{"kind":"compare","slug":"omnivoice","path":"/compare/omnivoice","title":"OmniVoice 替代方案：CPU 上更快的 TTS | SuperClonica","description":"OmniVoice 与 SuperClonica 对比：600 多种语言加零样本声音克隆，还是在 CPU 上跑得快的小型 Supertonic 3 引擎？速度、音质和许可证一次讲清。","h1":"OmniVoice 与 SuperClonica：语言更多，还是速度更快","intro":"k2-fsa 项目的 OmniVoice 是一款零样本语音合成模型，语言覆盖极广，文档里列了 646 种语言，其中也包括中文。它能用几秒音频克隆声音，也能按属性设计声音。SuperClonica 只支持 31 种语言（不含中文），但在普通 CPU 上运行很快，而且换一句台词音色也很稳定。\n\n下面是两者的区别和选择方法。","sections":[{"heading":"OmniVoice 擅长什么","paragraphs":["OmniVoice 采用类似扩散语言模型的架构，基础模型是 Qwen3-0.6B。"],"points":["语言：官方列表 646 种，训练数据 58.1 万小时。","声音克隆：用 3–10 秒的参考音频和对应文本，无需训练即可复刻。","声音设计：按性别、年龄、音高、口音、耳语等属性生成声音。不过开发者说明，这个模式只用中文和英文数据训练，在部分低资源语言上可能不稳定。","GPU 速度：开发者称 RTF 最低可到 0.025，即实时的 40 倍。","硬件：支持 NVIDIA GPU、Apple Silicon 和 Intel Arc。"]},{"heading":"OmniVoice 在 CPU 上的速度：我们的估算","paragraphs":["在 8 核台式机 CPU 上，我们估计 OmniVoice 比实时慢 2–5 倍，一句 10 秒的话要 20–50 秒。SuperClonica 大约 1.0–1.6 秒，是实时的 6.4–9.7 倍。","估算的出发点是 OmniVoice 的 HTTP 服务 omnivoice-server：它报告在 CPU 上 32 步时 RTF 为 4.92，但没写具体 CPU 型号。我们假设换成较快的台式机 CPU 最多能快 2.5 倍。和表里所有速度一样，这只是估算，不是实测。"]},{"heading":"音质与音色稳定性","paragraphs":["在我们的表里，OmniVoice 和 SuperClonica 的音质都是 4.5 分（满分 5 分）。差别在别处：CPU 性能（2 分对 5 分）和音色稳定性（4 分对 5 分）。像游戏、有声书这种同一个角色要说几百句台词的场景，音色稳定很关键。","OmniVoice 的文档也印证了这一点：克隆是最稳定的模式，参考音频太长会拖慢生成，还可能让效果变差。"]},{"heading":"许可证","paragraphs":["OmniVoice 的代码采用 Apache 2.0，但模型卡写明，由于训练数据的原因，预训练模型采用 CC-BY-NC 许可证。按默认条款，官方权重不能商用。","Supertonic 3 采用 OpenRAIL-M 许可证：免费，没有“仅限非商业”条款，而是列出了禁止用途。在 SuperClonica 购买的音色用于您自己的项目，不能转售。"]},{"heading":"语音合成选 OmniVoice 还是 SuperClonica","paragraphs":["先看语言，再看硬件。"],"points":["OmniVoice：需要中文或 Supertonic 3 不支持的语言，想要即时克隆，并且有 GPU。","SuperClonica：语言在 Supertonic 3 的 31 种之内，需要在 CPU、浏览器或应用里离线快速运行，并让同一个声音读完成千上万句台词。"]}],"faq":[{"question":"OmniVoice 能在 CPU 上运行吗？","answer":"能。omnivoice-server 把 CPU 列为可用模式，RTF 为 4.92，约比实时慢 5 倍。按我们的估算，较快的台式机 CPU 能做到比实时慢 2–5 倍。长文本差距会累积：1 小时音频需要 2–5 小时。"},{"question":"OmniVoice 可以免费商用吗？","answer":"代码是 Apache 2.0，但模型卡显示预训练模型为 CC-BY-NC，因此默认许可证不允许商用。"},{"question":"离线 TTS 有比 OmniVoice 更快的替代方案吗？","answer":"如果您的语言在 Supertonic 3 支持的 31 种之内，可以用 Supertonic 3 音色加 SuperClonica 引擎。它只用 CPU 就能离线运行，按我们的估算在 8 核台式机 CPU 上是实时的 6.4–9.7 倍。"},{"question":"OmniVoice 支持的语言比 SuperClonica 多吗？","answer":"多得多：646 种对 31 种。而且 OmniVoice 支持中文，Supertonic 3 目前不支持。"