Install coqui-ai TTS in Conda on Windows tutorial

[tacitus-git]

clone git
conda create -n coquitts python=3.10
conda activate coquitts
conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia
cd (directory of tts)
pip install -r requirements.txt
python setup.py develop
#use python script to produce tts results

This is not a detailed tutorial, but it is damn better than what I had. Hopefully this will help someone.

[king-dahmanus]

What about a tutorial for those without a GPU?

[casiimir]

You can use the CPU flag, setting torch too:

import torch
from TTS.api import TTS

device = torch.device("cpu")
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=False)

[Hexatona]

Awesome - I had to wade through an hour long video just to figure all this out. You're very to the point - appreciated.

[Cem-Kaya]

thank you, this also worked for me

git clone https://github.com/coqui-ai/TTS # clone repo
conda create -n coquitts python=3.10 # create python env
conda activate coquitts
conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia # works with cuda 12.1
cd (directory of tts)
pip install -e .[all,dev,notebooks] # adding notebooks requirements fixed a bug in notebooks
python setup.py develop

[bobsteele]

As far as I know, only Nvidia offers CUDU in their GPU's. If that is the case, then its suggesting that TTS requires a Nvidia GPU. Am I correct in my understanding?

[TheMechanicX32]

No. There are instructions on the pytorch website for installing pytorch on CPU-only devices. I am running xtts on a 6-year-old MacBook Pro perfectly fine.

Please hit me up if anyone has any questions.

[jj449]

clone git conda create -n coquitts python=3.10 conda activate coquitts conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia cd (directory of tts) pip install -r requirements.txt python setup.py develop #use python script to produce tts results

This is not a detailed tutorial, but it is damn better than what I had. Hopefully this will help someone.

Thank you so much , very useful . Installed successfully according this !

[Panquesito0]

I tried your method using a MacBook Air M1. After creating an enviornment on Conda using Python 3.9.19 with the name "ai_stuff".
I continue with the next steps:

git clone https://github.com/coqui-ai/TTS
conda activate ai_stuff
cd (directory of tts)
pip install -r requirements.txt
python setup.py develop

Thanks. Really helpful.

[VictorJulianiR]

for those that don't want to install conda, this line did the trick for me:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

[TheMechanicX32]

I totally understand the desire to avoid Conda, but conda is really useful for building environments for XTTS or something along those lines. Allows me to create an environment where I can install very specific versions of packages without affecting my other environments.