Voice consistency

#1
by johnblues - opened

The voice seems to change a bit for every sentence. Is there any way to ensure consistency across each sentence?

Hey John, not completely possible with this model. Giving a name to the description helps but not perfect. You need to train a model with data from only one speaker. I will upload a tutorial soon on how to do it. Will keep you posted.

Ok, thank you. I will look forward to that.

Hey John, I've just created a tutorial on how to fine-tune a text-to-speech model in any language. If your dataset consists of only one person's voice, you'll get great voice consistency in the output.

The tutorial video:
https://www.youtube.com/watch?v=TZIBQ24UCgA

Model weights:
https://huggingface.co/emirhanbilgic/speecht5_finetuned_emirhan_tr

Training codes:
https://github.com/emirhanbilgic/Turkish-TTS

Dataset I use:
https://huggingface.co/datasets/erenfazlioglu/turkishvoicedataset

Hugging face demo to try:
https://huggingface.co/spaces/emirhanbilgic/Text-to-speech-Turkish

Sign up or log in to comment