AllTalk TTS

AllTalk is an updated version of the Coqui_tts extension for Text Generation web UI. Features include:

Custom Start-up Settings: Adjust your default start-up settings. Screenshot
Narrarator: Use different voices for main character and narration. Example Narration
Low VRAM mode: Great for people with small GPU memory or if your VRAM is filled by your LLM. Screenshot
DeepSpeed: A 3-4x performance boost generating TTS. DeepSpeed Windows/Linux Instructions Screenshot
Local/Custom models: Use any of the XTTSv2 models (API Local and XTTSv2 Local).
Optional wav file maintenance: Configurable deletion of old output wav files. Screenshot
Finetuning Train the model specifically on a voice of your choosing for better reproduction.
Documentation: Fully documented with a built in webpage. Screenshot
Console output Clear command line output for any warnings or issues.
API Suite and 3rd Party support via JSON calls Can be used with 3rd party applications via JSON calls.
Can be run as a standalone app Not just inside of text-generation-webui.

Index

🟩 Installation
🟪 Updating & problems with updating
🟫 Screenshots
🟨 Help with problems
⚫ Finetuning a model
🔵🟢🟡 DeepSpeed Installation (Windows & Linux)
🟦 Running AllTalk as a standalone app
🟠 API Suite and JSON-CURL
🔴 Future to-do list & Upcoming updates

🔄 Updates

Please check the below link to find a list of all recent updates and changes.

🔄 Updates list & bug fixes list can be found here

🟩 Installation on Text generation web UI

This has been tested on the current Dec 2023 release of Text generation webUI. If you have not updated it for a while, you may wish to update Text generation webUI, instructions here

If you want to watch a video of how to do the below link here

In a command prompt/terminal window you need to move into your Text generation webUI folder:

cd text-generation-webui
Start the Text generation webUI Python environment for your OS with whichever one of the below is correct for your OS:

cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat

Loading Text-generation-webui's Python Environment is $\textcolor{red}{\textsf{VERY IMPORTANT}}$. If you are uncertain what a loaded Python environment looks like, image here and video here
Move into your extensions folder:

cd extensions
Once there git clone this repository:

git clone https://github.com/erew123/alltalk_tts
Move into the alltalk_tts folder:

cd alltalk_tts
Install one of the two requirements files. Whichever one of the two is correct for your machine type:

Nvidia graphics card machines - pip install -r requirements_nvidia.txt

Other machines (mac, amd etc) - pip install -r requirements_other.txt
(Optional DeepSpeed) If you have an Nvidia Graphics card on a system running Linux or Windows and wish to use DeepSpeed please follow these instructions here. However, I would highly reccommend before you install DeepSpeed, you start text-generation-webui up, confirm AllTalk starts correctly and everything is working, as DeepSpeed can add another layer of complications troubleshooting any potential start-up 6D47 issues. If necessary you can pip uninstall deepspeed.
You can now start move back to the main Text generation webUI folder cd .. (a few times), start Text generation webUI with whichever one of the startup scripts is correct for your OS (start_windows.bat,./start_linux.sh, start_macos.sh or start_wsl.bat) and load the AllTalk extension in the Text generation webUI session tab.

Starting Text-generation-webui with its correct start-up script is $\textcolor{red}{\textsf{VERY IMPORTANT}}$.
Please read the note below about start-up times and also the note about ensuring your character cards are set up correctly
Some extra voices downloadable here

🟩 Other installation notes

On first startup, AllTalk will download the Coqui XTTSv2 2.0.2 model to its models folder (1.8GB space required). Check the command prompt/terminal window if you want to know what its doing. After it says "Model Loaded" the Text generation webUI is usually available on its IP address a few seconds later, for you to connect to in your browser.

Once the extension is loaded, please find all documentation and settings on the link provided in the interface (as shown in the screenshot below).

Where to find voices https://aiartes.com/voiceai or https://commons.wikimedia.org/ or interviews on youtube etc. Instructions on how to cut down and prepare a voice sample are within the built in documentation.

🟩 The one thing I cant easily work around

Messages intended for the Narrator should be enclosed in asterisks * and those for the character inside quotation marks ". However, AI systems often deviate from these rules, resulting in text that is neither in quotes nor asterisks. Sometimes, text may appear with only a single asterisk, and AI models may vary their formatting mid-conversation. For example, they might use asterisks initially and then switch to unmarked text. A properly formatted line should look like this:

"Hey! I'm so excited to finally meet you. I've heard so many great things about you and I'm eager to pick your brain about computers." *She walked across the room and picked up her cup of coffee*

Most narrator/character systems switch voices upon encountering an asterisk or quotation marks, which is somewhat effective. AllTalk has undergone several revisions in its sentence splitting and identification methods. While some irregularities and AI deviations in message formatting are inevitable, any line beginning or ending with an asterisk should now be recognized as Narrator dialogue. Lines enclosed in double quotes are identified as Character dialogue. For any other text, you can choose how AllTalk handles it: whether it should be interpreted as Character or Narrator dialogue (most AI systems tend to lean more towards one format when generating text not enclosed in quotes or asterisks).

With improvements to the splitter/processor, I'm confident it's functioning well. You can monitor what AllTalk identifies as Narrator lines on the command line and adjust its behavior if needed (Text Not Inside - Function).

🟪 Updating

This is pretty much a repeat of the installation process.

