You can now fine-tune my 3.96M-parameter TTS on your own voice or language
A new toolkit enables fine-tuning a 3.96M-parameter TTS model on your voice or language.
Following the release of Inflect v2, users expressed a strong interest in training the TTS model on their own voice or language. In response, the developer created a toolkit that allows users to fine-tune the model using their single-speaker recordings and transcripts. This new toolkit facilitates warm-starting either Nano or Micro models, resuming training, and exporting results to PyTorch or ONNX.