Hey HN,
I'm excited to share Ind-QwenTTS, a mini TTS system I trained from scratch focusing on Indian languages and accents. It's built on top of Qwen2.5-0.5B and uses SNAC for discrete audio codecs, turning speech synthesis into a causal language modeling task. The goal was to address the lack of high-quality TTS for low-resource Indian setups like Indian-accented English