SemBridge: Semantic Token Anchoring for Continuous-Latent Autoregressive Speech Generation

Hanke Xie, Haopeng Lin, Jiale Qian, Dake Guo, Yuepeng Jiang, Zhichao Wang, Wenxiao Cao, Jingbin Hu, Guobin Ma, Wenhao Li, Huakang Chen, Chengyou Wang, Ming Tao, Zhonghua Fu, Lei Xie, Xinsheng Wang

Paper Demo Code Model

• A training-only semantic-token anchoring framework that directly supervises the states of a continuous-latent AR language model with discrete semantic tokens, while inference remains entirely continuous.
• Consistently improves content accuracy (WER/CER) on both zero-shot TTS and score-conditioned singing voice synthesis.

SoulX-Podcast: Towards Realistic Long-form Podcasts with Dialectal and Paralinguistic Diversity

Hanke Xie^*, Haopeng Lin^*, Wenxiao Cao, Dake Guo, Wenjie Tian, Jun Wu, Hanlin Wen, Ruixuan Shang, Hongmei Liu, Zhiqi Jiang, Yuepeng Jiang, Wenxi Chen, Ruiqi Yan, Jiale Qian, Yichao Yan, Shunshun Yin, Ming Tao, Xie Chen, Lei Xie, Xinsheng Wang

Paper Demo Code Model

• A podcast-style multi-turn, multi-speaker dialogue speech generation system with paralinguistic controls, supporting Mandarin, English and Chinese dialects (Sichuanese, Henanese and Cantonese).
• Continuously generates over 90 minutes of conversation with stable speaker timbre and smooth speaker transitions.

FlashTTS: Fast Streaming TTS with MTP Acceleration and X-pred Mean Flow Distillation

Hanke Xie, Xiaming Ren, Dake Guo, Ruonan You, Wenhao Li, Jingbin Hu, Guobin Ma, Huakang Chen, Kejie Xu, Rui Huang, Weiguo Tan, Xianrong Wang, Lei Xie

Paper Demo Code

• An open-source low-latency streaming TTS framework with a lagged multi-track architecture for native streaming text/speech input, parallel Multi-Token Prediction (MTP), and a 2-NFE X-pred mean-flow decoder.
• Reduces first-packet latency to 325 ms while preserving strong zero-shot voice cloning.

X-Pred MeanFlow for Streaming Token-to-Mel Speech Decoding

Hanke Xie, Xiaming Ren, Qirui Zhan, Jingbin Hu, Wenhao Li, Haoyu Zhang, Ruonan You, Chengyou Wang, Yunxiang Chen, Houdun Liu, Su Feng, Lei Xie

Paper Demo

• A few-step streaming token-to-mel decoder that reparameterizes MeanFlow with mel-space prediction, together with layer-selective block-wise attention for chunk-wise streaming generation.

DialoSpeech: Dual-Speaker Dialogue Generation with LLM and Flow Matching

Hanke Xie, Dake Guo, Chengyou Wang, Yue Li, Wenjie Tian, Xinfa Zhu, Xinsheng Wang, Xiulin Li, Guanqiong Miao, Bo Liu, Lei Xie

Paper Demo

• A dual-track architecture combining an LLM with chunked flow matching for human-like dual-speaker dialogue generation with coherent turn-taking and natural overlaps, plus a pipeline for building dual-track dialogue data.