so-vits-svc/README.md

533 lines
35 KiB
Markdown
Raw Normal View History

2023-03-12 02:54:00 +00:00
# SoftVC VITS Singing Voice Conversion
2023-03-11 05:23:24 +00:00
2023-03-12 02:46:22 +00:00
[**English**](./README.md) | [**中文简体**](./README_zh_CN.md)
2023-06-21 01:35:47 +00:00
#### ✨ A studio that contains visible f0 editor, speaker mix timeline editor and other features (Where the Onnx models are used) : [MoeVoiceStudio](https://github.com/NaruseMioShirakana/MoeVoiceStudio)
2023-05-18 05:25:25 +00:00
2023-06-21 01:35:47 +00:00
#### ✨ A fork with a greatly improved user interface: [34j/so-vits-svc-fork](https://github.com/34j/so-vits-svc-fork)
2023-03-11 05:23:24 +00:00
2023-03-20 11:07:59 +00:00
#### ✨ A client supports real-time conversion: [w-okada/voice-changer](https://github.com/w-okada/voice-changer)
2023-06-21 01:35:47 +00:00
**This project differs fundamentally from VITS, as it focuses on Singing Voice Conversion (SVC) rather than Text-to-Speech (TTS). In this project, TTS functionality is not supported, and VITS is incapable of performing SVC tasks. It's important to note that the models used in these two projects are not interchangeable or universally applicable.**
2023-05-14 17:15:05 +00:00
## Announcement
2023-06-21 01:35:47 +00:00
The purpose of this project was to enable developers to have their beloved anime characters perform singing tasks. The developers' intention was to focus solely on fictional characters and avoid any involvement of real individuals, anything related to real individuals deviates from the developer's original intention.
2023-05-14 17:15:05 +00:00
## Disclaimer
2023-04-11 13:24:55 +00:00
2023-06-21 01:35:47 +00:00
This project is an open-source, offline endeavor, and all members of SvcDevelopTeam, as well as other developers and maintainers involved (hereinafter referred to as contributors), have no control over the project. The contributors have never provided any form of assistance to any organization or individual, including but not limited to dataset extraction, dataset processing, computing support, training support, inference, and so on. The contributors do not and cannot be aware of the purposes for which users utilize the project. Therefore, any AI models and synthesized audio produced through the training of this project are unrelated to the contributors. Any issues or consequences arising from their use are the sole responsibility of the user.
2023-04-11 13:24:55 +00:00
2023-06-21 01:35:47 +00:00
This project is run completely offline and does not collect any user information or gather user input data. Therefore, contributors to this project are not aware of all user input and models and therefore are not responsible for any user input.
2023-05-14 17:15:05 +00:00
2023-06-21 01:35:47 +00:00
This project serves as a framework only and does not possess speech synthesis functionality by itself. All functionalities require users to train the models independently. Furthermore, this project does not come bundled with any models, and any secondary distributed projects are independent of the contributors of this project.
2023-05-14 17:15:05 +00:00
2023-03-20 11:07:59 +00:00
## 📏 Terms of Use
2023-06-21 01:35:47 +00:00
# Warning: Please ensure that you address any authorization issues related to the dataset on your own. You bear full responsibility for any problems arising from the usage of non-authorized datasets for training, as well as any resulting consequences. The repository and its maintainer, svc develop team, disclaim any association with or liability for the consequences.
2023-03-20 11:07:59 +00:00
2023-06-21 01:35:47 +00:00
1. This project is exclusively established for academic purposes, aiming to facilitate communication and learning. It is not intended for deployment in production environments.
2023-06-30 19:18:32 +00:00
2. Any sovits-based video posted to a video platform must clearly specify in the introduction the input source vocals and audio used for the voice changer conversion, e.g., if you use someone else's video/audio and convert it by separating the vocals as the input source, you must give a clear link to the original video or music; if you use your own vocals or a voice synthesized by another voice synthesis engine as the input source, you must also state this in your introduction.
2023-06-29 17:20:04 +00:00
3. You are solely responsible for any infringement issues caused by the input source and all consequences. When using other commercial vocal synthesis software as an input source, please ensure that you comply with the regulations of that software, noting that the regulations of many vocal synthesis engines explicitly state that they cannot be used to convert input sources!
2023-06-21 01:35:47 +00:00
4. Engaging in illegal activities, as well as religious and political activities, is strictly prohibited when using this project. The project developers vehemently oppose the aforementioned activities. If you disagree with this provision, the usage of the project is prohibited.
2023-06-29 18:01:47 +00:00
5. If you continue to use the program, you will be deemed to have agreed to the terms and conditions set forth in README and README has discouraged you and is not responsible for any subsequent problems.
2023-06-21 01:35:47 +00:00
6. If you intend to employ this project for any other purposes, kindly contact and inform the maintainers of this repository in advance.
2023-03-11 05:23:24 +00:00
2023-03-20 11:07:59 +00:00
## 📝 Model Introduction
2023-03-11 05:23:24 +00:00
2023-06-28 12:10:10 +00:00
The singing voice conversion model uses SoftVC content encoder to extract speech features from the source audio. These feature vectors are directly fed into VITS without the need for conversion to a text-based intermediate representation. As a result, the pitch and intonations of the original audio are preserved. Meanwhile, the vocoder was replaced with [NSF HiFiGAN](https://github.com/openvpi/DiffSinger/tree/refactor/modules/nsf_hifigan) to solve the problem of sound interruption.
2023-03-11 05:23:24 +00:00
2023-05-22 15:28:53 +00:00
### 🆕 4.1-Stable Version Update Content
2023-04-15 03:33:07 +00:00
2023-06-21 01:35:47 +00:00
- Feature input is changed to the 12th Layer of [Content Vec](https://github.com/auspicious3000/contentvec) Transformer output, And compatible with 4.0 branches.
