Fine-Tuning Chatterbox for Multilingual & Emotion-Aware Conversations (Persian Included)
Hi ResembleAI Team,
We really appreciate your work on Chatterbox — it’s an impressive conversational model with strong emotional understanding and expressive dialogue abilities.
We’d like to contribute by helping extend Chatterbox to new languages, especially Persian (Farsi). Our goal is to improve multilingual performance and also integrate emotional and affective layers into the responses to make interactions feel more natural across cultures.
We have access to powerful compute resources (GPU clusters) and a dedicated team ready to work on data curation and fine-tuning.
Could you please share a technical guide or documentation on how to:
Fine-tune or extend Chatterbox on new datasets.
Adjust the emotional tone or expression in dialogue.
Properly evaluate and align the model for new languages.
We’d be glad to collaborate or follow your best practices if there’s an internal pipeline or API structure for model training.
Thanks again for making this model public — we’re excited to help expand it further!
Best,
Hamed
The model won’t be released publicly, which means finding the full training code will be quite difficult.