[{"data":1,"prerenderedAt":173},["ShallowReactive",2],{"repo-phantivia-cortico-world-gpt-sovits-for-coopanion-zh-CN":3,"i18n-phantivia-cortico-world-gpt-sovits-for-coopanion-zh-CN":170},{"mirrorPool":4,"detail":13,"relatedProjects":49},[5,9],{"key":6,"label":7,"base":8,"type":6},"github","mirror.github","https:\u002F\u002Fgithub.com\u002F",{"key":10,"label":11,"base":12,"type":10},"auto-proxy","mirror.autoProxy","https:\u002F\u002Fgithub.xxlab.tech\u002F",{"owner":14,"repo":15,"repoId":16,"repoUrl":17,"starCount":18,"language":19,"description":20,"languages":21,"screenshots":25,"homepage":26,"pushedAt":27,"ownerAvatar":28,"ownerName":14,"ownerBio":29,"lastSyncTime":30,"lastReleaseTime":31,"tags":32,"license":33,"category":34,"subgroup":35,"releases":36},"Phantivia","cortico-world-gpt-sovits-for-coopanion",1423986824,"https:\u002F\u002Fgithub.com\u002FPhantivia\u002Fcortico-world-gpt-sovits-for-coopanion",1,"TypeScript","Cortico World for Coopanion: the pet's lines spoken with GPT-SoVITS (V5 on NVIDIA GPUs, v2ProPlus on CPU)",{"TypeScript":22,"Shell":23,"JavaScript":24,"CSS":18},93.1,4.5,1.4,[],"","2026-11-04","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F52902663?v=4","NLPer","2026-11-10T18:09:22.676736+00:00","2026-10-18T04:37:06Z",[],"MIT","ai","AI语音工具",[37],{"tag":38,"publishAt":31,"semVer":39,"prerelease":41,"assets":42},"runtime-e91506f-1",[40,40,40],0,false,[43],{"fileName":44,"sizeMb":45,"sha256":46,"downloadCount":47,"originUrl":48},"gpt-sovits-e91506f-1-win-x64-cpu.zip",988.3,"9ce348074e941388544b4db072ee0b9cfc50c9bd420e61f13ee9364fe90152e1",268,"https:\u002F\u002Fgithub.com\u002FPhantivia\u002Fcortico-world-gpt-sovits-for-coopanion\u002Freleases\u002Fdownload\u002Fruntime-e91506f-1\u002Fgpt-sovits-e91506f-1-win-x64-cpu.zip",[50,68,104,134,157],{"owner":51,"repo":52,"repoId":53,"category":34,"subgroup":35,"language":54,"tags":55,"starCount":60,"ownerUid":61,"latestTag":62,"platforms":63,"lastReleaseTime":65,"downloadCount":66,"description":67},"pbakaus","impeccable",1097346685,"Rust",[56,34,57,58,59],"agent-skills","anti-slop","design","slop",109440,43004,"skill-v4.9.0",[64],"other","2026-11-13T06:18:33Z",5463,"一个为AI编码代理提供设计指导的工具，包含1个技能、23个命令、59条确定性检测规则和实时浏览器迭代功能，帮助AI生成更优秀的前端设计。",{"owner":69,"repo":70,"repoId":71,"category":34,"subgroup":35,"language":72,"tags":73,"starCount":92,"ownerUid":93,"latestTag":94,"platforms":95,"lastReleaseTime":101,"downloadCount":102,"description":103},"debpalash","VoiceStudio",1206390571,"Python",[34,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91],"audiobook","cuda","dubbing","elevenlabs-alternative","huggingface","local-first","mlx","omnivoice-studio","speech-to-text","text-to-speech","transcription","translate","tts","voice-ai","voice-cloning","voice-generation","voicestudio","workflow",109017,4178343,"v0.5.8",[96,97,98,99,100],"windows","mac","linux","android","ios","2026-11-12T08:53:29Z",10774,"一款完全本地运行的开源AI语音工作室，支持语音克隆、语音设计、视频配音、听写、转录和有声书创作，支持646种语言，是ElevenLabs的开源替代品。",{"owner":105,"repo":106,"repoId":107,"category":34,"subgroup":35,"language":72,"tags":108,"starCount":127,"ownerUid":128,"latestTag":129,"platforms":130,"lastReleaseTime":131,"downloadCount":132,"description":133},"modelscope","FunASR",569959091,[109,110,111,112,113,114,115,116,117,118,119,120,121,122,82,123,84,124,125,126],"asr","audio","chinese","emotion-recognition","funasr","mcp-server","multilingual-asr","openai-compatible-api","paraformer","punctuation","pytorch","real-time-asr","speaker-diarization","speech-recognition","streaming-asr","vllm","voice-activity-detection","whisper-alternative",21017,109945100,"runtime-torch-v0.1.7",[64],"2026-11-13T09:50:06Z",11,"FunASR 是阿里巴巴开源的工业级语音识别工具包（Python），覆盖离线与流式转写、端点检测、标点恢复与说话人分离，支持 CPU\u002FGPU 部署并提供 OpenAI 兼容 API 与 MCP Server。",{"owner":135,"repo":136,"repoId":137,"category":34,"subgroup":35,"language":19,"tags":138,"starCount":150,"ownerUid":151,"latestTag":152,"platforms":153,"lastReleaseTime":154,"downloadCount":155,"description":156},"leemysw","yovoice",1371687666,[139,110,140,141,142,143,144,145,146,147,83,86,148,88,149,96],"ai-music","desktop-app","local-ai","macos","mcp","midi","music-generation","offline-ai","skill","voice","voice-cloning-app",2166,40636507,"v0.7.0",[96,97,98],"2026-11-11T12:55:18Z",181,"yovoice 是开源的本地优先 AI 音频与音乐创作工作室，在 macOS 与 Windows 上离线完成语音合成、声音克隆、带歌词的歌曲生成与多轨乐谱，零云 API 费用。",{"owner":158,"repo":159,"repoId":160,"category":34,"subgroup":35,"language":161,"tags":162,"starCount":163,"ownerUid":164,"latestTag":165,"platforms":166,"lastReleaseTime":167,"downloadCount":168,"description":169},"Gary23333","opentts",1356280885,"Swift",[],2,120565299,"v5.0.3",[97],"2026-09-03T18:33:50Z",6,"锻造回声 ForgedEcho 是一款 macOS 原生的声音克隆与 AI 语音生产工作台，把多种开源语音大模型封装为图形界面，支持声音克隆、情绪控制、字幕批量配音、有声书生成与本地 LoRA 微调，全程离线运行。",{"description":171,"readmeHtml":172},"想让 Coopanion 桌宠开口说话，GPTSovits 语音扩展在 pet_say 前调用本机 GPT-SoVITS 流式合成把回复念出来，支持显卡与 CPU 两档推理、预设与自定义音色，配置位于 worlds.gpt-sovits。","\u003Ch2 id=\"项目概述\">项目概述\u003C\u002Fh2>\u003Cp>这个扩展由 Phantivia 开发，是 Coopanion 扩展体系里的语音 World：Coo 每次 \u003Ccode>pet_say\u003C\u002Fcode> 之前在同一次回复里先调用 \u003Ccode>sovits_speak\u003C\u002Fcode>，扩展用本机 GPT-SoVITS 流式合成这句话并从声卡播出。\u003Ccode>sovits_speak\u003C\u002Fcode> 等第一块音频写进声卡才返回，所以紧跟其后的 \u003Ccode>pet_say\u003C\u002Fcode> 气泡与声音同步出现。\u003C\u002Fp>\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>项目\u003C\u002Fth>\n\u003Cth>内容\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>宿主\u003C\u002Ftd>\n\u003Ctd>Coopanion（桌宠）\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>推理引擎\u003C\u002Ftd>\n\u003Ctd>GPT-SoVITS api_v2\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>推理档\u003C\u002Ftd>\n\u003Ctd>显卡 V5 turbo、CPU v2ProPlus，默认自动选择\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>配置\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>worlds.gpt-sovits\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\u003C\u002Ftable>\n\u003Cblockquote>\n\u003Cp>桌宠开口说话，声音由本机 GPT-SoVITS 实时合成。