1berry/wokada-voice-changer
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

指摘事項の修正

Browse Source
This commit is contained in:
nadare 2023-04-28 11:07:58 +00:00
parent 91a35fc688
commit 90c9c764df
2 changed files with 62 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

65
tutorials/tutorial_rvc_en_latest.md
View File

@ -1,119 +1,146 @@
Realtime Voice Changer Client for RVC Tutorial (v.1.5.2.4) # Realtime Voice Changer Client for RVC Tutorial (v.1.5.2.4)
================================================================
# Introduction # Introduction
This application is client software for real-time voice conversion that supports various voice conversion models. This document provides a description for voice conversion limited to [RVC(Retrieval-based-Voice-Conversion)](https://github.com/liujing04/Retrieval-based-Voice-Conversion-WebUI). This application is client software for real-time voice conversion that supports various voice conversion models. This document provides a description for voice conversion limited to [RVC(Retrieval-based-Voice-Conversion)](https://github.com/liujing04/Retrieval-based-Voice-Conversion-WebUI).
## Notes ## Notes
- Model training must be done separately. - Model training must be done separately.
- If you want to learn by yourself, please go to [RVC(Retrieval-based-Voice-Conversion)](https://github.com/liujing04/Retrieval-based-Voice-Conversion-WebUI). - If you want to learn by yourself, please go to [RVC(Retrieval-based-Voice-Conversion)](https://github.com/liujing04/Retrieval-based-Voice-Conversion-WebUI).
- [Recording app on Github Pages](https://w-okada.github.io/voice-changer/) is convenient for preparing voice for learning on the browser. - [Recording app on Github Pages](https://w-okada.github.io/voice-changer/) is convenient for preparing voice for learning on the browser.
- [Commentary video] (https://youtu.be/s_GirFEGvaA) - [Commentary video] (https://youtu.be/s_GirFEGvaA)
- [TIPS for training](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/training_tips_en.md) has been published, so please refer to it. - [TIPS for training](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/training_tips_en.md) has been published, so please refer to it.
# 起動まで
## HuBERTのインストール
RVCの実行にはHuBERTが必要です。
[このリポジトリ](https://huggingface.co/lj1995/VoiceConversionWebUI/tree/main)から`hubert_base.pt`をダウンロードして、バッチファイルがあるフォルダに格納してください。
# Steps up to startup # Steps up to startup
## Installing HuBERT ## Installing HuBERT
HuBERT is required to run RVC. HuBERT is required to run RVC.
Download `hubert_base.pt` from [this repository](https://huggingface.co/lj1995/VoiceConversionWebUI/tree/main) and store it in the folder containing the batch file. Download `hubert_base.pt` from [this repository](https://huggingface.co/lj1995/VoiceConversionWebUI/tree/main) and store it in the folder containing the batch file.
## Start GUI ## Start GUI
### Windows version, ### Windows version,
Unzip the downloaded zip file and run `start_http.bat`. Unzip the downloaded zip file and run `start_http.bat`.
### Mac version ### Mac version
After extracting the download file, execute `startHttp.command`. If it shows that the developer cannot be verified, press the control key again and click to execute (or right-click to execute). After extracting the download file, execute `startHttp.command`. If it shows that the developer cannot be verified, press the control key again and click to execute (or right-click to execute).
### Precautions when connecting remotely ### Precautions when connecting remotely
When connecting remotely, please use `.bat` file (win) and `.command` file (mac) where http is replaced with https. When connecting remotely, please use `.bat` file (win) and `.command` file (mac) where http is replaced with https.
## client selection ## client selection
It is successful if the Launcher screen like the one below appears. Select RVC from this screen. It is successful if the Launcher screen like the one below appears. Select RVC from this screen.
<img src="/tutorials/images/launcher.png" alt="launcher" width="800" loading="lazy"> <img src="/tutorials/images/launcher.png" alt="launcher" width="800" loading="lazy">
## Screen for RVC ## Screen for RVC
It is successful if the following screen appears. It is successful if the following screen appears.
<img src="/tutorials/images/RVC_GUI.png" alt="launcher" width="800" loading="lazy"> <img src="/tutorials/images/RVC_GUI.png" alt="launcher" width="800" loading="lazy">
## GUI item details ## GUI item details
## server control ## server control
### start ### start
`start` starts the server, `stop` stops the server `start` starts the server, `stop` stops the server
### monitor ### monitor
Indicates the status of real-time conversion. Indicates the status of real-time conversion.