2023-05-22 15:28:53 +00:00
- Update the shallow diffusion, you can use the shallow diffusion model to improve the sound quality.
2023-06-21 01:35:47 +00:00
- Added Whisper-PPG encoder support
2023-05-29 16:40:29 +00:00
- Added static/dynamic sound fusion
- Added loudness embedding
2023-06-27 14:08:18 +00:00
- Added Functionality of feature retrieval from [RVC](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)
2023-05-25 15:34:51 +00:00
2023-05-18 02:45:46 +00:00
### 🆕 Questions about compatibility with the 4.0 model
2023-05-14 13:43:53 +00:00
2023-06-21 01:35:47 +00:00
- To support the 4.0 model and incorporate the speech encoder, you can make modifications to the `config.json` file. Add the `speech_encoder` field to the "model" section as shown below:
2023-05-14 13:43:53 +00:00
```
"model": {
.........
2023-05-22 15:28:53 +00:00
"ssl_dim": 256,
2023-05-14 13:43:53 +00:00
"n_speakers": 200,
"speech_encoder":"vec256l9"
}
```
2023-06-21 01:35:47 +00:00
### 🆕 Shallow diffusion
2023-05-18 02:45:46 +00:00
![Diagram](shadowdiffusion.png)
2023-06-21 01:35:47 +00:00
## 💬 Python Version
2023-03-24 15:13:44 +00:00
2023-06-21 01:35:47 +00:00
Based on our testing, we have determined that the project runs stable on `Python 3.8.9`.
2023-03-24 15:13:44 +00:00
2023-03-20 11:07:59 +00:00
## 📥 Pre-trained Model Files
2023-03-11 05:23:24 +00:00
2023-03-12 02:46:22 +00:00
#### **Required**
2023-06-21 01:35:47 +00:00
**You need to select one encoder from the list below**
2023-05-14 13:43:53 +00:00
2023-05-29 16:40:29 +00:00
##### **1. If using contentvec as speech encoder(recommended)**
2023-06-04 11:13:08 +00:00
`vec768l12` and `vec256l9` require the encoder
2023-03-11 05:23:24 +00:00
- ContentVec: [checkpoint_best_legacy_500.pt](https://ibm.box.com/s/z1wgl1stco8ffooyatzdwsqn2psd9lrr)
2023-05-14 13:43:53 +00:00
- Place it under the `pretrain` directory
2023-03-11 05:23:24 +00:00
2023-06-03 15:54:28 +00:00
Or download the following ContentVec, which is only 199MB in size but has the same effect:
2023-06-15 17:08:20 +00:00
- ContentVec: [hubert_base.pt](https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/hubert_base.pt)
2023-06-03 15:54:28 +00:00
- Change the file name to `checkpoint_best_legacy_500.pt` and place it in the `pretrain` directory
2023-03-11 05:23:24 +00:00
```shell
# contentvec
2023-05-14 13:43:53 +00:00
wget -P pretrain/ http://obs.cstcloud.cn/share/obs/sankagenkeshi/checkpoint_best_legacy_500.pt
2023-03-11 05:23:24 +00:00
# Alternatively, you can manually download and place it in the hubert directory
```
2023-05-25 15:34:51 +00:00
##### **2. If hubertsoft is used as the speech encoder**
2023-06-15 17:08:20 +00:00
- soft vc hubert: [hubert-soft-0d54a1f4.pt](https://github.com/bshall/hubert/releases/download/v0.1/hubert-soft-0d54a1f4.pt)
2023-05-14 13:43:53 +00:00
- Place it under the `pretrain` directory
2023-05-25 13:37:36 +00:00
##### **3. If whisper-ppg as the encoder**
2023-06-04 11:13:08 +00:00
- download model at [medium.pt](https://openaipublic.azureedge.net/main/whisper/models/345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1/medium.pt), the model fits `whisper-ppg`
- or download model at [large-v2.pt](https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt), the model fits `whisper-ppg-large`
2023-06-28 11:43:01 +00:00
- Place it under the `pretrain` directory
2023-05-25 15:34:51 +00:00
2023-06-01 18:15:42 +00:00
##### **4. If cnhubertlarge as the encoder**
- download model at [chinese-hubert-large-fairseq-ckpt.pt](https://huggingface.co/TencentGameMate/chinese-hubert-large/resolve/main/chinese-hubert-large-fairseq-ckpt.pt)
2023-06-28 11:43:01 +00:00
- Place it under the `pretrain` directory
2023-06-01 18:15:42 +00:00
2023-06-01 18:44:18 +00:00
##### **5. If dphubert as the encoder**
- download model at [DPHuBERT-sp0.75.pth](https://huggingface.co/pyf98/DPHuBERT/resolve/main/DPHuBERT-sp0.75.pth)
2023-06-28 11:43:01 +00:00
- Place it under the `pretrain` directory
2023-06-01 18:44:18 +00:00
2023-06-07 11:22:47 +00:00
##### **6. If WavLM is used as the encoder**
- download model at [WavLM-Base+.pt](https://valle.blob.core.windows.net/share/wavlm/WavLM-Base+.pt?sv=2020-08-04&st=2023-03-01T07%3A51%3A05Z&se=2033-03-02T07%3A51%3A00Z&sr=c&sp=rl&sig=QJXmSJG9DbMKf48UDIU1MfzIro8HQOf3sqlNXiflY1I%3D), the model fits `wavlmbase+`
2023-06-28 11:43:01 +00:00
- Place it under the `pretrain` directory
2023-06-07 11:22:47 +00:00
##### **7. If OnnxHubert/ContentVec as the encoder**
2023-05-29 16:40:29 +00:00
- download model at [MoeSS-SUBModel](https://huggingface.co/NaruseMioShirakana/MoeSS-SUBModel/tree/main)
2023-05-25 13:37:36 +00:00
- Place it under the `pretrain` directory
2023-05-22 16:20:37 +00:00
#### **List of Encoders**
- "vec768l12"
- "vec256l9"
- "vec256l9-onnx"
- "vec256l12-onnx"
- "vec768l9-onnx"
- "vec768l12-onnx"
- "hubertsoft-onnx"
- "hubertsoft"
2023-05-25 13:37:36 +00:00
- "whisper-ppg"
2023-06-01 18:15:42 +00:00
- "cnhubertlarge"
2023-06-01 18:44:18 +00:00
- "dphubert"
2023-06-04 11:13:08 +00:00
- "whisper-ppg-large"
2023-06-09 11:10:29 +00:00
- "wavlmbase+"
2023-05-22 16:20:37 +00:00
2023-03-12 02:46:22 +00:00
#### **Optional(Strongly recommend)**
- Pre-trained model files: `G_0.pth` `D_0.pth`
- Place them under the `logs/44k` directory
2023-05-18 02:45:46 +00:00
- Diffusion model pretraining base model file: `model_0.pt`
- Put it in the `logs/44k/diffusion` directory
2023-05-21 06:57:59 +00:00
Get Sovits Pre-trained model from svc-develop-team(TBD) or anywhere else.