In a command prompt/terminal window you need to move into your Text generation webUI folder:

cd text-generation-webui and start the Python environment for your OS with whichever one of the below is correct for your OS:

cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat
Move into your extensions and alltalk_tts folder:

cd extensions then cd alltalk_tts
At the command prompt/terminal, type:

git pull
Install the correct requirements for your machine:

Nvidia graphics card machines - pip install -r requirements_nvidia.txt

Other machines (mac, amd etc) - pip install -r requirements_other.txt

🟪 Updating "git pull" error

Click to expand

I did leave a mistake in the /extensions/alltalk_tts/.gitignore file at one point. If your git pull doesnt work, you can either follow the Problems Updating section below, or edit the .gitignore file and replace its entire contents with the below, save the file, then re-try the git pull

voices/*.*
models/*.*
outputs/*.*
finetune/*.*
config.json
confignew.json
models.json
diagnostics.log

🟪 Updating other problems

Click to expand

If you do experience any problems, the simplest method to resolve this will be:

re-name the existing alltalk_tts folder to something like alltalk_tts.old
Start a console/terminal then:

cd text-generation-webui and start your python environment cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat
Move into the extensions folder, same as if you were doing a fresh installation:

cd extensions then

git clone https://github.com/erew123/alltalk_tts

This will download a fresh installation.

Move into the alltalk_tts folder:

cd alltalk_tts
Install the correct requirements for your machine:

Nvidia graphics card machines - pip install -r requirements_nvidia.txt

Other machines (mac, amd etc) - pip install -r requirements_other.txt
Before starting it up, copy/merge the models, voices and outputs folders over from the alltalk_tts.old folder to the newly created alltalk_tts folder. This will keep your voices history and also stop it re-downloading the model again.

You can now start text-generation-webui or AllTalk (standalone) and it should start up fine. You will need to re-set any saved configuration changes on the configuration page.

Assuming its all working fine and you are happy, you can delete the old alltalk_tts.old folder.

🟫 Screenshots

🟨 Help with problems

🔄 Minor updates/bug fixes list can be found here

🟨 How to make a diagnostics report file

Click to expand

Open a command prompt window, move into your text-generation-webui folder, you can now start the Python environment for text-generation-webui:

cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat
Move into the alltalk_tts folder:

cd extensions and then cd alltalk_tts
Run the diagnostics and select the requirements file name you installed AllTalk with:

python diagnostics.py
You will have an on screen output showing your environment setttings, file versions request vs whats installed and details of your graphics card (if Nvidia). This will also create a file called diagnostics.log in the alltalk_tts folder, that you can upload if you need to create a support ticket on here.

🟨 [AllTalk Startup] Warning TTS Subprocess has NOT started up yet, Will keep trying for 120 seconds maximum. Please wait. It times out after 120 seconds.

Click to expand

When the subprocess is starting 2x things are occurring:

A) Its trying to load the voice model into your graphics card VRAM (assuming you have a Nvidia Graphics card, otherwise its your system RAM)
B) Its trying to start up the mini-webserver and send the "ready" signal back to the main process.

Before giving other possibilities a go, some people with old machines are finding their startup times are very slow 2-3 minutes. Ive extended the allowed time within the script from 1 minute to 2 minutes. If you have an older machine and wish to try extending this further, you can do so by editing script.py and changing line 251 timeout = 120 changing the timeout to a larger value e.g timeout = 240 (4 minutes).

Note: If you need to create a support ticket, please create a diagnostics.log report file to submit with a support request. Details on doing this are above.

Other possibilities for this issue are:

You are starting AllTalk in both your CMD FLAG.txt and settings.yaml file. The CMD FLAG.txt you would have manually edited and the settings.yaml is the one you change and save in the session tab of text-generation-webui and you can Save UI defaults to settings.yaml. Please only have one of those two starting up AllTalk.
You are not starting text-generation-webui with its normal Python environment. Please start it with start_{your OS version} as detailed here (start_windows.bat,./start_linux.sh, start_macos.sh or start_wsl.bat) OR (cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat and then python server.py).
You have installed the wrong version of DeepSpeed on your system, for the wrong version of Python/Text-generation-webui. You can go to your text-generation-webui folder in a terminal/command prompt and run the correct cmd version for your OS e.g. (cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat) and then you can type pip uninstall deepspeed then try loading it again. If that works, please see here for the correct instructions for installing DeepSpeed here.
You have an old version of text-generation-webui (pre Dec 2023) I have not tested on older versions of text-generation-webui, so cannot confirm viability on older versions. For instructions on updating the text-generation-webui, please look here (update_linux.sh, update_windows.bat, update_macos.sh, or update_wsl.bat).
You already have something running on port 7851 on your computer, so the mini-webserver cant start on that port. You can change this port number by editing the confignew.json file and changing "port_number": "7851" to "port_number": "7602" or any port number you wish that isn’t reserved. Only change the number and save the file, do not change the formatting of the document. This will at least discount that you have something else clashing on the same port number.
You have antivirus/firewalling that is blocking that port from being accessed. If you had to do something to allow text-generation-webui through your antivirus/firewall, you will have to do that for this too.
You have quite old graphics drivers and may need to update them.
Something within text-generation-webui is not playing nicely for some reason. You can go to your text-generation-webui folder in a terminal/command prompt and run the correct cmd version for your OS e.g. (cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat) and then you can type python extensions\alltalk_tts\script.py and see if AllTalk starts up correctly. If it does then something else is interfering.
Something else is already loaded into your VRAM or there is a crashed python process. Either check your task manager for erroneous Python processes or restart your machine and try again.
You are running DeepSpeed on a Linux machine and although you are starting with ./start_linux.sh AllTalk is failing there on starting. This is because text-generation-webui will overwrite some environment variables when it loads its python environment. To see if this is the problem, from a terminal go into your text-generation-webui folder and ./cmd_linux.sh then set your environment variable again e.g. export CUDA_HOME=/usr/local/cuda (this may vary depending on your OS, but this is the standard one for Linux, and assuming you have installed the CUDA toolkit), then python server.py and see if it starts up. If you want to edit the environment permanently you can do so, I have not managed to write full instructions yet, but here is the conda guide here.
You have built yourself a custom Python environment and something is funky with it. This is very hard to diagnose as its not a standard environment. You may want to updating text-generation-webui and re installing its requirements file (whichever one you use that comes down with text-generation-webui).

