Coqui has introduced a Python API for TTS since it was added to Manim Voiceover. This means that the workaround implemented to call it can be removed and the service simplified.
This also removes the word boundary generation logic that was previously implemented, but we have to do it anyway since TTS has been refactored and its latest version doesn't work with the workaround. We can fall back to word boundary detection with Whisper instead: CoquiService(transcription_model='base').
Coqui has introduced a Python API for TTS since it was added to Manim Voiceover. This means that the workaround implemented to call it can be removed and the service simplified.
This also removes the word boundary generation logic that was previously implemented, but we have to do it anyway since TTS has been refactored and its latest version doesn't work with the workaround. We can fall back to word boundary detection with Whisper instead:
CoquiService(transcription_model='base').