2023-06-21 01:35:47 +00:00
Diffusion model references [Diffusion-SVC](https://github.com/CNChTu/Diffusion-SVC) diffusion model. The pre-trained diffusion model is universal with the DDSP-SVC's. You can go to [Diffusion-SVC](https://github.com/CNChTu/Diffusion-SVC)'s repo to get the pre-trained diffusion model.
2023-03-12 02:46:22 +00:00
2023-06-21 01:35:47 +00:00
While the pretrained model typically does not pose copyright concerns, it is essential to remain vigilant. It is advisable to consult with the author beforehand or carefully review the description to ascertain the permissible usage of the model. This helps ensure compliance with any specified guidelines or restrictions regarding its utilization.
2023-03-12 02:53:36 +00:00
2023-04-08 17:07:12 +00:00
#### **Optional(Select as Required)**
2023-06-21 01:35:47 +00:00
If you are using the `NSF-HIFIGAN enhancer` or `shallow diffusion`, you will need to download the pre-trained NSF-HIFIGAN model.
2023-04-08 17:07:12 +00:00
- Pre-trained NSF-HIFIGAN Vocoder: [nsf_hifigan_20221211.zip](https://github.com/openvpi/vocoders/releases/download/nsf-hifigan-v1/nsf_hifigan_20221211.zip)
2023-04-09 03:36:55 +00:00
- Unzip and place the four files under the `pretrain/nsf_hifigan` directory
2023-04-08 17:07:12 +00:00
```shell
# nsf_hifigan
2023-05-14 13:43:53 +00:00
wget -P pretrain/ https://github.com/openvpi/vocoders/releases/download/nsf-hifigan-v1/nsf_hifigan_20221211.zip
2023-06-28 11:47:36 +00:00
unzip -od pretrain/nsf_hifigan pretrain/nsf_hifigan_20221211.zip
2023-04-08 17:07:12 +00:00
# Alternatively, you can manually download and place it in the pretrain/nsf_hifigan directory
2023-06-15 17:08:20 +00:00
# URL: https://github.com/openvpi/vocoders/releases/tag/nsf-hifigan-v1
2023-04-08 17:07:12 +00:00
```
2023-03-20 11:07:59 +00:00
## 📊 Dataset Preparation
2023-03-11 05:23:24 +00:00
2023-06-21 01:35:47 +00:00
Simply place the dataset in the `dataset_raw` directory with the following file structure:
2023-03-11 05:23:24 +00:00
2023-03-24 08:59:47 +00:00
```
2023-03-11 05:23:24 +00:00
dataset_raw
├───speaker0
│ ├───xxx1-xxx1.wav
│ ├───...
│ └───Lxx-0xx8.wav
└───speaker1
├───xx2-0xxx2.wav
├───...
└───xxx7-xxx007.wav
```
2023-06-21 01:35:47 +00:00
You can customize the speaker's name as showed below:
2023-03-24 08:59:47 +00:00
```
dataset_raw
└───suijiSUI
├───1.wav
├───...
└───25788785-20221210-200143-856_01_(Vocals)_0_0.wav
```
2023-03-20 11:07:59 +00:00
## 🛠️ Preprocessing
2023-03-11 05:23:24 +00:00
2023-04-09 03:41:03 +00:00
### 0. Slice audio
2023-04-09 03:36:55 +00:00
2023-06-29 18:17:40 +00:00
To avoid video memory overflow during training or pre-processing, it is recommended to limit the length of audio clips. Cutting the audio to a length of "5s - 15s" is more recommended. Slightly longer times are acceptable, however, excessively long clips may cause problems such as `torch.cuda.OutOfMemoryError`.
2023-04-09 03:36:55 +00:00
2023-06-21 01:35:47 +00:00
To facilitate the slicing process, you can use [audio-slicer-GUI](https://github.com/flutydeer/audio-slicer) or [audio-slicer-CLI](https://github.com/openvpi/audio-slicer)
2023-04-09 03:36:55 +00:00
2023-06-21 01:35:47 +00:00
In general, only the `Minimum Interval` needs to be adjusted. For spoken audio, the default value usually suffices, while for singing audio, it can be adjusted to around `100` or even `50`, depending on the specific requirements.
2023-04-09 03:36:55 +00:00
2023-06-21 01:35:47 +00:00
After slicing, it is recommended to remove any audio clips that are excessively long or too short.
