Publications
CONFERENCE (INTERNATIONAL) BitTTS: Highly Compact Text-to-Speech Using 1.58-bit Quantization and Weight Indexing
Masaya Kawamura, Takuya Hasumi, Yuma Shirahata, Ryuichi Yamamoto
The 26th Annual Conference of the International Speech Communication Association (INTERSPEECH 2025)
August 21, 2025
This paper proposes a highly compact, lightweight text-to-speech (TTS) model for on-device applications. To reduce the model size, the proposed model introduces two techniques. First, we introduce quantization-aware training (QAT), which quantizes model parameters during training to as low as 1.58-bit. In this case, most of 32-bit model parameters are quantized to ternary values {-1, 0, 1}. Second, we propose a method named weight indexing. In this method, we save a group of 1.58-bit weights as a single int8 index. This allows for efficient storage of model parameters, even on hardware that treats values in units of 8-bit. Experimental results demonstrate that the proposed method achieved 83 % reduction in model size, while outperforming the baseline of similar model size without quantization in synthesis quality.
Paper :
BitTTS: Highly Compact Text-to-Speech Using 1.58-bit Quantization and Weight Indexing
(external link)
Software :
https://masayakawamura.github.io/bittts-demo/
(external link)