Request to Make Model Weights Public for Community Learning and Research
Hello Fahad Qazi 👋
First of all, thank you for your amazing contribution through this model! 🙌 It looks very promising and could be incredibly useful for the research and developer community.
I wanted to kindly ask whether you have any plans to make the model weights public in the future. Many of us would love to explore and learn from your work — especially for educational and non-commercial purposes.
Open access to the weights could help students, researchers, and AI enthusiasts experiment, fine-tune, and contribute to improving the model further.
We completely understand if there are licensing or project-related reasons for keeping it private, but it would be great to know if an open-source release is on your roadmap.
Thanks again for your hard work and contribution to the community! 💡
Warm regards,
Amanat Ali
Hello Amanat,
I am so sorry to not have been able to respond to this earlier. I built these models for a larger purpose which was to make an end-to-end system that could dub English and Hindi/Urdu videos into Sindhi so that Sindhi language could be promoted. Over the past year, we have played with and trained so many models for Sindhi, and we learned so much from them and have finally incorporated the best parts from all of them to create our own end-to-end audio dubbing model.
That being said, I originally did not think that I will make the weights of these tts and stt models public. But soon when we finish building our dubbing system and release it, we will release weights of these tts & stt models to the public as well.
Regards,
Fahad Maqsood Qazi