2023-04-09 03:36:55 +00:00
2023-06-21 01:35:47 +00:00
**If you are using whisper-ppg encoder for training, the audio clips must shorter than 30s.**
2023-05-25 13:48:39 +00:00
2023-04-09 03:41:03 +00:00
### 1. Resample to 44100Hz and mono
2023-03-11 05:23:24 +00:00
```shell
python resample.py
```
2023-06-29 16:51:25 +00:00
#### Cautions
2023-05-25 15:34:51 +00:00
2023-06-30 19:18:32 +00:00
Although this project has resample.py scripts for resampling, mono and loudness matching, the default loudness matching is to match to 0db. This can cause damage to the sound quality. While python's loudness matching package pyloudnorm does not limit the level, this can lead to sonic boom. Therefore, it is recommended to consider using professional sound processing software, such as `adobe audition` for loudness matching. If you are already using other software for loudness matching, add the parameter `-skip_loudnorm` to the run command:
2023-05-25 15:34:51 +00:00
2023-06-02 04:41:25 +00:00
```shell
python resample.py --skip_loudnorm
2023-05-25 15:34:51 +00:00
```
2023-04-09 03:41:03 +00:00
### 2. Automatically split the dataset into training and validation sets, and generate configuration files.
2023-03-11 05:23:24 +00:00
```shell
2023-05-25 15:34:51 +00:00
python preprocess_flist_config.py --speech_encoder vec768l12
2023-03-11 05:23:24 +00:00
```
2023-06-07 11:22:47 +00:00
speech_encoder has the following options
2023-05-14 13:43:53 +00:00
```
vec768l12
vec256l9
hubertsoft
2023-05-25 13:43:02 +00:00
whisper-ppg
2023-06-01 18:15:42 +00:00
cnhubertlarge
2023-06-01 18:44:18 +00:00
dphubert
2023-06-04 11:13:08 +00:00
whisper-ppg-large
2023-06-09 11:10:29 +00:00
wavlmbase+
2023-05-14 13:43:53 +00:00
```
2023-06-21 01:35:47 +00:00
If the speech_encoder argument is omitted, the default value is `vec768l12`
2023-05-14 13:43:53 +00:00
2023-05-29 16:40:29 +00:00
**Use loudness embedding**
2023-05-29 22:35:53 +00:00
Add `--vol_aug` if you want to enable loudness embedding:
```shell
python preprocess_flist_config.py --speech_encoder vec768l12 --vol_aug
```
2023-06-21 01:35:47 +00:00
After enabling loudness embedding, the trained model will match the loudness of the input source; otherwise, it will match the loudness of the training set.
2023-05-29 16:40:29 +00:00
2023-05-30 10:07:07 +00:00
#### You can modify some parameters in the generated config.json and diffusion.yaml
2023-06-21 01:35:47 +00:00
* `keep_ckpts`: Keep the the the number of previous models during training. Set to `0` to keep them all. Default is `3`.
2023-05-30 10:07:07 +00:00
2023-06-15 17:08:20 +00:00
* `all_in_mem`: Load all dataset to RAM. It can be enabled when the disk IO of some platforms is too low and the system memory is **much larger** than your dataset.
2023-05-30 10:07:07 +00:00
2023-06-21 01:35:47 +00:00
* `batch_size`: The amount of data loaded to the GPU for a single training session can be adjusted to a size lower than the GPU memory capacity.
2023-05-30 10:07:07 +00:00
2023-06-15 17:08:20 +00:00
* `vocoder_name`: Select a vocoder. The default is `nsf-hifigan`.
##### diffusion.yaml
* `cache_all_data`: Load all dataset to RAM. It can be enabled when the disk IO of some platforms is too low and the system memory is **much larger** than your dataset.
* `duration`: The duration of the audio slicing during training, can be adjusted according to the size of the video memory, **Note: this value must be less than the minimum time of the audio in the training set!**
* `batch_size`: The amount of data loaded to the GPU for a single training session can be adjusted to a size lower than the video memory capacity.
* `timesteps`: The total number of steps in the diffusion model, which defaults to 1000.
2023-06-15 18:06:51 +00:00
* `k_step_max`: Training can only train `k_step_max` step diffusion to save training time, note that the value must be less than `timesteps`, 0 is to train the entire diffusion model, **Note: if you do not train the entire diffusion model will not be able to use only_diffusion!**
2023-05-30 10:07:07 +00:00
2023-06-02 12:18:16 +00:00
##### **List of Vocoders**
```
nsf-hifigan
nsf-snake-hifigan
```
2023-05-30 10:07:07 +00:00
2023-04-09 03:41:03 +00:00
### 3. Generate hubert and f0
2023-03-11 05:23:24 +00:00
```shell
2023-05-14 13:43:53 +00:00
python preprocess_hubert_f0.py --f0_predictor dio
```
f0_predictor has four options
```
crepe
dio
pm
harvest
2023-03-11 05:23:24 +00:00
```
2023-06-29 18:33:35 +00:00
If the training set is too noisy,it is recommended to use `crepe` to handle f0
2023-05-14 13:43:53 +00:00
2023-06-21 01:35:47 +00:00
If the f0_predictor parameter is omitted, the default value is `dio`
2023-05-14 13:43:53 +00:00
2023-06-21 01:35:47 +00:00
If you want shallow diffusion (optional), you need to add the `--use_diff` parameter, for example:
2023-05-18 02:45:46 +00:00
```shell
python preprocess_hubert_f0.py --f0_predictor dio --use_diff
```
2023-03-11 05:23:24 +00:00
After completing the above steps, the dataset directory will contain the preprocessed data, and the dataset_raw folder can be deleted.
