Training for Turkish Support
I want to train this model for Turkish. Is it possible? Can I add new language by training?
Just released this today: https://github.com/Sedherthe/llm-tts-factory. Rough scripts on how you can train your own model. It's still very messy though, but gives you an idea how you can do it. I plan to structure it nicely so anybody can just run the scripts on their data to get a good model, but could take time
Has there been any recent progress in fine-tuning base models? If I fine-tune a model on a new language, can it actually gain the ability to speak or understand that language effectively?
Sometimes, for fine-tuning on new language it does help to take a pre-trained model as base. This model would lose all the old language's capabilities. But, might help. It could be as effective as training from scratch. Understanding a new language is much difficult than understanding a new voice.