🟨 I think AllTalks requirements file has installed something another extension doesn't like

Click to expand

Ive paid very close attention to not impact what Text-generation-webui is requesting on a factory install. This is one of the requirements of submitting an extension to Text-generation-webui. If you want to look at a comparison of a factory fresh text-generation-webui installed packages (with cuda 12.1, though AllTalk's requirements were set on cuda 11.8) you can find that comparison here. This comparison shows that AllTalk is requesting the same package version numbers as Text-generation-webui or even lower version numbers (meaning AllTalk will not update them to a later version). What other extensions do, I cant really account for that.

I will note that the TTS engine downgrades Pandas data validator to 1.5.3 though its unlikely to cause any issues. You can upgrade it back to text-generation-webui default (december 2023) with pip install pandas==2.1.4 when inside of the python environment. I have noticed no ill effects from it being a lower or higher version, as far as AllTalk goes. This is also the same behaviour as the Coqui_tts extension that comes with Text-generation-webui.

Other people are reporting issues with extensions not starting with errors about Pydantic e.g. pydantic.errors.PydanticImportError: BaseSettings` has been moved to the pydantic-settings package. See https://docs.pydantic.dev/2.5/migration/#basesettings-has-moved-to-pydantic-settings for more details.

Im not sure if the Pydantic version has been recently updated by the Text-generation-webui installer, but this is nothing to do with AllTalk. The other extension you are having an issue with, need to be updated to work with Pydantic 2.5.x. AllTalk was updated in mid december to work with 2.5.x. I am not specifically condoning doing this, as it may have other knock on effects, but within the text-gen Python environment, you can use pip install pydantic==2.5.0 or pip install pydantic==1.10.13 to change the version of Pydantic installed.

🟨 I activated DeepSpeed in the settings page, but I didnt install DeepSpeed yet and now I have issues starting up

Click to expand

You can either follow the Problems Updating and fresh install your config. Or you can edit the confignew.json file within the alltalk_tts folder. You would look for '"deepspeed_activate": true,' and change the word true to false `"deepspeed_activate": false,' ,then save the file and try starting again.

If you want to use DeepSpeed, you need an Nvidia Graphics card and to install DeepSpeed on your system. Instructions are here

🟨 I am having problems updating/some other issue where it wont start up/Im sure this is a bug

Click to expand

Please see Problems Updating. If that doesnt help you can raise an ticket here. It would be handy to have any log files from the console where your error is being shown. I can only losely support custom built Python environments and give general pointers. Please create a diagnostics.log report file to submit with a support request.

Also, is your text-generation-webui up to date? instructions here

🟨 I am having problems getting AllTalk to start after changing settings or making a custom setup/model setup.

Click to expand

I would suggest following Problems Updating and if you still have issues after that, you can raise an issue here

🟨 I see some red "asyncio" messages

Click to expand

As far as I am aware, these are to do with the chrome browser the gradio text-generation-webui in some way. I raised an issue about this on the text-generation-webui here where you can see that AllTalk is not loaded and the messages persist. Either way, this is more a warning than an actual issue, so shouldnt affect any functionality of either AllTalk or text-generation-webui, they are more just an annoyance.

🟨 I have multiple GPU's and I have problems running Finetuning

Click to expand

Finetuning pulls in various other scripts and some of those scripts can have issues with multiple Nvidia GPU's being present. Until the people that created those other scripts fix up their code, there is a workaround to temporarily tell your system to only use the 1x of your Nvidia GPU's. To do this:

Windows - You will start the script with set CUDA_VISIBLE_DEVICES=0 && python finetune.py
After you have completed training, you can reset back with set CUDA_VISIBLE_DEVICES=
Linux - You will start the script with CUDA_VISIBLE_DEVICES=0 python finetune.py
After you have completed training, you can reset back with unset CUDA_VISIBLE_DEVICES

Rebooting your system will also unset this. The setting is only applied temporarily.

Depending on which of your Nvidia GPU's is the more powerful one, you can change the 0 to 1 or whichever of your GPU's is the most powerful.