The lag from voicing to conversion is `buf + res seconds`. Adjust so that the buf time is longer than res. The lag from voicing to conversion is `buf + res seconds`. Adjust so that the buf time is longer than res.
#### vol #### vol
This is the volume after voice conversion. This is the volume after voice conversion.
#### buf #### buf
It is the length (ms) of one section to cut the audio. Shortening the Input Chunk reduces this number. It is the length (ms) of one section to cut the audio. Shortening the Input Chunk reduces this number.
#### res #### res
This is the time it takes to convert data that is the sum of Input Chunk and Extra Data Length. Shortening both Input Chunk and Extra Data Length will reduce the number. This is the time it takes to convert data that is the sum of Input Chunk and Extra Data Length. Shortening both Input Chunk and Extra Data Length will reduce the number.
### Model Info ### Model Info
Get information held by the server. If information synchronization between server and client seems not to be successful, please press the Reload button. Get information held by the server. If information synchronization between server and client seems not to be successful, please press the Reload button.
### Switch Model ### Switch Model
You can switch between uploaded models. You can switch between uploaded models.
## Model Setting ## Model Setting
### Model Uploader ### Model Uploader
If enable PyTorch is turned on, you can select the PyTorch model (extension is pth). If you turn this on when using a model converted from RVC, the PyTorch item will appear. （From the next version, you can only choose either PyTorch or ONNX for each slot. If enable PyTorch is turned on, you can select the PyTorch model (extension is pth). If you turn this on when using a model converted from RVC, the PyTorch item will appear. （From the next version, you can only choose either PyTorch or ONNX for each slot.
#### Model Slot #### Model Slot
You can choose which frame to set the model in. The set model can be switched with Switch Model in Server Control. You can choose which frame to set the model in. The set model can be switched with Switch Model in Server Control.
#### Onnx(.onnx) #### Onnx(.onnx)
Specify the model in .onnx format here. This or PyTorch (.pth) is required. Specify the model in .onnx format here. This or PyTorch (.pth) is required.
#### PyTorch(.pth) #### PyTorch(.pth)
Specify the model in .pth format here. This or Onnx (.onnx) is required. Specify the model in .pth format here. This or Onnx (.onnx) is required.
If you trained with RVC-WebUI, it's in `/logs/weights`. If you trained with RVC-WebUI, it's in `/logs/weights`.
#### feature(.npy) #### feature(.npy)
This is an additional function that brings the features extracted by HuBERT closer to the training data. Used in pairs with index(.index). This is an additional function that brings the features extracted by HuBERT closer to the training data. Used in pairs with index(.index).
If you trained with RVC-WebUI, it is saved as `/logs/weights/total_fea.npy`. If you trained with RVC-WebUI, it is saved as `/logs/weights/total_fea.npy`.
#### index(.index) #### index(.index)
This is an additional function that brings the features extracted by HuBERT closer to the training data. Used in pairs with feature(.npy). This is an additional function that brings the features extracted by HuBERT closer to the training data. Used in pairs with feature(.npy).
If you trained with RVC-WebUI, it is saved as `/logs/weights/add_XXX.index`. If you trained with RVC-WebUI, it is saved as `/logs/weights/add_XXX.index`.
#### half-precision #### half-precision
You can choose to infer precision as float32 or float16. You can choose to infer precision as float32 or float16.
This selection can be speeded up at the expense of accuracy. This selection can be speeded up at the expense of accuracy.
Turn it off if it doesn't work. Turn it off if it doesn't work.
#### Default Tune #### Default Tune
Enter the default value for how much the pitch of the voice should be converted. You can also convert during inference. Below is a guideline for the settings. Enter the default value for how much the pitch of the voice should be converted. You can also convert during inference. Below is a guideline for the settings.
- +12 for male voice to female voice conversion - +12 for male voice to female voice conversion
- -12 for female voice to male voice conversion - -12 for female voice to male voice conversion
#### upload #### upload
After setting the above items, press to make the model ready for use. After setting the above items, press to make the model ready for use.