2023-07-01 19:27:38 +00:00
## 🏋️‍ Training
2023-03-11 05:23:24 +00:00
2023-06-15 17:08:20 +00:00
### Sovits Model
2023-05-18 02:45:46 +00:00
```shell
2023-06-15 17:08:20 +00:00
python train.py -c configs/config.json -m 44k
2023-05-18 02:45:46 +00:00
```
2023-06-15 17:08:20 +00:00
### Diffusion Model (optional)
If the shallow diffusion function is needed, the diffusion model needs to be trained. The diffusion model training method is as follows:
2023-05-18 02:45:46 +00:00
2023-03-11 05:23:24 +00:00
```shell
2023-06-15 17:08:20 +00:00
python train_diff.py -c configs/diffusion.yaml
2023-03-11 05:23:24 +00:00
```
2023-06-21 01:35:47 +00:00
During training, the model files will be saved to `logs/44k`, and the diffusion model will be saved to `logs/44k/diffusion`
2023-05-18 02:45:46 +00:00
2023-03-20 11:07:59 +00:00
## 🤖 Inference
2023-03-11 05:23:24 +00:00
Use [inference_main.py](https://github.com/svc-develop-team/so-vits-svc/blob/4.0/inference_main.py)
```shell
# Example
python inference_main.py -m "logs/44k/G_30400.pth" -c "configs/config.json" -n "君の知らない物語-src.wav" -t 0 -s "nen"
```
Required parameters:
2023-04-05 11:01:30 +00:00
- `-m` | `--model_path`: path to the model.
- `-c` | `--config_path`: path to the configuration file.
2023-06-21 01:35:47 +00:00
- `-n` | `--clean_names`: a list of wav file names located in the `raw` folder.
- `-t` | `--trans`: pitch shift, supports positive and negative (semitone) values.
2023-06-26 20:54:48 +00:00
- `-s` | `--spk_list`: Select the speaker ID to use for conversion.
2023-06-21 01:35:47 +00:00
- `-cl` | `--clip`: Forced audio clipping, set to 0 to disable(default), setting it to a non-zero value (duration in seconds) to enable.
2023-03-11 05:23:24 +00:00
Optional parameters: see the next section
2023-04-05 11:01:30 +00:00
- `-lg` | `--linear_gradient`: The cross fade length of two audio slices in seconds. If there is a discontinuous voice after forced slicing, you can adjust this value. Otherwise, it is recommended to use the default value of 0.
2023-06-21 01:35:47 +00:00
- `-f0p` | `--f0_predictor`: Select a F0 predictor, options are `crepe`, `pm`, `dio`, `harvest`, default value is `pm`(note: f0 mean pooling will be enable when using `crepe`)
- `-a` | `--auto_predict_f0`: automatic pitch prediction, do not enable this when converting singing voices as it can cause serious pitch issues.
- `-cm` | `--cluster_model_path`: Cluster model or feature retrieval index path, if left blank, it will be automatically set as the default path of these models. If there is no training cluster or feature retrieval, fill in at will.
2023-05-31 18:45:01 +00:00
- `-cr` | `--cluster_infer_ratio`: The proportion of clustering scheme or feature retrieval ranges from 0 to 1. If there is no training clustering model or feature retrieval, the default is 0.
2023-06-21 01:35:47 +00:00
- `-eh` | `--enhance`: Whether to use NSF_HIFIGAN enhancer, this option has certain effect on sound quality enhancement for some models with few training sets, but has negative effect on well-trained models, so it is disabled by default.
- `-shd` | `--shallow_diffusion`: Whether to use shallow diffusion, which can solve some electrical sound problems after use. This option is disabled by default. When this option is enabled, NSF_HIFIGAN enhancer will be disabled
- `-usm` | `--use_spk_mix`: whether to use dynamic voice fusion
- `-lea` | `--loudness_envelope_adjustment`The adjustment of the input source's loudness envelope in relation to the fusion ratio of the output loudness envelope. The closer to 1, the more the output loudness envelope is used
- `-fr` | `--feature_retrieval`Whether to use feature retrieval If clustering model is used, it will be disabled, and `cm` and `cr` parameters will become the index path and mixing ratio of feature retrieval
2023-05-29 16:40:29 +00:00
2023-05-23 07:37:20 +00:00
Shallow diffusion settings:
2023-05-29 16:40:29 +00:00
- `-dm` | `--diffusion_model_path`: Diffusion model path
2023-06-21 01:35:47 +00:00
- `-dc` | `--diffusion_config_path`: Diffusion config file path
- `-ks` | `--k_step`: The larger the number of k_steps, the closer it is to the result of the diffusion model. The default is 100
- `-od` | `--only_diffusion`: Whether to use Only diffusion mode, which does not load the sovits model to only use diffusion model inference
- `-se` | `--second_encoding`which involves applying an additional encoding to the original audio before shallow diffusion. This option can yield varying results - sometimes positive and sometimes negative.
2023-06-29 16:51:25 +00:00
### Cautions
2023-05-25 15:42:03 +00:00
2023-06-21 01:35:47 +00:00
If inferencing using `whisper-ppg` speech encoder, you need to set `--clip` to 25 and `-lg` to 1. Otherwise it will fail to infer properly.
2023-05-25 15:42:03 +00:00
2023-03-20 11:07:59 +00:00
## 🤔 Optional Settings
2023-03-11 05:23:24 +00:00
2023-06-29 18:33:35 +00:00
If you are satisfied with the previous results, or if you do not feel you understand what follows, you can skip it and it will have no effect on the use of the model. The impact of these optional settings mentioned is relatively small, and while they may have some impact on specific datasets, in most cases the difference may not be significant.
2023-03-11 05:23:24 +00:00
### Automatic f0 prediction
2023-06-21 01:35:47 +00:00
During the training of the 4.0 model, an f0 predictor is also trained, which enables automatic pitch prediction during voice conversion. However, if the results are not satisfactory, manual pitch prediction can be used instead. Please note that when converting singing voices, it is advised not to enable this feature as it may cause significant pitch shifting.
