arXiv cs.CL· Dawei Han, Shidong Li, Shijia Liao, Jiahua Liu, Songting Liu, Qingzheng Wang, Yuxuan Wang, Xin Chen, Yifan Cheng, Tianyu Li, Gangin Park, Xuehao Yang, Chang Zeng, Haoyang Zhang, Ruoyi Zhang, Xiangyu Zhang·· 3 天前AI 评分60
Fish Audio Drama 3 技术报告：35B MoE 开源语音生成模型
Fish Audio Drama 3 Technical Report
AI 导读
Fish Audio Drama 3 是一个开源语音生成模型，基于 35B 参数的 MoE 主干、每个 token 激活 3B 参数，可将文本转为带表演指导的语音。
正文 · 原文
Authors:Dawei Han, Shidong Li, Shijia Liao, Jiahua Liu, Songting Liu, Qingzheng Wang, Yuxuan Wang, Xin Chen, Yifan Cheng, Tianyu Li, Gangin Park, Xuehao Yang, Chang Zeng, Haoyang Zhang, Ruoyi Zhang, Xiangyu Zhang
Abstract:We introduce Fish Audio Drama 3, an open-source speech generation model that turns text into directed performances, built on a 35B-parameter mixture-of-experts backbone that activates 3B parameters per token. It clones voices from multiple references, follows natural-language direction, revises a take when the next turn gives a note, voices multi-character scenes, edits existing speech by infilling, and synthesizes long-form and streaming audio. Multi-turn revision, scene dubbing, editing by infilling, and streaming text input are new in Drama 3, all in one model. Drama 3 pairs the hybrid linear-attention backbone with a new codebook multi-token-prediction audio decoder that predicts the acoustic codebooks of each frame in dedicated layers that attend over the entire sequence. The model is trained with a three-stage pre-training recipe, instruction fine-tuning, and large-batch reinforcement learning whose only reward is a single learned model of human preference. Its inference engine decodes a single stream at 271 frames per second on one H200, against 177 for Fish Audio S2, although Drama 3 has more than seven times as many parameters. The model weights are available on GitHub (this https URL) and Hugging Face (this https URL), and inference is available through SGLang-Omni (this https URL). We highly encourage readers to visit this https URL to direct their own performances.
| Subjects: | Sound (cs.SD); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.18010 [cs.SD] |
| (or arXiv:2610.18010v2 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2610.18010 arXiv-issued DOI via DataCite |
Submission history
From: Qingzheng Wang [view email]
[v1]
Tue, 13 Oct 2026 09:06:28 UTC (698 KB)
[v2]
Wed, 14 Oct 2026 09:28:32 UTC (698 KB)
来源：arXiv cs.CL · arxiv.org