⚫ Finetuning a model

If you have a voice that the model doesnt quite reproduce correctly, or indeed you just want to improve the reproduced voice, then finetuning is a way to train your "XTTSv2 local" model (stored in /alltalk_tts/models/xxxxx/) on a specific voice. For this you will need:

An Nvidia graphics card. (Please see this note if you have multiple Nvidia GPU's).
To install a few portions of the Nvidia CUDA 11.8 Toolkit (this will not impact text-generation-webui's cuda setup.
18GB of disk space free (most of this is used temporarily)
At least 2 minutes of good quality speech from your chosen speaker in mp3, wav or flacc format, in one or more files (have tested as far as 20 minutes worth of audio).

⚫ How will this work/How complicated is it?

Everything has been done to make this as simple as possible. At its simplest, you can literally just download a large chunk of audio from an interview, and tell the finetuning to strip through it, find spoken parts and build your dataset. You can literally click 4 buttons, then copy a few files and you are done. At it's more complicated end you will clean up the audio a little beforehand, but its still only 4x buttons and copying a few files.

⚫ The audio you will use

I would suggest that if its in an interview format, you cut out the interviewer speaking in audacity or your chosen audio editing package. You dont have to worry about being perfect with your cuts, the finetuning Step 1 will go and find spoken audio and cut it out for you. Is there is music over the spoken parts, for best quality you would cut out those parts, though its not 100% necessary. As always, try to avoid bad quality audio with noises in it (humming sounds, hiss etc). You can try something like Audioenhancer to try clean up noisier audio. There is no need to down-sample any of the audio, all of that is handled for you. Just give the finetuning some good quality audio to work with.

⚫ Important requirements CUDA 11.8

As mentioned you must have a small portion of the Nvidia CUDA Toolkit 11.8 installed. Not higher or lower versions. Specifically 11.8. You do not have to uninstall any other versions, change any graphics drivers, reinstall torch or anything like that. To keep the download+install as small as possible, you will need to:

Download the xxx (network) install of the Nvidia Cuda Toolkit 11.8 from here
Run the installer. At minimum, you need to [minimally] install the nvcc compiler and the CUBLAS development and runtime libraries:
- Select Custom Advanced as your installation type.
- Uncheck all the checkboxes in the list.
- Now check the following elements:
  - CUDA > Development > Compiler > nvcc
  - CUDA > Development > Compiler > Libraries > CUBLAS
  - CUDA > Runtime > Libraries > CUBLAS
- You can now proceed through the install.
When that has installed, open a terminal/command prompt and type nvcc --version. If it reports back Cuda compilation tools, release 11.8. you are good to go. Specifically, 11.8. If not continue to the next step.
For both Windows and Linux, you need to ensure that nvcc and the 11.8 cuda library files are in your environments search path. You can undo the changes below after finetuning if you prefer:
- Windows: Edit the Windows Path environment variable and add C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin
- Linux: The path may be different depending on what flavour of Linux you are running, so you may need to seek out specific instructions on the internet. Generic paths may be:
  - export LD_LIBRARY_PATH=/usr/local/cuda-11.8/lib64:${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}} and
  - export LD_LIBRARY_PATH=/usr/local/cuda-11.8/bin
  - Add these to your '~/.bashrc' if you want this to be permanent and not something you have to set each time you open a new terminal.
When you have made the changes, open a new terminal/command prompt (in order to load the new search paths) and nvcc --version. It should report back Cuda compilation tools, release 11.8. at which point, you are good to go.
If it doesn't report that, check you have correctly set the search environment paths, dont have overlapping other versions of cuda paths etc.

⚫ Starting Finetuning

NOTE: Please make sure you have started AllTalk at least once after updating, so that it downloads the additional files needed for finetuning.

Close all other applications that are using your GPU/VRAM and copy your audio samples into:

/alltalk_tts/finetune/put-voice-samples-in-here/
In a command prompt/terminal window you need to move into your Text generation webUI folder:

cd text-generation-webui
Start the Text generation webUI Python environment for your OS:

cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat
You can double check your search path environment still works correctly with nvcc --version. It should report back 11.8:

Cuda compilation tools, release 11.8.
Move into your extensions folder:

cd extensions
Move into the alltalk_tts folder:

cd alltalk_tts
Install the finetune requirements file: pip install -r requirements_finetune.txt
Type python finetune.py and it should start up.
Follow the on-screen instructions when the web interface starts up.
When you have finished finetuning, the final tab will tell you what to do with your files and how to move your newly trained model to the correct location on disk.

⚫ Using a Finetuned model in Text-generation-webui

At the end of the finetune process, you will have an option to Compact and move model to /trainedmodel/ this will compact the raw training file and move it to /model/trainedmodel/. When AllTalk starts up within Text-generation-webui, if it finds a model in this location a new loader will appear in the interface for XTTSv2 FT and you can use this to load your finetuned model.

Be careful not to train a new model from the base model, then overwrite your current /model/trainedmodel/ if you want a seperately trained model. This is why there is an OPTION B to move your just trained model to /models/lastfinetuned/.

⚫ Training one model with multiple voices

At the end of the finetune process, you will have an option to Compact and move model to /trainedmodel/ this will compact the raw training file and move it to /model/trainedmodel/. This model will become available when you start up finetuning. You will have a choice to train the Base Model or the Existing finetuned model (which is the one in /model/trainedmodel/). So you can use this to keep further training this model with additional voices, then copying it back to /model/trainedmodel/ at the end of training.

⚫ Do I need to keep the raw training data/model?

