parameter footprint
0.1B
Released Apr 10, 2026
Realtime voice cloning
OpenMOSS Team
MOSI.AI
Fudan NLP Lab
Multilingual Tiny Speech Generation Model
Build faster voice products with MOSS-TTS-Nano.
A deployment-first TTS model designed for realtime speech generation, voice cloning, and lightweight integration. It stays compact enough for practical CPU use while still covering Chinese, English, and a broad multilingual set.
这个页面现在聚焦 MOSS-TTS-Nano 的宣传与试听展示， 不再依赖 Vue 路由或组件系统。下面的样例按中文、英文和多语种拆开， 方便直接听对比，也更适合静态部署到 GitHub Pages。
stereo audio output
48 kHzlanguages shown here
--demo clips on this page
--Why This Page Looks Different
A landing page tuned for MOSS-TTS-Nano, not video understanding.
Visual direction is warmer, lighter, and more audio-centric. The emphasis moves from Q&A and video grids to model positioning, deployment value, and direct side-by-side listening for prompt and generated speech.
Tiny enough for practical deployment
MOSS-TTS-Nano keeps the stack intentionally small so local demos, server-side generation, and embedded product experiments stay lightweight.
Voice cloning with prompt speech
Each sample on this page pairs a prompt clip with synthesized output, making it easy to hear how style, speaker identity, and language transfer behave.
Multilingual from the start
This showcase separates Chinese, English, and additional languages so the multilingual coverage is visible at a glance instead of buried in a single list.
Multilingual Coverage
Chinese and English lead the page, with the rest grouped into a dedicated multilingual gallery.
国旗已经直接体现在样例名里，所以这里也保留了旗帜语言标签，方便一眼看出覆盖范围。
Listen First
Multilingual demo showcase
每张卡片都统一展示 name、prompt speech、
text 和 generated speech。中文和英文单独成栏，
其他语言放在最后一个多语种区块里。
Chinese Showcase
中文样例
包含常规播报、长文本晚安、台湾腔和京味胡同闲聊，比较适合突出 voice cloning 的风格保持能力。
English Showcase
English samples
English cards cover institutional introduction, long-form reading, news-style delivery, and a softer reminder tone.
Multilingual Gallery
更多语种
Japanese, Korean, Spanish, French, German, Italian, Hungarian, Russian, Persian, Arabic, Polish, Portuguese, Czech, Danish, Swedish, Greek, and Turkish are grouped together for quick scanning.