Use espeak-ng for coqui-ai/TTS VITS English models. #466

csukuangfj · 2023-12-05T11:18:24Z

Usage

git clone https://github.com/k2-fsa/sherpa-onnx
cd sherpa-onnx
mkdir build
cmake ..
make -j

wget https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-coqui-en-ljspeech.tar.bz2
tar xf vits-coqui-en-ljspeech.tar.bz2


./bin/sherpa-onnx-offline-tts-play \
  --vits-model=./vits-coqui-en-ljspeech/model.onnx \
  --vits-tokens=./vits-coqui-en-ljspeech/tokens.txt \
  --vits-data-dir=./vits-coqui-en-ljspeech/espeak-ng-data \
  --num-threads=3 \
  --output-filename=./abc.wav \
  --debug=1 \
"Friends fell out often because life was changing so fast. The easiest thing in the world was to lose touch with someone. The best day of the month was payday. But in a way it was the worst day, too. After you had worked hard for so long, it was infuriating to see how much money had been docked for silly things: being a few minutes late one morning, or taking a half day off for feeling sick, or having to pay extra when the winter uniforms switched to summer ones. On payday, everyone crowded the post office to wire money to their families. Girls who had just come out from home were crazy about sending money back, but the ones who had been out longer laughed at them. Some girls set up savings accounts for themselves, especially if they already had boyfriends. Everyone knew which girls were the best savers and how many thousands they had saved. Everyone knew the worst savers, too, with their lip gloss and silver mobile phones and heart-shaped lockets and their many pairs of high-heeled shoes. The only way to find a better job was to quit the one you had. Interviews took time away from work, and a new hire was expected to start right away. Leaving a job was also the best guarantee of getting a new one: The pressing need for a place to eat and sleep was incentive to find work fast. Girls often quit a factory in groups, finding courage in numbers and pledging to join a new factory together, although that usually turned out to be impossible."

You can find 3 VITS English models from coqui-ai/TTS below
https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models