展示HN:Supertonic 2 – 快速、设备内置的多语言文本转语音系统

1作者: login588大约 1 个月前原帖
Supertonic 现在支持 5 种语言: 韩语 · 西班牙语 · 法语 · 葡萄牙语 · 英语 它被设计为可以在各种环境中本地运行——从设备上的离线使用到通过 WebGPU 的浏览器推理,以及在移动设备和树莓派等小型设备上运行。 以下是主要特点: (1) 超快速度 — 在 M4 Pro 上的响应时间为 0.006 秒 (2) 轻量级 — 66M 参数 (3) 设备端文本转语音 — 完全隐私,零网络延迟 (4) 灵活部署 — 可在浏览器、PC、移动设备和边缘设备上运行 (5) 10 种预设声音 — 选择适合您使用场景的声音 (6) 开放权重模型 — 允许商业使用(OpenRAIL-M) 我们希望 Supertonic 能为许多应用提供帮助。 谢谢! 演示: [https://huggingface.co/spaces/Supertone/supertonic-2](https://huggingface.co/spaces/Supertone/supertonic-2) 模型: [https://huggingface.co/Supertone/supertonic-2](https://huggingface.co/Supertone/supertonic-2) 代码: [https://github.com/supertone-inc/supertonic](https://github.com/supertone-inc/supertonic)
查看原文
Supertonic now supports 5 languages: 한국어 · Español · Français · Português · English<p>It’s built to run locally across setups—from on-device&#x2F;offline use to browser inference with WebGPU, and small devices like mobile and Raspberry Pi.<p>Here are key features:<p>(1) Lightning fast — RTF 0.006 on M4 Pro<p>(2) Lightweight — 66M parameters<p>(3) On-device TTS — Complete privacy, zero network latency<p>(4) Flexible deployment — Runs on browsers, PCs, mobiles, and edge devices<p>(5) 10 preset voices — Pick the voice that fits your use cases<p>(6) Open-weight model — Commercial use allowed (OpenRAIL-M)<p>We hope Supertonic is useful for many applications. Thank you!<p>Demo: <a href="https:&#x2F;&#x2F;huggingface.co&#x2F;spaces&#x2F;Supertone&#x2F;supertonic-2" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;spaces&#x2F;Supertone&#x2F;supertonic-2</a><p>Model: <a href="https:&#x2F;&#x2F;huggingface.co&#x2F;Supertone&#x2F;supertonic-2" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;Supertone&#x2F;supertonic-2</a><p>Code: <a href="https:&#x2F;&#x2F;github.com&#x2F;supertone-inc&#x2F;supertonic" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;supertone-inc&#x2F;supertonic</a>