What's Changed

• Run XTTS models by direct name with versions by @erogol in https://github.com/coqui-ai/TTS/pull/3318
• fix: correctly strip/restore initial punctuation by @eginhard in https://github.com/coqui-ai/TTS/pull/3336
• Fix link to installation instructions by @Vuizur in https://github.com/coqui-ai/TTS/pull/3329

New Contributors

• @Vuizur made their first contribution in https://github.com/coqui-ai/TTS/pull/3329

Full Changelog: https://github.com/coqui-ai/TTS/compare/v0.21.1...v0.21.2

This PR allows for running XTTS models with version tags. So you the user can access any version they like.

from TTS.api import TTS

# get v2.0.2
tts = TTS(model_name="xtts_v2.0.2", gpu=True)

# get the latest version
tts = TTS(model_name="xtts", gpu=True)

# generate speech by cloning a voice using default settings
tts.tts_to_file(text="Here is my sample text.",
                file_path="output.wav",
                speaker_wav=["reference.wav", "reference1.wav"],
                language="en")

Making automatic sentence splitting optional. So you can apply any custom logic for processing the text before passing it to the model. Set split_sentences False.

from TTS.api import TTS

# get v2.0.2
tts = TTS(model_name="xtts_v2.0.2", gpu=True)

# generate speech by cloning a voice using default settings
tts.tts_to_file(text="Here is my sample text.",
                file_path="output.wav",
                speaker_wav=["reference.wav", "reference1.wav"],
                language="en",
                split_sentences=False)