}],"cta":{"catalog":"试听能在 CPU 上运行的音色","install":"配置 SuperClonica 引擎"},"sources":[{"label":"OmniVoice 代码仓库（GitHub）","url":"https://github.com/k2-fsa/OmniVoice"},{"label":"OmniVoice 支持语言列表","url":"https://github.com/k2-fsa/OmniVoice/blob/master/docs/languages.md"},{"label":"OmniVoice 模型卡（Hugging Face）","url":"https://huggingface.co/k2-fsa/OmniVoice"},{"label":"omnivoice-server 0.2.0：CPU 运行报告","url":"https://pypi.org/project/omnivoice-server/0.2.0/"},{"label":"Supertonic 3 模型卡","url":"https://huggingface.co/supertone-oss-archive/supertonic-3"}],"published":"2026-10-17","updated":"2026-10-23","pages":[{"slug":"text-to-speech","path":"/text-to-speech","h1":"在自己电脑上运行的语音合成"},{"slug":"supertonic","path":"/compare/supertonic","h1":"Supertonic 与 SuperClonica：同一模型，更快更干净"},{"slug":"fish-audio","path":"/compare/fish-audio","h1":"Fish Audio、Fish Speech 与 SuperClonica 对比"},{"slug":"omnivoice","path":"/compare/omnivoice","h1":"OmniVoice 与 SuperClonica：语言更多，还是速度更快"},{"slug":"qwen3-tts","path":"/compare/qwen3-tts","h1":"本地语音合成：Qwen3-TTS 还是 SuperClonica"},{"slug":"voxcpm","path":"/compare/voxcpm","h1":"VoxCPM 2 与 SuperClonica：录音棚音质还是 CPU 速度"},{"slug":"f5-tts","path":"/compare/f5-tts","h1":"F5-TTS 与 SuperClonica：零样本克隆还是 CPU 上的快速合成"},{"slug":"elevenlabs","path":"/compare/elevenlabs","h1":"可离线使用的 ElevenLabs 替代方案"},{"slug":"kokoro","path":"/compare/kokoro","h1":"Kokoro TTS 与 SuperClonica：两个小模型的本地语音合成"},{"slug":"piper","path":"/compare/piper","h1":"Piper TTS 与 SuperClonica：小设备还是定制音色"},{"slug":"xtts","path":"/compare/xtts","h1":"Coqui XTTS-v2 与 SuperClonica：Coqui TTS 的替代方案"},{"slug":"chatterbox","path":"/compare/chatterbox","h1":"Chatterbox TTS 与 SuperClonica：即时克隆还是训练好的音色文件"}],"related":[{"slug":"offline-text-to-speech","path":"/guides/offline-text-to-speech","h1":"如何在自己的电脑上离线把文字转成语音"},{"slug":"voice-cloning","path":"/guides/voice-cloning","h1":"声音克隆：如何得到一个像录音里那样的声音"},{"slug":"best-free-tts-models","path":"/guides/best-free-tts-models","h1":"2026 年适合本地运行的免费语音合成模型"},{"slug":"text-to-speech-for-videos","path":"/guides/text-to-speech-for-videos","h1":"为视频、有声书、游戏和应用做文字转语音配音"},{"slug":"supertonic-3-voices","path":"/guides/supertonic-3-voices","h1":"Supertonic 3 音色：自定义音色的来源与使用方法"},{"slug":"python-text-to-speech","path":"/guides/python-text-to-speech","h1":"Python 文字转语音：离线、免费、自然的音色"},{"slug":"text-to-speech-api-docker","path":"/guides/text-to-speech-api-docker","h1":"在 Docker 中自建文字转语音 API"}]}