Training Model

[blessziamah]

Is it possible for me to train a new model with a dataset that contains only voice recordings? No transcript

[samarasimhapeyala]

Training a text-to-speech model typically requires both audio recordings and corresponding transcripts (text) to learn the mapping between spoken sounds and written language. If you only have voice recordings without transcripts, you would face a significant challenge because the model needs the text to understand what to synthesize. You can use ASR models like whisper to generate transcripts for your audios.

Please mark it answered if your doubt is resolved.