\u003C\u002Fp>\n\u003C\u002Fblockquote>\n\u003Ch2 id=\"主要功能\">主要功能\u003C\u002Fh2>\u003Cul>\n\u003Cli>\u003Cstrong>两档推理\u003C\u002Fstrong>：有 NVIDIA 显卡走显卡档（V5 turbo，实测首包 0.5–1.0 秒、显存 3.5 GB），其余走 CPU 档（v2ProPlus，首包 0.8–1.1 秒、内存 5.8 GB）\u003C\u002Fli>\n\u003Cli>\u003Cstrong>预设音色\u003C\u002Fstrong>：随包带女声、男声两个音色（由 MOSS-VoiceGenerator 生成，参数在 \u003Ccode>assets\u002Fvoices\u002Fprompts.json\u003C\u002Fcode>），也可上传 3–10 秒自定义参考音频\u003C\u002Fli>\n\u003Cli>\u003Cstrong>三种后端接入\u003C\u002Fstrong>：已有服务直连 \u003Ccode>serverUrl\u003C\u002Fcode>、自备 GPT-SoVITS 目录填 \u003Ccode>runtimeDir\u003C\u002Fcode>、或在面板上「安装并启动」托管运行时（独立 CPython 3.12 与权重，按 SHA-256 校验）\u003C\u002Fli>\n\u003Cli>\u003Cstrong>工具与事件\u003C\u002Fstrong>：\u003Ccode>sovits_speak(text)\u003C\u002Fcode> 排队合成，\u003Ccode>sovits_hush()\u003C\u002Fcode> 停止并清空队列，事件 \u003Ccode>gpt-sovits.state\u003C\u002Fcode> 广播语音可用性\u003C\u002Fli>\n\u003Cli>\u003Cstrong>参数可调\u003C\u002Fstrong>：语速、音量（0–1）、输出设备名、提示词语言（默认 \u003Ccode>zh\u003C\u002Fcode>）均可在配置页修改\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2 id=\"安装与启用\">安装与启用\u003C\u002Fh2>\u003Col>\n\u003Cli>在 Coopanion 高级模式打开「GPTSovits 语音」设置页\u003C\u002Fli>\n\u003Cli>默认 \u003Ccode>tier: auto\u003C\u002Fcode>：\u003Ccode>nvidia-smi\u003C\u002Fcode> 列得出显卡就用显卡档，首次启动显卡档要编译 CUDA 内核（实测第一次 74 秒，之后约 11 秒）\u003C\u002Fli>\n\u003Cli>中、英、日各一句预热，预热完成才算可用\u003C\u002Fli>\n\u003Cli>不装扩展或未调 \u003Ccode>sovits_speak\u003C\u002Fcode> 时，\u003Ccode>pet_say\u003C\u002Fcode> 保持静音，不影响桌宠本体\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Ch2 id=\"权限与影响范围\">权限与影响范围\u003C\u002Fh2>\u003Cul>\n\u003Cli>\u003Cstrong>本机服务\u003C\u002Fstrong>：推理服务以 \u003Ccode>runtime\u002Fpython.exe -I -X utf8 -u api_v2.py\u003C\u002Fcode> 启动，只监听 127.0.0.1 的空闲端口\u003C\u002Fli>\n\u003Cli>\u003Cstrong>磁盘占用\u003C\u002Fstrong>：托管安装把运行时与权重装到部署根的 \u003Ccode>runtimes\u002Fgpt-sovits\u002F<版本>\u002F<档>\u002F\u003C\u002Fcode> 与 \u003Ccode>models\u002Fgpt-sovits\u002Fpretrained_models\u002F\u003C\u002Fcode>\u003C\u002Fli>\n\u003Cli>\u003Cstrong>音量补偿\u003C\u002Fstrong>：v2ProPlus 输出电平低（峰值约 −27 dBFS），播放时自动乘 8 倍并软限幅\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2 id=\"注意事项\">注意事项\u003C\u002Fh2>\u003Cul>\n\u003Cli>\u003Cstrong>显卡档不适合 CPU 跑\u003C\u002Fstrong>：V5 turbo 在 CPU 上首包 3–4 秒、实时率 1.2–1.5 跟不上播放，所以 CPU 档固定用 v2ProPlus\u003C\u002Fli>\n\u003Cli>\u003Cstrong>改档需重启\u003C\u002Fstrong>：\u003Ccode>tier\u003C\u002Fcode> 与后端相关配置改动后要重启推理服务生效\u003C\u002Fli>\n\u003Cli>\u003Cstrong>状态以事件为准\u003C\u002Fstrong>：环境提示词里的语音状态只是前缀重建时的快照，实时可用性看 \u003Ccode>gpt-sovits.state\u003C\u002Fcode> 事件\u003C\u002Fli>\n\u003C\u002Ful>\n",1794593785463]