swiss-tts
A Text-to-Speech model trained on Swiss German dialects
overview
Developed a custom TTS model to tackle the unique challenge of Swiss German dialects, which traditionally lack standardized text-to-speech solutions.
tech stack
PythonPyTorchTransformer ArchitectureCUDAAudio Processing
development
Late night training sessions and model fine-tuning
audio samples
Zürich dialect - 'Halli hallo, was lauft bi dier?'
Bern dialect - 'Äuä, das isch guet'
challenges & learnings
- •Handling dialect variations and non-standardized pronunciations
- •Training data collection and preprocessing for Swiss German
- •Model optimization for real-time inference