- Set `auto_predict_f0` to `true` in `inference_main.py`.
2023-03-11 05:23:24 +00:00
### Cluster-based timbre leakage control
2023-06-21 01:35:47 +00:00
Introduction: The clustering scheme implemented in this model aims to reduce timbre leakage and enhance the similarity of the trained model to the target's timbre, although the effect may not be very pronounced. However, relying solely on clustering can reduce the model's clarity and make it sound less distinct. Therefore, a fusion method is adopted in this model to control the balance between the clustering and non-clustering approaches. This allows manual adjustment of the trade-off between "sounding like the target's timbre" and "have clear enunciation" to find an optimal balance.
2023-03-11 05:23:24 +00:00
2023-06-21 01:35:47 +00:00
No changes are required in the existing steps. Simply train an additional clustering model, which incurs relatively low training costs.
2023-03-11 05:23:24 +00:00
- Training process:
2023-06-21 01:35:47 +00:00
- Train on a machine with good CPU performance. According to extant experience, it takes about 4 minutes to train each speaker on a Tencent Cloud machine with 6-core CPU.
2023-04-09 03:36:55 +00:00
- Execute `python cluster/train_cluster.py`. The output model will be saved in `logs/44k/kmeans_10000.pt`.
2023-05-18 13:16:25 +00:00
- The clustering model can currently be trained using the gpu by executing `python cluster/train_cluster.py --gpu`
2023-03-11 05:23:24 +00:00
- Inference process:
- Specify `cluster_model_path` in `inference_main.py`. If not specified, the default is `logs/44k/kmeans_10000.pt`.
2023-04-09 03:36:55 +00:00
- Specify `cluster_infer_ratio` in `inference_main.py`, where `0` means not using clustering at all, `1` means only using clustering, and usually `0.5` is sufficient.
2023-03-11 05:23:24 +00:00
2023-05-31 18:45:01 +00:00
### Feature retrieval
2023-06-21 01:35:47 +00:00
Introduction: As with the clustering scheme, the timbre leakage can be reduced, the enunciation is slightly better than clustering, but it will reduce the inference speed. By employing the fusion method, it becomes possible to linearly control the balance between feature retrieval and non-feature retrieval, allowing for fine-tuning of the desired proportion.
2023-05-31 18:45:01 +00:00
2023-06-15 17:08:20 +00:00
- Training process:
First, it needs to be executed after generating hubert and f0:
2023-05-31 18:45:01 +00:00
```shell
python train_index.py -c configs/config.json
```
The output of the model will be in `logs/44k/feature_and_index.pkl`
2023-06-15 17:08:20 +00:00
- Inference process:
2023-05-31 18:45:01 +00:00
- The `--feature_retrieval` needs to be formulated first, and the clustering mode automatically switches to the feature retrieval mode.
- Specify `cluster_model_path` in `inference_main.py`. If not specified, the default is `logs/44k/feature_and_index.pkl`.
2023-05-31 18:45:01 +00:00
- Specify `cluster_infer_ratio` in `inference_main.py`, where `0` means not using feature retrieval at all, `1` means only using feature retrieval, and usually `0.5` is sufficient.
2023-05-23 07:37:20 +00:00
### [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/svc-develop-team/so-vits-svc/blob/4.1-Stable/sovits4_for_colab.ipynb) [sovits4_for_colab.ipynb](https://colab.research.google.com/github/svc-develop-team/so-vits-svc/blob/4.1-Stable/sovits4_for_colab.ipynb)
2023-03-12 03:24:13 +00:00
2023-06-21 01:35:47 +00:00
## 🗜️ Model compression
2023-05-25 06:11:11 +00:00
2023-06-21 01:35:47 +00:00
The generated model contains data that is needed for further training. If you confirm that the model is final and not be used in further training, it is safe to remove these data to get smaller file size (about 1/3).
2023-05-25 06:11:11 +00:00
```shell
# Example
python compress_model.py -c="configs/config.json" -i="logs/44k/G_30400.pth" -o="logs/44k/release.pth"
```
2023-05-29 16:40:29 +00:00
## 👨‍🔧 Timbre mixing
2023-06-26 21:02:14 +00:00
### Static Tone Mixing
2023-05-29 16:40:29 +00:00
2023-06-27 09:19:15 +00:00
**Refer to `webUI.py` file for stable Timbre mixing of the gadget/lab feature.**
2023-05-29 16:40:29 +00:00
2023-06-21 01:39:46 +00:00
Introduction: This function can combine multiple models into one model (convex combination or linear combination of multiple model parameters) to create mixed voice that do not exist in reality
2023-05-29 16:40:29 +00:00
**Note:**
2023-06-26 20:56:39 +00:00
1. This feature is only supported for single-speaker models
2023-06-26 21:02:14 +00:00
2. If you force a multi-speaker model, it is critical to make sure there are the same number of speakers in each model. This will ensure that sounds with the same SpeakerID can be mixed correctly.
2023-06-21 01:39:46 +00:00
3. Ensure that the `model` fields in config.json of all models to be mixed are the same
4. The mixed model can use any config.json file from the models being synthesized. However, the clustering model will not be functional after mixed.
2023-05-29 16:40:29 +00:00
5. When batch uploading models, it is best to put the models into a folder and upload them together after selecting them
6. It is suggested to adjust the mixing ratio between 0 and 100, or to other numbers, but unknown effects will occur in the linear combination mode
7. After mixing, the file named output.pth will be saved in the root directory of the project
8. Convex combination mode will perform Softmax to add the mix ratio to 1, while linear combination mode will not
### Dynamic timbre mixing
**Refer to the `spkmix.py` file for an introduction to dynamic timbre mixing**
Character mix track writing rules:
2023-05-29 17:48:41 +00:00
2023-05-29 16:40:29 +00:00
Role ID: \[\[Start time 1, end time 1, start value 1, start value 1], [Start time 2, end time 2, start value 2]]
2023-05-29 17:48:41 +00:00
2023-05-29 16:40:29 +00:00
The start time must be the same as the end time of the previous one. The first start time must be 0, and the last end time must be 1 (time ranges from 0 to 1).