If you've compacted and moved your model, its highly unlikely you would want to keep that data, however the choice is there to keep it if you wish. It will be between 5-10GB in size, so most people will want to delete it.

🔵🟢🟡 DeepSpeed Installation Options

NOTE: You DO NOT need to set Text-generation-webUI's --deepspeed setting for AllTalk to be able to use DeepSpeed. These are two completely separate things and incorrectly setting that on Text-generation-webUI may cause other complications.

🔵 Linux Installation

Click to expand: Linux DeepSpeed installation

➡️DeepSpeed requires an Nvidia Graphics card!⬅️

Preferably use your built in package manager to install CUDA tools. Alternatively download and install the Nvidia Cuda Toolkit for Linux Nvidia Cuda Toolkit 11.8 or 12.1
Open a terminal console.
Install libaio-dev (however your Linux version installs things) e.g. sudo apt install libaio-dev
Move into your Text generation webUI folder e.g. cd text-generation-webui
Start the Text generation webUI Python environment ./cmd_linux.sh
Text generation webUI overwrites the CUDA_HOME environment variable each time you ./cmd_linux.sh or ./start_linux.sh, so you will need to either permanently change within the python environment OR set CUDA_HOME it each time you ./cmd_linux.sh. Details to change it each time are on the next step. Below is a link to Conda's manual and changing environment variables permanently though its possible changing it permanently could affect other extensions, you would have to test.

Conda manual - Environment variables
You can temporarily set the CUDA_HOME environment with (Standard paths on Ubuntu, but could vary on other Linux flavours):

export CUDA_HOME=/etc/alternatives/cuda

every time you run ./cmd_linux.sh.

If you try to start DeepSpeed with the CUDA_HOME path set incorrectly, expect an error similar to [Errno 2] No such file or directory: /home/yourname/text-generation-webui/installer_files/env/bin/nvcc
Now install deepspeed with pip install deepspeed
You can now start Text generation webUI python server.py ensuring to activate your extensions.

Just to reiterate, starting Text-generation-webUI with ./start_linux.sh will overwrite the CUDA_HOME variable unless you have permanently changed it, hence always starting it with ./cmd_linux.sh then setting the environment variable manually (step 7) and then python server.py, which is how you would need to run it each time, unless you permanently set the environment variable for CUDA_HOME within Text-generation-webUI's standard Python environment.

Removal - If it became necessary to uninstall DeepSpeed, you can do so with ./cmd_linux.sh and then pip uninstall deepspeed

🟢🟡 Windows Installation

DeepSpeed v11.2 will work on the current default text-generation-webui Python 3.11 environment! You have 2x options for how to setup DeepSpeed on Windows. A quick way (🟢Option 1) and a long way (🟡Option 2).

Thanks to @S95Sedan - They managed to get DeepSpeed 11.2 working on Windows via making some edits to the original Microsoft DeepSpeed v11.2 installation. The original post is here.

🟢 OPTION 1 - Quick and easy!

Click to expand: Pre-Compiled Wheel Deepspeed v11.2 (Python 3.11 and 3.10)

➡️DeepSpeed requires an Nvidia Graphics card!⬅️

Download the correct wheel version for your Python/Cuda from here and save the file it inside your text-generation-webui folder.
Open a command prompt window, move into your text-generation-webui folder, you can now start the Python environment for text-generation-webui:

cmd_windows.bat
With the file that you saved in the text-generation-webui folder you now type the following, replacing YOUR-VERSION with the name of the file you have:

pip install "deepspeed-0.11.2+YOUR-VERSION-win_amd64.whl"
This should install through cleanly and you should now have DeepSpeed v11.2 installed within the Python 3.11/3.10 environment of text-generation-webui.
When you start up text-generation-webui, and AllTalk starts, you should see [AllTalk Startup] DeepSpeed Detected
Within AllTalk, you will now have a checkbox for Activate DeepSpeed though remember you can only change 1x setting every 15 or so seconds, so dont try to activate DeepSpeed and LowVRAM/Change your model simultantiously. Do one of those, wait 15-20 seconds until the change is confirmed in the terminal/command prompt, then you can change the other. When you are happy it works, you can set the default start-up settings in the settings page.

Removal - If it became necessary to uninstall DeepSpeed, you can do so with cmd_windows.bat and then pip uninstall deepspeed

🟡 OPTION 2 - A bit more complicated!

Click to expand: Manual Build DeepSpeed v11.2 (Python 3.11 and 3.10)

➡️DeepSpeed requires an Nvidia Graphics card!⬅️

This will take about 1 hour to complete and about 6GB of disk space.

Download the 11.2 release of DeepSpeed extract it to a folder.
Install Visual C++ build tools, such as VS2019 C++ x64/x86 build tools.
Download and install the Nvidia Cuda Toolkit 11.8 or 12.1
OPTIONAL If you do not have an python environment already created and you are not going to use Text-generation-webui's environment, you can install Miniconda, then at a command prompt, create and activate your environment with:

conda create -n pythonenv python=3.11
activate pythonenv
Launch the Command Prompt cmd with Administrator privilege as it requires admin to allow creating symlink folders.
If you are using the Text-generation-webui python environment, then in the text-generation-webui folder you will run cmd_windows.bat to start the python evnironment.

