Logo
Published on

XTTS-v2 Text to Speech Installation guide

Authors
  • Name
    Francesco C

This is a synthetic guide to my experiments with TTS models using XTTS-v2. The other guides of TTS are: Bark Guide, Cloning anyone voice tutorial, Create a million dollar business with this simple pipeline.

Follow me

Frank Since I was 12 I loved to spend time at the computer, from programming to hacking and gaming. This channel is a…www.youtube.com

XTTS VS Bark

After testing both models, this model produce much better quality

for example:

This was done with Bark, and the video below with XTTS-v2

https://www.youtube.com/playlist?list=PLEST2cB76tXvU206gFPeIaCEbHMzTuk7q

Guide

1

git clone https://huggingface.co/coqui/XTTS-v2

2

 python -m venv .env

3

Activate virtual environment (my guide to virtual environments)

# Activate the virtual environment# On Windows:.\.env\Scripts\activate# On Linux/macOS:source .env/bin/activate

4

install requirements.txt

pip install -r requirements.txt

create a requirements.txt with the following dependencies

absl-py==2.1.0aiohappyeyeballs==2.4.3aiohttp==3.11.2aiosignal==1.3.1annotated-types==0.7.0anyascii==0.3.2asttokens==2.4.1attrs==24.2.0audioread==3.0.1babel==2.16.0bangla==0.0.2blinker==1.9.0blis==0.7.11bnnumerizer==0.0.2bnunicodenormalizer==0.1.7catalogue==2.0.10certifi==2024.8.30cffi==1.17.1charset-normalizer==3.4.0click==8.1.7cloudpathlib==0.20.0colorama==0.4.6confection==0.1.5contourpy==1.3.1coqpit==0.0.17cycler==0.12.1cymem==2.0.8Cython==3.0.11dateparser==1.1.8decorator==5.1.1docopt==0.6.2einops==0.8.0encodec==0.1.1executing==2.1.0filelock==3.13.1Flask==3.1.0fonttools==4.55.0frozenlist==1.5.0fsspec==2024.2.0g2pkk==0.1.2grpcio==1.68.0gruut==2.2.3gruut-ipa==0.13.0gruut_lang_de==2.0.1gruut_lang_en==2.0.1gruut_lang_es==2.0.1gruut_lang_fr==2.0.2hangul-romanize==0.1.0huggingface-hub==0.26.2idna==3.10inflect==7.4.0ipython==8.29.0itsdangerous==2.2.0jamo==0.4.1jedi==0.19.2jieba==0.42.1Jinja2==3.1.3joblib==1.4.2jsonlines==1.2.0kiwisolver==1.4.7langcodes==3.4.1language_data==1.2.0lazy_loader==0.4librosa==0.10.2.post1llvmlite==0.43.0marisa-trie==1.2.1Markdown==3.7markdown-it-py==3.0.0MarkupSafe==2.1.5matplotlib==3.9.2matplotlib-inline==0.1.7mdurl==0.1.2more-itertools==10.5.0mpmath==1.3.0msgpack==1.1.0multidict==6.1.0murmurhash==1.0.10networkx==2.8.8nltk==3.9.1num2words==0.5.13numba==0.60.0numpy==1.26.3packaging==24.2pandas==1.5.3parso==0.8.4pillow==10.2.0platformdirs==4.3.6pooch==1.8.2preshed==3.0.9prompt_toolkit==3.0.48propcache==0.2.0protobuf==5.28.3psutil==6.1.0pure_eval==0.2.3pycparser==2.22pydantic==2.9.2pydantic_core==2.23.4Pygments==2.18.0pynndescent==0.5.13pyparsing==3.2.0pypinyin==0.53.0pysbd==0.3.4python-crfsuite==0.9.11python-dateutil==2.9.0.post0pytz==2024.2PyYAML==6.0.2regex==2024.11.6requests==2.32.3rich==13.9.4safetensors==0.4.5scikit-learn==1.5.2scipy==1.14.1shellingham==1.5.4six==1.16.0smart-open==7.0.5soundfile==0.12.1soxr==0.5.0.post1spacy==3.7.5spacy-legacy==3.0.12spacy-loggers==1.0.5srsly==2.4.8stack-data==0.6.3SudachiDict-core==20241021SudachiPy==0.6.8sympy==1.13.1tensorboard==2.18.0tensorboard-data-server==0.7.2thinc==8.2.5threadpoolctl==3.5.0tokenizers==0.20.3torch==2.5.1+cu124torchaudio==2.5.1+cu124torchvision==0.20.1+cu124tqdm==4.67.0trainer==0.0.36traitlets==5.14.3transformers==4.46.2typeguard==4.4.1typer==0.13.0typing_extensions==4.12.2tzdata==2024.2tzlocal==5.2umap-learn==0.5.7Unidecode==1.3.8urllib3==2.2.3wasabi==1.1.3wcwidth==0.2.13weasel==0.4.1Werkzeug==3.1.3wrapt==1.16.0yarl==1.17.1

5

The installation can take up to 40 min, so don’t worry if is super slow

pip install TTS==0.22.0 --no-deps

You are going to read a lot of this stuffs and that’s okay, in the beginning i thought it was broken but i went for a walk and back and it was fine.

Collecting transformers>=4.33.0 (from TTS)  Obtaining dependency information for transformers>=4.33.0 from https://files.pythonhosted.org/packages/75/d5/294a09a62bdd88da9a1007a341d4f8fbfc43be520c101e6afb526000e9f4/transformers-4.46.1-py3-none-any.whl.metadata  Using cached transformers-4.46.1-py3-none-any.whl.metadata (44 kB)  Obtaining dependency information for transformers>=4.33.0 from https://files.pythonhosted.org/packages/f9/9d/030cc1b3e88172967e22ee1d012e0d5e0384eb70d2a098d1669d549aea29/transformers-4.45.2-py3-none-any.whl.metadata  Using cached transformers-4.45.2-py3-none-any.whl.metadata (44 kB)

6

Visit this page and install pytorch

https://pytorch.org/get-started/locally/

Resources:

Stackademic 🎓

Thank you for reading until the end. Before you go: