https://github.com/x-lance/unicats-ctx-vec2wav
[AAAI 2024] Code for CTX-vec2wav in UniCATS
https://github.com/x-lance/unicats-ctx-vec2wav
self-supervised-speech semantic-token speech-synthesis unicats vocoder vocoding
Last synced: 6 months ago
JSON representation
[AAAI 2024] Code for CTX-vec2wav in UniCATS
- Host: GitHub
- URL: https://github.com/x-lance/unicats-ctx-vec2wav
- Owner: X-LANCE
- Created: 2023-08-20T20:27:46.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2024-06-11T15:36:24.000Z (12 months ago)
- Last Synced: 2024-12-05T11:24:25.904Z (6 months ago)
- Topics: self-supervised-speech, semantic-token, speech-synthesis, unicats, vocoder, vocoding
- Language: Python
- Homepage: https://cpdu.github.io/unicats/
- Size: 1000 KB
- Stars: 123
- Watchers: 10
- Forks: 16
- Open Issues: 6
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# CTX-vec2wav, the Acoustic Context-Aware Vocoder
> This is the official implementation of **CTX-vec2wav** vocoder in the AAAI-2024 paper [UniCATS: A Unified Context-Aware Text-to-Speech Framework with Contextual VQ-Diffusion and Vocoding](https://arxiv.org/abs/2306.07547).
See also: [The official implementation of CTX-txt2vec](https://github.com/X-LANCE/UniCATS-CTX-txt2vec), the acoustic model with contextual VQ-diffusion, proposed in the paper.
## Environment Setup
This repo is tested on **python 3.9** on Linux. You can set up the environment with conda
```shell
# Install required packages
conda create -n ctxv2w python=3.9 # or any name you like
conda activate ctxv2w
pip install -r requirements.txt
# Then, set PATH and PYTHONPATH
source path.sh # change the env name if you don't use "ctxv2w"
```
The scripts in `utils/` should be executable. You can run `chmod +x utils/*` to ensure this.
The following process will also need `bash` and `perl` commands in your Linux environment.
## Inference (Vocoding with acoustic context)
For utterances that are already registered in `data/`, the inference (VQ index + acoustic prompt) can be done by
```shell
bash run.sh --stage 3 --stop_stage 3
# You can specify the dataset to be constructed by "--eval_set $which_set", i.e. "--eval_set dev_all"
```
You can also create a subset can perform inference on it by
```shell
subset_data_dir.sh data/eval_all 200 data/eval_subset # randomly select 200 utts from data/eval_all
bash run.sh --stage 3 --stop_stage 3 --eval_set "eval_subset"
```
The program loads the latest checkpoint in the experiment dir `exp/train_all_ctxv2w.v1/*pkl`.
💡Note: the stage 3 in `run.sh` automatically selects the prompt for each utterance by random (see `local/build_prompt_feat.py`).
You can customize this process and perform inference yourself:
1. Prepare a `feats.scp` that specifies each utterance (for inference) with its VQ index sequence in `(L, 2)` shape.
2. Run `feat-to-len.py scp:/path/to/feats.scp > /path/to/utt2num_frames`.
3. Prepare a `prompt.scp` that specifies each utterance with its acoustic (mel) prompt in `(L', 80)` shape.
4. Run inference via
```shell
# might change sampling rate.
decode.py \
--sampling-rate 16000 \
--feats-scp /path/to/feats.scp \
--prompt-scp /path/to/prompt.scp \
--num-frames /path/to/utt2num_frames \
--config /path/to/config.yaml \
--vq-codebook /path/to/codebook.npy \
--checkpoint /path/to/checkpoint \
--outdir /path/to/output/wav
```
## Training
First, you need to properly construct `data` and `feats` directory. Please check out [data_prep](data_prep.md) for details.
> 💡Note: here we provide the 16kHz version of model and data. Meanwhile, the original paper uses 24kHz data, which was accomplished by using the features extracted in 16kHz and increasing the `upsample_scales` in the config yaml.
Then, training on LibriTTS (all training partitions) can be done by
```shell
bash run.sh --stage 2 --stop_stage 2
# You can provide different config file by --conf $your_config
# Checkout run.sh for all the parameters. You can specify every bash variable there as "--key value" in CLI.
```
This will create `exp/train_all_ctxv2w.v1` for logging. The script **automatically handles multi-GPU training** if you specify the $CUDA_VISIBLE_DEVICES env variable.
## Pre-trained model parameters
We release two versions of model parameters (generator) on LibriTTS train-all set. These refer to two sampling rates of the target waveforms.
Note that the acoustic features (fbank+ppe) are all extracted from 16k waveform. The only difference is the upsample rate in the HifiGAN generator.
* 16k: [link](https://huggingface.co/cantabile-kwok/ctx_vec2wav_libritts_all/resolve/main/ctx_v2w.pkl?download=true)
* 24k: [link](https://huggingface.co/cantabile-kwok/ctx_vec2wav_libritts_all/resolve/main/ctx_v2w_24k.pkl?download=true) with the corresponding config file `conf/ctxv2w.24k.yaml`.
The usage is the same as the "Inference" section. Feel free to use these checkpoints for vocoding!
**CMVN file**: in order to perform inference on out-of-set utterances, we provide the `cmvn.ark` file [here](https://huggingface.co/cantabile-kwok/ctx_vec2wav_libritts_all/resolve/main/cmvn.ark). You should extract mel-spectrogram, normalize by that file (computed on LibriTTS), and then feed the model.
## Acknowledgement
During the development, the following repositories were referred to:
* [ESPnet](https://github.com/espnet/espnet), for most network modules in `ctx_vec2wav/models/conformer`.
* [Kaldi](https://github.com/kaldi-asr/kaldi), for most utility scripts in `utils/`.
* [ParallelWaveGAN](https://github.com/kan-bayashi/ParallelWaveGAN), whose training and decoding pipeline is adopted.
## Citation
```
@article{du2023unicats,
title={UniCATS: A Unified Context-Aware Text-to-Speech Framework with Contextual VQ-Diffusion and Vocoding},
author={Du, Chenpeng and Guo, Yiwei and Shen, Feiyu and Liu, Zhijun and Liang, Zheng and Chen, Xie and Wang, Shuai and Zhang, Hui and Yu, Kai},
journal={arXiv preprint arXiv:2306.07547},
year={2023}
}
```