2023-05-29 17:48:41 +00:00
2023-05-29 16:40:29 +00:00
All roles must be filled in. For unused roles, fill \[\[0., 1., 0., 0.]]
2023-05-29 17:48:41 +00:00
The fusion value can be filled in arbitrarily, and the linear change from the start value to the end value within the specified period of time. The
internal linear combination will be automatically guaranteed to be 1 (convex combination condition), so it can be used safely
2023-05-29 16:40:29 +00:00
Use the `--use_spk_mix` parameter when reasoning to enable dynamic timbre mixing
2023-03-20 11:07:59 +00:00
## 📤 Exporting to Onnx
2023-03-11 05:23:24 +00:00
Use [onnx_export.py](https://github.com/svc-develop-team/so-vits-svc/blob/4.0/onnx_export.py)
2023-03-16 14:57:05 +00:00
- Create a folder named `checkpoints` and open it
- Create a folder in the `checkpoints` folder as your project folder, naming it after your project, for example `aziplayer`
- Rename your model as `model.pth`, the configuration file as `config.json`, and place them in the `aziplayer` folder you just created
2023-05-18 05:19:48 +00:00
- Modify `"NyaruTaffy"` in `path = "NyaruTaffy"` in [onnx_export.py](https://github.com/svc-develop-team/so-vits-svc/blob/4.0/onnx_export.py) to your project name, `path = "aziplayer"`onnx_export_speaker_mix makes you can mix speaker's voice
2023-03-16 14:57:05 +00:00
- Run [onnx_export.py](https://github.com/svc-develop-team/so-vits-svc/blob/4.0/onnx_export.py)
2023-03-11 05:23:24 +00:00
- Wait for it to finish running. A `model.onnx` will be generated in your project folder, which is the exported model.
2023-04-09 03:36:55 +00:00
Note: For Hubert Onnx models, please use the models provided by MoeSS. Currently, they cannot be exported on their own (Hubert in fairseq has many unsupported operators and things involving constants that can cause errors or result in problems with the input/output shape and results when exported.)
2023-03-11 05:23:24 +00:00
2023-06-14 13:40:34 +00:00
## 📎 Reference
| URL | Designation | Title | Implementation Source |
| --- | ----------- | ----- | --------------------- |
|[2106.06103](https://arxiv.org/abs/2106.06103) | VITS (Synthesizer)| Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech | [jaywalnut310/vits](https://github.com/jaywalnut310/vits) |
|[2111.02392](https://arxiv.org/abs/2111.02392) | SoftVC (Speech Encoder)| A Comparison of Discrete and Soft Speech Units for Improved Voice Conversion | [bshall/hubert](https://github.com/bshall/hubert) |
|[2204.09224](https://arxiv.org/abs/2204.09224) | ContentVec (Speech Encoder)| ContentVec: An Improved Self-Supervised Speech Representation by Disentangling Speakers | [auspicious3000/contentvec](https://github.com/auspicious3000/contentvec) |
|[2212.04356](https://arxiv.org/abs/2212.04356) | Whisper (Speech Encoder) | Robust Speech Recognition via Large-Scale Weak Supervision | [openai/whisper](https://github.com/openai/whisper) |
|[2110.13900](https://arxiv.org/abs/2110.13900) | WavLM (Speech Encoder) | WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing | [microsoft/unilm/wavlm](https://github.com/microsoft/unilm/tree/master/wavlm) |
|[2305.17651](https://arxiv.org/abs/2305.17651) | DPHubert (Speech Encoder) | DPHuBERT: Joint Distillation and Pruning of Self-Supervised Speech Models | [pyf98/DPHuBERT](https://github.com/pyf98/DPHuBERT) |
|[DOI:10.21437/Interspeech.2017-68](http://dx.doi.org/10.21437/Interspeech.2017-68) | Harvest (F0 Predictor) | Harvest: A high-performance fundamental frequency estimator from speech signals | [mmorise/World/harvest](https://github.com/mmorise/World/blob/master/src/harvest.cpp) |
|[aes35-000039](https://www.aes.org/e-lib/online/browse.cfm?elib=15165) | Dio (F0 Predictor) | Fast and reliable F0 estimation method based on the period extraction of vocal fold vibration of singing voice and speech | [mmorise/World/dio](https://github.com/mmorise/World/blob/master/src/dio.cpp) |
|[8461329](https://ieeexplore.ieee.org/document/8461329) | Crepe (F0 Predictor) | Crepe: A Convolutional Representation for Pitch Estimation | [maxrmorrison/torchcrepe](https://github.com/maxrmorrison/torchcrepe) |
|[DOI:10.1016/j.wocn.2018.07.001](https://doi.org/10.1016/j.wocn.2018.07.001) | Parselmouth (F0 Predictor) | Introducing Parselmouth: A Python interface to Praat | [YannickJadoul/Parselmouth](https://github.com/YannickJadoul/Parselmouth) |
|[2010.05646](https://arxiv.org/abs/2010.05646) | HIFIGAN (Vocoder) | HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis | [jik876/hifi-gan](https://github.com/jik876/hifi-gan) |
|[1810.11946](https://arxiv.org/abs/1810.11946.pdf) | NSF (Vocoder) | Neural source-filter-based waveform model for statistical parametric speech synthesis | [openvpi/DiffSinger/modules/nsf_hifigan](https://github.com/openvpi/DiffSinger/tree/refactor/modules/nsf_hifigan)
|[2006.08195](https://arxiv.org/abs/2006.08195) | Snake (Vocoder) | Neural Networks Fail to Learn Periodic Functions and How to Fix It | [EdwardDixon/snake](https://github.com/EdwardDixon/snake)
|[2105.02446v3](https://arxiv.org/abs/2105.02446v3) | Shallow Diffusion (PostProcessing)| DiffSinger: Singing Voice Synthesis via Shallow Diffusion Mechanism | [CNChTu/Diffusion-SVC](https://github.com/CNChTu/Diffusion-SVC) |
|[K-means](https://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=01D65490BADCC216F350D06F84D721AD?doi=10.1.1.308.8619&rep=rep1&type=pdf) | Feature K-means Clustering (PreProcessing)| Some methods for classification and analysis of multivariate observations | This repo |
| | Feature TopK Retrieval (PreProcessing)| Retrieval based Voice Conversion | [RVC-Project/Retrieval-based-Voice-Conversion-WebUI](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI) |
2023-03-24 08:59:47 +00:00
## ☀️ Previous contributors
2023-03-22 13:04:52 +00:00
For some reason the author deleted the original repository. Because of the negligence of the organization members, the contributor list was cleared because all files were directly reuploaded to this repository at the beginning of the reconstruction of this repository. Now add a previous contributor list to README.md.