#### Framework #### Framework
Choose which of the uploaded model files to use (PyTorch or ONNX). It will be gone in the next version. Choose which of the uploaded model files to use (PyTorch or ONNX). It will be gone in the next version.
## Device Setting ## Device Setting
### Audio Input ### Audio Input
Choose an input device Choose an input device
### Audio Output ### Audio Output
Choose an output terminal Choose an output terminal
#### output record #### output record
Audio is recorded from when you press start until you press stop. Audio is recorded from when you press start until you press stop.
Pressing this button does not start real-time conversion. Pressing this button does not start real-time conversion.
Press Server Control for real-time conversion Press Server Control for real-time conversion
@ -121,50 +148,62 @@ Press Server Control for real-time conversion
## Quality Control ## Quality Control
### Noise Supression ### Noise Supression
On/Off of the browser's built-in noise removal function. On/Off of the browser's built-in noise removal function.
### Gain Control ### Gain Control
- input: Increase or decrease the volume of the input audio to the model. 1 is the default value - input: Increase or decrease the volume of the input audio to the model. 1 is the default value
- output: Increase or decrease the volume of the output audio from the model. 1 is the default value - output: Increase or decrease the volume of the output audio from the model. 1 is the default value
### F0Detector ### F0Detector
Choose an algorithm for extracting the pitch. You can choose from the following two types. Choose an algorithm for extracting the pitch. You can choose from the following two types.
- Lightweight `pm` - Lightweight `pm`
- Highly accurate `harvest` - Highly accurate `harvest`
### Analyzer(Experimental) ### Analyzer(Experimental)
Record input and output on the server side. Record input and output on the server side.
As for the input, the sound of the microphone is sent to the server and recorded as it is. It can be used to check the communication path from the microphone to the server. As for the input, the sound of the microphone is sent to the server and recorded as it is. It can be used to check the communication path from the microphone to the server.
For output, the data output from the model is recorded in the server. You can see how the model behaves (once you've verified that your input is correct). For output, the data output from the model is recorded in the server. You can see how the model behaves (once you've verified that your input is correct).
## Speaker Setting ## Speaker Setting
### Destination Speaker Id ### Destination Speaker Id
It seems to be a setting when supporting multiple speakers, but it is not used at present because the RVC head office does not support it (it is unlikely). It seems to be a setting when supporting multiple speakers, but it is not used at present because the RVC head office does not support it (it is unlikely).
### Tuning ### Tuning
Adjust the pitch of your voice. Below is a guideline for the settings. Adjust the pitch of your voice. Below is a guideline for the settings.
- +12 for male voice to female voice conversion - +12 for male voice to female voice conversion
- -12 for female voice to male voice conversion - -12 for female voice to male voice conversion
### index ratio ### index ratio
Specify the ratio to shift to the features used in training. Effective when both feature and index are set in Model Setting. Specify the ratio to shift to the features used in training. Effective when both feature and index are set in Model Setting.
0 uses the output of HuBERT as it is, 1 brings it all back to the original features. 0 uses the output of HuBERT as it is, 1 brings it all back to the original features.
If the index ratio is greater than 0, the search may take a long time. If the index ratio is greater than 0, the search may take a long time.
### Silent Threshold ### Silent Threshold
The volume threshold for audio conversion. If the rms is smaller than this value, no voice conversion is performed and silence is returned. The volume threshold for audio conversion. If the rms is smaller than this value, no voice conversion is performed and silence is returned.
(In this case, the conversion process is skipped, so the load is less.) (In this case, the conversion process is skipped, so the load is less.)
## Converter Setting ## Converter Setting
### InputChunk Num(128sample / chunk) ### InputChunk Num(128sample / chunk)
Decide how much length to cut and convert in one conversion. The higher the value, the more efficient the conversion, but the larger the buf value, the longer the maximum time before the conversion starts. The approximate time is displayed in buff:. Decide how much length to cut and convert in one conversion. The higher the value, the more efficient the conversion, but the larger the buf value, the longer the maximum time before the conversion starts. The approximate time is displayed in buff:.
### Extra Data Length ### Extra Data Length
Determines how much past audio to include in the input when converting audio. The longer the past voice is, the better the accuracy of the conversion, but the longer the res is, the longer the calculation takes. Determines how much past audio to include in the input when converting audio. The longer the past voice is, the better the accuracy of the conversion, but the longer the res is, the longer the calculation takes.