Otherwise Install PyTorch, 2.1.0 with CUDA 11.8 or 12.1 into your Python 3.1x.x environment e.g:

activate pythonenv (activate your python environment)
conda install pytorch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 pytorch-cuda=12.1 -c pytorch -c nvidia
or
conda install pytorch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 pytorch-cuda=11.8 -c pytorch -c nvidia
Set your CUDA Windows environment variables in the command prompt to ensure that CUDA_HOME and CUDA_PATH are set to your Nvidia Cuda Toolkit path. (The folder above the bin folder that nvcc.exe is installed in). Examples are:

set CUDA_HOME=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1
set CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1
or
set CUDA_HOME=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8
set CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8
Navigate to wherever you extracted the deepspeed folder in the Command Prompt:

cd c:\DeepSpeed-0.11.2 (wherever you extracted it to)
Modify the following files:
(These modified files are included in the git-pull of AllTalk, in the DeepSpeed Windows folder and so can just be copied over the top of the exsting folders/files, but if you want to modify them yourself, please follow the below)

deepspeed-0.11.2/build_win.bat - at the top of the file, add:

set DS_BUILD_EVOFORMER_ATTN=0

deepspeed-0.11.2/csrc/quantization/pt_binding.cpp - lines 244-250 - change to:

    std::vector<int64_t> sz_vector(input_vals.sizes().begin(), input_vals.sizes().end());
    sz_vector[sz_vector.size() - 1] = sz_vector.back() / devices_per_node;  // num of GPU per nodes
    at::IntArrayRef sz(sz_vector);
    auto output = torch::empty(sz, output_options);

    const int elems_per_in_tensor = at::numel(input_vals) / devices_per_node;
    const int elems_per_in_group = elems_per_in_tensor / (in_groups / devices_per_node);
    const int elems_per_out_group = elems_per_in_tensor / out_groups;

deepspeed-0.11.2/csrc/transformer/inference/csrc/pt_binding.cpp lines 541-542 - change to:

									 {static_cast<unsigned>(hidden_dim * InferenceContext::Instance().GetMaxTokenLength()),
									  static_cast<unsigned>(k * InferenceContext::Instance().GetMaxTokenLength()),

lines 550-551 - change to:

						 {static_cast<unsigned>(hidden_dim * InferenceContext::Instance().GetMaxTokenLength()),
						  static_cast<unsigned>(k * InferenceContext::Instance().GetMaxTokenLength()),

line 1581 - change to:

		at::from_blob(intermediate_ptr, {input.size(0), input.size(1), static_cast<int64_t>(mlp_1_out_neurons)}, options);

deepspeed-0.11.2/deepspeed/env_report.py line 10 - add:

import psutil

line 83 - 100 - change to:

def get_shm_size():
    try:
        temp_dir = os.getenv('TEMP') or os.getenv('TMP') or os.path.join(os.path.expanduser('~'), 'tmp')
        shm_stats = psutil.disk_usage(temp_dir)
        shm_size = shm_stats.total
        shm_hbytes = human_readable_size(shm_size)
        warn = []
        if shm_size < 512 * 1024**2:
            warn.append(
                f" {YELLOW} [WARNING] Shared memory size might be too small, consider increasing it. {END}"
            )
            # Add additional warnings specific to your use case if needed.
        return shm_hbytes, warn
    except Exception as e:
        return "UNKNOWN", [f"Error getting shared memory size: {e}"]

While still in your command line with python environment enabled run:
build_win.bat and wait 10-20 minutes.
Now cd dist to go into your dist folder and you can now pip install deepspeed-YOURFILENAME.whl (or whatever your WHL file is called).

Removal - If it became necessary to uninstall DeepSpeed, you can do so with cmd_windows.bat and then pip uninstall deepspeed

🟦 Running AllTalk as a standalone app

AllTalk will run as a standalone app, as long as you install its requirements files into whatever Python environment you are using. You can follow the steps to install the AllTalk's requirements into whatever Python environment you wish. Because I dont know what Python environment you are wanting to use, I can only give you a loose set of installation instructions.

Please note, at time of writing, the TTS engine requires Python 3.9.x to 3.11.x TTS Engine details here. AllTalk and its requirements are tested on Python 3.11.x.

🟦 (Option 1) I already have AllTalk installed as an extension of Text-generation-webui

If you already have AllTalk as a extension of Text-generation-webui, and wish to run it as standalone, load Text-generation-webui's Python environment cmd_windows.bat, ./cmd_linux.sh, cmd_macos.sh or cmd_wsl.bat, move into the AllTalk folder cd extensions > cd alltalk_tts and start AllTalk with python script.py. There is nothing beyond this you would need to do.

🟦 (Option 2) I wish to do a custom install of AllTalk

In your chosen disk location git clone this repository:

git clone https://github.com/erew123/alltalk_tts
Move into the alltalk_tts folder:

cd alltalk_tts
If you have a custom Python environment, now is the time to load it!
Install one of the two requirements files. Whichever one of the two is correct for your machine type:

Nvidia graphics card machines - pip install -r requirements_nvidia.txt

Other machines (mac, amd etc) - pip install -r requirements_other.txt
Start AllTalk with python script.py

Deepspeed and other such things can be installed. Please read the relevant instructions for those items, however, make the relevant changes to load your correct Python environment when installing any requirements files and starting AllTalk.

🟠 API Suite and JSON-CURL

🟠Overview

The Text-to-Speech (TTS) Generation API allows you to generate speech from text input using various configuration options. This API supports both character and narrator voices, providing flexibility for creating dynamic and engaging audio content.