*Some members have not listed according to their personal wishes.*
<table>
<tr>
<td align="center"><a href="https://github.com/MistEO"><img src="https://avatars.githubusercontent.com/u/18511905?v=4" width="100px;" alt=""/><br /><sub><b>MistEO</b></sub></a><br /></td>
<td align="center"><a href="https://github.com/XiaoMiku01"><img src="https://avatars.githubusercontent.com/u/54094119?v=4" width="100px;" alt=""/><br /><sub><b>XiaoMiku01</b></sub></a><br /></td>
<td align="center"><a href="https://github.com/ForsakenRei"><img src="https://avatars.githubusercontent.com/u/23041178?v=4" width="100px;" alt=""/><br /><sub><b>しぐれ</b></sub></a><br /></td>
<td align="center"><a href="https://github.com/TomoGaSukunai"><img src="https://avatars.githubusercontent.com/u/25863522?v=4" width="100px;" alt=""/><br /><sub><b>TomoGaSukunai</b></sub></a><br /></td>
<td align="center"><a href="https://github.com/Plachtaa"><img src="https://avatars.githubusercontent.com/u/112609742?v=4" width="100px;" alt=""/><br /><sub><b>Plachtaa</b></sub></a><br /></td>
<td align="center"><a href="https://github.com/zdxiaoda"><img src="https://avatars.githubusercontent.com/u/45501959?v=4" width="100px;" alt=""/><br /><sub><b>zd小达</b></sub></a><br /></td>
<td align="center"><a href="https://github.com/Archivoice"><img src="https://avatars.githubusercontent.com/u/107520869?v=4" width="100px;" alt=""/><br /><sub><b>凍聲響世</b></sub></a><br /></td>
</tr>
</table>
2023-03-20 11:07:59 +00:00
## 📚 Some legal provisions for reference
2023-03-11 05:23:24 +00:00
2023-03-22 12:47:48 +00:00
#### Any country, region, organization, or individual using this project must comply with the following laws.
2023-03-11 05:23:24 +00:00
#### 《民法典》
##### 第一千零一十九条
2023-04-04 14:44:55 +00:00
任何组织或者个人不得以丑化、污损,或者利用信息技术手段伪造等方式侵害他人的肖像权。未经肖像权人同意,不得制作、使用、公开肖像权人的肖像,但是法律另有规定的除外。未经肖像权人同意,肖像作品权利人不得以发表、复制、发行、出租、展览等方式使用或者公开肖像权人的肖像。对自然人声音的保护,参照适用肖像权保护的有关规定。
2023-03-11 05:23:24 +00:00
##### 第一千零二十四条
2023-03-12 02:46:22 +00:00
【名誉权】民事主体享有名誉权。任何组织或者个人不得以侮辱、诽谤等方式侵害他人的名誉权。
2023-03-11 05:23:24 +00:00
##### 第一千零二十七条
2023-04-04 14:44:55 +00:00
【作品侵害名誉权】行为人发表的文学、艺术作品以真人真事或者特定人为描述对象,含有侮辱、诽谤内容,侵害他人名誉权的,受害人有权依法请求该行为人承担民事责任。行为人发表的文学、艺术作品不以特定人为描述对象,仅其中的情节与该特定人的情况相似的,不承担民事责任。
2023-03-22 12:47:48 +00:00
#### 《[中华人民共和国宪法](http://www.gov.cn/guoqing/2018-03/22/content_5276318.htm)》
#### 《[中华人民共和国刑法](http://gongbao.court.gov.cn/Details/f8e30d0689b23f57bfc782d21035c3.html?sw=%E4%B8%AD%E5%8D%8E%E4%BA%BA%E6%B0%91%E5%85%B1%E5%92%8C%E5%9B%BD%E5%88%91%E6%B3%95)》
2023-03-22 13:04:52 +00:00
#### 《[中华人民共和国民法典](http://gongbao.court.gov.cn/Details/51eb6750b8361f79be8f90d09bc202.html)》
2023-03-24 12:22:52 +00:00
2023-06-14 08:14:01 +00:00
#### 《[中华人民共和国合同法](http://www.npc.gov.cn/zgrdw/npc/lfzt/rlyw/2016-07/01/content_1992739.htm)》
2023-03-24 12:22:52 +00:00
## 💪 Thanks to all contributors for their efforts
2023-03-24 12:25:15 +00:00
<a href="https://github.com/svc-develop-team/so-vits-svc/graphs/contributors" target="_blank">
<img src="https://contrib.rocks/image?repo=svc-develop-team/so-vits-svc" />
</a>