(Probably because Transformer is a bottleneck, the calculation time will increase by the square of this length) (Probably because Transformer is a bottleneck, the calculation time will increase by the square of this length)
### GPU ### GPU
If you have 2 or more GPUs, you can choose your GPU here.
If you have 2 or more GPUs, you can choose your GPU here.

19
tutorials/tutorial_rvc_ja_latest.md
View File

@ -3,13 +3,13 @@ Realtime Voice Changer Client for RVC チュートリアル(v.1.5.2.5)
# はじめに # はじめに
本アプリケーションは、各種音声変換 AI(VC, Voice Conversion)を用いてリアルタイム音声変換を行うためのクライアントソフトウェアです。本ドキュメントでは[RVC(Retrieval-based-Voice-Conversion)](https://github.com/liujing04/Retrieval-based-Voice-Conversion-WebUI)に限定した音声変換のためのチュートリアルを行います。 本アプリケーションは、各種音声変換 AI(VC, Voice Conversion)を用いてリアルタイム音声変換を行うためのクライアントソフトウェアです。本ドキュメントでは[RVC(Retrieval-based-Voice-Conversion)](https://github.com/liujing04/Retrieval-based-Voice-Conversion-WebUI)に限定した音声変換のためのチュートリアルを行います。
以下、本家の[Retrieval-based-Voice-Conversion-WebUI](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)を本家RVCと表記し、ddPn08氏の作成した[RVC-WebUI](https://github.com/ddPn08/rvc-webui)をだだぱんRVCと記載します。 以下、本家の[Retrieval-based-Voice-Conversion-WebUI](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)を本家RVCと表記し、ddPn08氏の作成した[RVC-WebUI](https://github.com/ddPn08/rvc-webui)をddPn08RVCと記載します。
## 注意事項 ## 注意事項
- 学習については別途行う必要があります。 - 学習については別途行う必要があります。
- 自身で学習を行う場合は[RVC(Retrieval-based-Voice-Conversion)](https://github.com/liujing04/Retrieval-based-Voice-Conversion-WebUI)で行ってください。 - 自身で学習を行う場合は[本家RVC](https://github.com/liujing04/Retrieval-based-Voice-Conversion-WebUI)または[ddPn08RVC](https://github.com/ddPn08/rvc-webui)で行ってください。
- ブラウザ上で学習用の音声を用意するには[録音アプリ on Github Pages](https://w-okada.github.io/voice-changer/)が便利です。 - ブラウザ上で学習用の音声を用意するには[録音アプリ on Github Pages](https://w-okada.github.io/voice-changer/)が便利です。
- [解説動画](https://youtu.be/s_GirFEGvaA) - [解説動画](https://youtu.be/s_GirFEGvaA)
- [trainingのTIPS](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/training_tips_ja.md)が公開されているので参照してください。 - [trainingのTIPS](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/training_tips_ja.md)が公開されているので参照してください。
@ -32,12 +32,13 @@ RVCの実行にはHuBERTが必要です。
## クライアントの選択 ## クライアントの選択
下記のようなLauncher画面が出れば成功です。この画面からRVCを選択してください。 下記のようなLauncher画面が出れば成功です。この画面からRVCを選択してください。
<img src="/tutorials/images/launcher.png" alt="launcher" width="800" loading="lazy"> ![クライアントの選択画面](https://user-images.githubusercontent.com/23290400/235131650-9eeee978-96fa-478a-b728-3581ae0b8b67.png)
## RVC用の画面 ## RVC用の画面
下記のような画面が出れば成功です。右上の?ボタンから[マニュアル](https://zenn.dev/wok/books/0004_vc-client-v_1_5_1_x)に移動できます。 下記のような画面が出れば成功です。右上の?ボタンから[マニュアル](https://zenn.dev/wok/books/0004_vc-client-v_1_5_1_x)に移動できます。