🟠 Ready Endpoint

Check if the Text-to-Speech (TTS) service is ready to accept requests.

URL: http://127.0.0.1:7851/api/ready
- Method: GET

curl -X GET "http://127.0.0.1:7851/api/ready"

Response: Ready

🟠 Voices List Endpoint

Retrieve a list of available voices for generating speech.

URL: http://127.0.0.1:7851/api/voices
- Method: GET

curl -X GET "http://127.0.0.1:7851/api/voices"

JSON return: {"voices": ["voice1.wav", "voice2.wav", "voice3.wav"]}

🟠 Preview Voice Endpoint

Generate a preview of a specified voice with hardcoded settings.

URL: http://127.0.0.1:7851/api/previewvoice/
- Method: POST
- Content-Type: application/x-www-form-urlencoded

curl -X POST "http://127.0.0.1:7851/api/previewvoice/" -F "voice=female_01.wav"

Replace female_01.wav with the name of the voice sample you want to hear.

JSON return: {"status": "generate-success", "output_file_path": "/path/to/outputs/api_preview_voice.wav", "output_file_url": "http://127.0.0.1:7851/audio/api_preview_voice.wav"}

🟠 Switching Model Endpoint

URL: http://127.0.0.1:7851/api/reload
- Method: POST

curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=API%20Local"
curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=API%20TTS"
curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=XTTSv2%20Local"

Switch between the 3 models respectively.

curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=XTTSv2%20FT"

If you have a finetuned model in /models/trainedmodel/ (will error otherwise)

JSON return {"status": "model-success"}

🟠 Switch DeepSpeed Endpoint

URL: http://127.0.0.1:7851/api/deepspeed
- Method: POST

curl -X POST "http://127.0.0.1:7851/api/deepspeed?new_deepspeed_value=True"

Replace True with False to disable DeepSpeed mode.

JSON return {"status": "deepspeed-success"}

🟠 Switching Low VRAM Endpoint

URL: http://127.0.0.1:7851/api/lowvramsetting
- Method: POST

curl -X POST "http://127.0.0.1:7851/api/lowvramsetting?new_low_vram_value=True"

Replace True with False to disable Low VRAM mode.

JSON return {"status": "lowvram-success"}

🟠 TTS Generation Endpoint

URL: http://127.0.0.1:7851/api/tts-generate
- Method: POST
- Content-Type: application/x-www-form-urlencoded

🟠 Example command lines

Standard TTS speech Example (standard text) generating a time-stamped file

curl -X POST "http://127.0.0.1:7851/api/tts-generate" -d "text_input=All of this is text spoken by the character. This is text not inside quotes, though that doesnt matter in the slightest" -d "text_filtering=standard" -d "character_voice_gen=female_01.wav" -d "narrator_enabled=false" -d "narrator_voice_gen=male_01.wav" -d "text_not_inside=character" -d "language=en" -d "output_file_name=myoutputfile" -d "output_file_timestamp=true" -d "autoplay=true" -d "autoplay_volume=0.8"

Narrator Example (standard text) generating a time-stamped file

curl -X POST "http://127.0.0.1:7851/api/tts-generate" -d "text_input=*This is text spoken by the narrator* \"This is text spoken by the character\". This is text not inside quotes." -d "text_filtering=standard" -d "character_voice_gen=female_01.wav" -d "narrator_enabled=true" -d "narrator_voice_gen=male_01.wav" -d "text_not_inside=character" -d "language=en" -d "output_file_name=myoutputfile" -d "output_file_timestamp=true" -d "autoplay=true" -d "autoplay_volume=0.8"

Note that if your text that needs to be generated contains double quotes you will need to escape them with \" (Please see the narrator example).

🟠 Request Parameters

🟠 text_input: The text you want the TTS engine to produce. Use escaped double quotes for character speech and asterisks for narrator speech if using the narrator function. Example:

-d "text_input=*This is text spoken by the narrator* \"This is text spoken by the character\". This is text not inside quotes."

🟠 text_filtering: Filter for text. Options:

none No filtering. Whatever is sent will go over to the TTS engine as raw text, which may result in some odd sounds with some special characters.
standard Human-readable text and a basic level of filtering, just to clean up some special characters.
html HTML content. Where you are using HTML entity's like "

-d "text_filtering=none"
-d "text_filtering=standard"
-d "text_filtering=html"

Example:

Standard Example: *This is text spoken by the narrator* "This is text spoken by the character" This is text not inside quotes.
HTML Example: &ast;This is text spoken by the narrator&ast; "This is text spoken by the character" This is text not inside quotes.
None: Will just pass whatever characters/text you send at it.

🟠 character_voice_gen: The WAV file name for the character's voice.

-d "character_voice_gen=female_01.wav"

🟠 narrator_enabled: Enable or disable the narrator function. If true, minimum text filtering is set to standard. Anything between double quotes is considered the character's speech, and anything between asterisks is considered the narrator's speech.

-d "narrator_enabled=true"
-d "narrator_enabled=false"

🟠 narrator_voice_gen: The WAV file name for the narrator's voice.

-d "narrator_voice_gen=male_01.wav"

🟠 text_not_inside: Specify the handling of lines not inside double quotes or asterisks, for the narrator feature. Options:

character: Treat as character speech.
narrator: Treat as narrator speech.

