{"slug":"fish-audio-drama-3","name":"Fish Audio Drama 3","nameZh":"Fish Audio Drama 3（多角色长文本 TTS，35B-A3B）","category":"speech","categoryName":"Speech recognition & synthesis","audience":["ai-era"],"primary":"ai-era","openClosed":"open","pricing":"open-source","license":"Fish Audio Research License","maintained":"yes","officialUrl":"https://huggingface.co/fishaudio/drama-3","repo":null,"latestVersion":null,"latestRelease":null,"stars":null,"starsAsOf":null,"description":"Fish Audio’s next-gen TTS model (HF, 2026-10-01): a 35B-parameter MoE text backbone with 3B active per token plus a 10-codebook audio decoder; up to 16 characters per text via speaker tags, voice cloning, natural-language tags like [sighs]/[laughs], automatic chunking with streaming for long text; supported by SGLang-Omni and vLLM-Omni from day one. Research licence — check the terms before commercial use","descriptionZh":"Fish Audio 新一代语音合成模型（HF 2026-10-01）：35B 参数、每 token 激活 3B 的 MoE 文本骨干 + 10 码本音频解码器，一段文本最多 16 个角色（speaker 标签），支持声音克隆、[sighs]/[laughs] 等自然语言控制标签，长文本自动分段并可流式输出；首发即支持 SGLang-Omni 和 vLLM-Omni。许可为研究许可，商用前请看条款","callMethods":[],"platforms":[],"addedAt":"2026-10-22","updatedAt":null,"topics":[],"language":null,"lastPush":null,"formerly":null,"packages":[],"recentVersions":[],"alternatives":["whisper","voicestudio","gpt-sovits","voicebox","vibevoice","chattts","voxcpm","fish-speech"],"stacks":[],"scenarios":[],"problems":[],"reasonZh":null,"sources":["editorial"],"record":"https://aimap.page/api/v1/software/fish-audio-drama-3.json","pages":{"en":"https://aimap.page/en/software/fish-audio-drama-3/","zh-hans":"https://aimap.page/zh-hans/software/fish-audio-drama-3/"}}