6.0, 12.0].
import torch
from vocos import Vocos
vocos = Vocos.from_pretrained("BSC-LT/wavenext-encodec")
audio_tokens = torch.randint(low=0, high=1024, size=(8, 200)) # 8 codeboooks, 200 frames
features = vocos.codes_to_features(audio_tokens)
bandwidth_id = torch.tensor([2]) # 6 kbps
audio = vocos.decode(features, bandwidth_id=bandwidth_id)
Copy-synthesis from a file:
import torchaudio
y, sr = torchaudio.load(YOUR_AUDIO_FILE)
if y.size(0) > 1: # mix to mono
y = y.mean(dim=0, keepdim=True)
y = torchaudio.functional.resample(y, orig_freq=sr, new_freq=24000)
y_hat = vocos(y, bandwidth_id=bandwidth_id)
Training Details
Training Data
The model was trained on 4 speech datasets
Dataset | Language | Hours |
---|---|---|
LibriTTS-r | en | 585 |
LJSpeech | en | 24 |
Festcat | ca | 22 |
OpenSLR69 | ca | 5 |
Training Procedure
The model was trained for 1M steps and 99 epochs with a batch size of 16 for stability. We used a Cosine scheduler with a initial learning rate of 1e-4.
Training Hyperparameters
- initial_learning_rate: 1e-4
- scheduler: cosine without warmup or restarts
- mel_loss_coeff: 45
- mrd_loss_coeff: 0.1
- batch_size: 16
- num_samples: 16384
Evaluation
Evaluation was done using the metrics on the original vocos repo, Note that this metrics are calculated using the codecs corresponding to a bandwidth of 1.5 kbps, after 99 epochs we achieve:
- val_loss: 5.52
- f1_score: 0.93
- mel_loss: 0.53
- periodicity_loss:0.14
- pesq_score: 2.12
- pitch_loss: 47.73
- utmos_score: 2.89
Citation
If this code contributes to your research, please cite the work:
@INPROCEEDINGS{10389765,
author={Okamoto, Takuma and Yamashita, Haruki and Ohtani, Yamato and Toda, Tomoki and Kawai, Hisashi},
booktitle={2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)},
title={WaveNeXt: ConvNeXt-Based Fast Neural Vocoder Without ISTFT layer},
year={2023},
volume={},
number={},
pages={1-8},
keywords={Fourier transforms;Vocoders;Conferences;Automatic speech recognition;ConvNext;end-to-end text-to-speech;linear layer-based upsampling;neural vocoder;Vocos},
doi={10.1109/ASRU57964.2023.10389765}}
@article{siuzdak2023vocos,
title={Vocos: Closing the gap between time-domain and Fourier-based neural vocoders for high-quality audio synthesis},
author={Siuzdak, Hubert},
journal={arXiv preprint arXiv:2306.00814},
year={2023}
}
Additional information
Author
The Language Technologies Unit from Barcelona Supercomputing Center.
Contact
For further information, please send an email to [email protected].
Copyright
Copyright(c) 2024 by Language Technologies Unit, Barcelona Supercomputing Center.
License
Funding
This work has been promoted and financed by the Generalitat de Catalunya through the Aina project.
- Downloads last month
- 3
Model tree for BSC-LT/wavenext-encodec
Base model
facebook/encodec_24khz