<img src="/tutorials/images/RVC_GUI.png" alt="launcher" width="800" loading="lazy"> ![v1.5.2.5 RVC初期画面](https://user-images.githubusercontent.com/23290400/235131679-5b4a1b7e-5d1d-4163-bfe7-84f3a4585ab7.png)
# クイックスタート # クイックスタート
日本語版では[マニュアル](https://zenn.dev/wok/books/0004_vc-client-v_1_5_1_x/viewer/003-1_quick-start)が用意されているのでこちらを参照してください。 日本語版では[マニュアル](https://zenn.dev/wok/books/0004_vc-client-v_1_5_1_x/viewer/003-1_quick-start)が用意されているのでこちらを参照してください。
@ -74,7 +75,7 @@ Input ChunkとExtra Data Lengthを足したデータを変換にかかる時間
3. モデルが用いる特徴量のチャンネル数(大きいほど情報を持っていて重い) 3. モデルが用いる特徴量のチャンネル数(大きいほど情報を持っていて重い)
4. 学習に用いられたクライアント 4. 学習に用いられたクライアント
- org: [本家RVC](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)で学習したモデルです。 - org: [本家RVC](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)で学習したモデルです。
- webui:[だだぱんRVC](https://github.com/ddPn08/rvc-webui)で学習したモデルです。 - webui:[ddPn08RVC](https://github.com/ddPn08/rvc-webui)で学習したモデルです。
### EXPORT ONNX ### EXPORT ONNX
ONNXモデルを出力します。PyTorchのモデルをONNXモデルに変換すると、推論が高速化される場合があります。 ONNXモデルを出力します。PyTorchのモデルをONNXモデルに変換すると、推論が高速化される場合があります。
@ -91,17 +92,17 @@ enable PyTorchをオンにするとPyTorchのモデル(拡張子がpth)を選ぶ
ONNX形式(.onnx)かPyTorch形式(.pth)のいずれかを選択可能です。 ONNX形式(.onnx)かPyTorch形式(.pth)のいずれかを選択可能です。
本家RVCで学習させた場合、`/logs/weights`に入っています。 本家RVCで学習させた場合、`/logs/weights`に入っています。
だだぱんRVCで学習させた場合、`/models/checkpoints`に入っています。 ddPn08RVCで学習させた場合、`/models/checkpoints`に入っています。
#### feature(.npy) #### feature(.npy)
HuBERTで抽出した特徴を訓練データに近づける追加機能です。index(.index)とペアで使用します。 HuBERTで抽出した特徴を訓練データに近づける追加機能です。index(.index)とペアで使用します。
本家RVCで学習させた場合、`/logs/実験名/total_fea.npy`という名前で保存されています。(2023/04/26にtotal_fea.npyを省略するアップデートが入ったので今後不要になる可能性があります) 本家RVCで学習させた場合、`/logs/実験名/total_fea.npy`という名前で保存されています。(2023/04/26にtotal_fea.npyを省略するアップデートが入ったので今後不要になる可能性があります)
だだぱんRVCで学習させた場合、`/models/checkpoints/モデル名_index/モデル名.0.big.npy`という名前で保存されています。 ddPn08RVCで学習させた場合、`/models/checkpoints/モデル名_index/モデル名.0.big.npy`という名前で保存されています。
#### index(.index) #### index(.index)
HuBERTで抽出した特徴を訓練データに近づける追加機能です。feature(.npy)とペアで使用します。 HuBERTで抽出した特徴を訓練データに近づける追加機能です。feature(.npy)とペアで使用します。
本家RVCで学習させた場合、`/logs/実験名/add_XXX.index`という名前で保存されています。 本家RVCで学習させた場合、`/logs/実験名/add_XXX.index`という名前で保存されています。
だだぱんRVCで学習させた場合、`/models/checkpoints/モデル名_index/モデル名.0.index`という名前で保存されています。 ddPn08RVCで学習させた場合、`/models/checkpoints/モデル名_index/モデル名.0.index`という名前で保存されています。
#### half-precision #### half-precision
精度をfloat32かfloat16のどちらで推論するか選べます。 精度をfloat32かfloat16のどちらで推論するか選べます。
@ -152,7 +153,7 @@ startをおしてからstopを押すまでの音声が記録されます。
## Speaker Setting ## Speaker Setting
### Destination Speaker Id ### Destination Speaker Id
複数話者に対応した時の設定かと思われますが、RVC本家が対応していない(見込みもない)ので現状は使わない項目です。 複数話者に対応した時の設定かと思われますが、[本家RVC](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)が対応していないので現状は使わない項目です。
### Tuning ### Tuning
声のピッチを調整します。以下は設定の目安です。 声のピッチを調整します。以下は設定の目安です。