-d "text_not_inside=character"
-d "text_not_inside=narrator"

🟠 language: Choose the language for TTS. Options:

ar Arabic
zh-cn Chinese (Simplified)
cs Czech
nl Dutch
en English
fr French
de German
hu Hungarian
it Italian
ja Japanese
ko Korean
pl Polish
pt Portuguese
ru Russian
es Spanish
tr Turkish

-d "language=en"

🟠 output_file_name: The name of the output file (excluding the .wav extension).

-d "output_file_name=myoutputfile"

🟠 output_file_timestamp: Add a timestamp to the output file name. If true, each file will have a unique timestamp; otherwise, the same file name will be overwritten each time you generate TTS.

-d "output_file_timestamp=true"
-d "output_file_timestamp=false"

🟠 autoplay: Enable or disable playing the generated TTS to your standard sound output device at time of TTS generation.

-d "autoplay=true"
-d "autoplay=false"

🟠 autoplay_volume: Set the autoplay volume. Should be between 0.1 and 1.0. Needs to be specified in the JSON request even if autoplay is false.

-d "autoplay_volume=0.8"

🟠 TTS Generation Response

The API returns a JSON object with the following properties:

status Indicates whether the generation was successful (generate-success) or failed (generate-failure).
output_file_path The on-disk location of the generated WAV file.
output_file_url The HTTP location for accessing the generated WAV file for browser playback.
output_cache_url The HTTP location for accessing the generated WAV file as a pushed download.

Example JSON TTS Generation Response:

{"status":"generate-success","output_file_path":"C:\\text-generation-webui\\extensions\\alltalk_tts\\outputs\\myoutputfile_1704141936.wav","output_file_url":"http://127.0.0.1:7851/audio/myoutputfile_1704141936.wav","output_cache_url":"http://127.0.0.1:7851/audiocache/myoutputfile_1704141936.wav"}

🔴 Future to-do list

Voice output within the command prompt/terminal (TBD).
Correct a few spelling mistakes in the documentation.
Possibly add some additional TTS engines (TBD).
Have a break!

Name		Name	Last commit message	Last commit date
Latest commit History 269 Commits
.github/ISSUE_TEMPLATE		.github/ISSUE_TEMPLATE
.vscode		.vscode
Deepspeed-Windows		Deepspeed-Windows
finetune		finetune
templates		templates
voices		voices
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
confignew.json		confignew.json
diagnostics.py		diagnostics.py
finetune.py		finetune.py
harvard_sentences.txt		harvard_sentences.txt
languages.json		languages.json
modeldownload.json		modeldownload.json
modeldownload.py		modeldownload.py
requirements_finetune.txt		requirements_finetune.txt
requirements_nvidia.txt		requirements_nvidia.txt
requirements_other.txt		requirements_other.txt
script.py		script.py
tts_server.py		tts_server.py

Uh oh!

License

erew123/alltalk_tts

Folders and files

Latest commit

History

Repository files navigation

AllTalk TTS

Index

🔄 Updates

🔄 Updates list & bug fixes list can be found here

🟩 Installation on Text generation web UI

🟩 Other installation notes

🟩 The one thing I cant easily work around

🟪 Updating

🟪 Updating "git pull" error

🟪 Updating other problems

🟫 Screenshots

🟨 Help with problems

🔄 Minor updates/bug fixes list can be found here

🟨 How to make a diagnostics report file

🟨 [AllTalk Startup] Warning TTS Subprocess has NOT started up yet, Will keep trying for 120 seconds maximum. Please wait. It times out after 120 seconds.

🟨 I think AllTalks requirements file has installed something another extension doesn't like

🟨 I activated DeepSpeed in the settings page, but I didnt install DeepSpeed yet and now I have issues starting up

🟨 I am having problems updating/some other issue where it wont start up/Im sure this is a bug

🟨 I am having problems getting AllTalk to start after changing settings or making a custom setup/model setup.

🟨 I see some red "asyncio" messages

🟨 I have multiple GPU's and I have problems running Finetuning

⚫ Finetuning a model

⚫ How will this work/How complicated is it?

⚫ The audio you will use

⚫ Important requirements CUDA 11.8

⚫ Starting Finetuning

⚫ Using a Finetuned model in Text-generation-webui

⚫ Training one model with multiple voices

⚫ Do I need to keep the raw training data/model?

🔵🟢🟡 DeepSpeed Installation Options

🔵 Linux Installation

🟢🟡 Windows Installation

🟢 OPTION 1 - Quick and easy!

🟡 OPTION 2 - A bit more complicated!

🟦 Running AllTalk as a standalone app

🟦 (Option 1) I already have AllTalk installed as an extension of Text-generation-webui

🟦 (Option 2) I wish to do a custom install of AllTalk

🟠 API Suite and JSON-CURL

🟠Overview

🟠 Ready Endpoint

🟠 Voices List Endpoint

🟠 Preview Voice Endpoint

🟠 Switching Model Endpoint

🟠 Switch DeepSpeed Endpoint

🟠 Switching Low VRAM Endpoint

🟠 TTS Generation Endpoint

🟠 Example command lines

🟠 Request Parameters

🟠 TTS Generation Response

🔴 Future to-do list

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 18

Sponsor this project

Uh oh!

Uh oh!

Contributors 12

Uh oh!

Languages