Ting Chen Liang

unhear the world @MosiAI_Official
@Open_MOSS 🦋 ex @novita_labs

[Ting Chen Liang @ting_]

🤗 our latest tts model is open source on @huggingface

> 48hz
> 30+ languages supported
> streaming output
> inline pause control

[OpenMOSS @Open_MOSS]

🤗 MOSS-TTS-Local Transformer v1.5 is now open source.

Built with a pure autoregressive Audio Tokenizer + LLM paradigm:

>MOSS-Audio-Tokenizer-v2, 2B params
>Qwen3-4B backbone
>Native 48 kHz stereo audio
>Streaming output with theoretical sub-100 ms TTFT
>Zero-shot voice cloning
>Inline [pause] control
>🇺🇸 🇯🇵 🇰🇷 31 language synthesis
>SGLang-Omni Day0 support 🎉 @sgl_project @lmsysorg

Designed for voice agents, digital humans, game NPCs, audiobooks, and real-time speech generation.

👇

[Ting Chen Liang @ting_]

@haxfenx beautiful

[Psalter @Psalteric]

That man just made speaker diarization a solved problem — with a 0.9B model you can run yourself.

It's called MOSS-Transcribe-Diarize. Drop in any audio or video file, and it returns a full transcript with timestamps and speaker labels. One model, one pass. (watch the demo below 👇)

Every other stack stitches together separate ASR + diarization pipelines — labels drift, speakers get merged, timestamps fall apart. This one does it all end-to-end.

→ Up to 90 minutes of audio in a single inference pass

→ 50+ languages for transcription AND diarization

→ Consistent speaker labels ([S01], [S02]…) with no separate pipeline

→ 7.37 cpCER on the Podcast benchmark — beating Gemini 2.5 Pro (10.23), Doubao (10.54) and ElevenLabs (11.34)

→ Custom hotword prompting for your domain-specific terms

→ Serves via vLLM / SGLang through the standard OpenAI transcription API

→ Ships with a web app that exports JSON, SRT & ASS subtitles — or burns them straight into an MP4

It just took 🏆 1st place at the MLC-SLM Challenge @ INTERSPEECH 2026 — across 14 languages. And it's only 0.9B parameters. 92K downloads last month and most people still haven't heard of it.

Meetings, podcasts, interviews, lectures — who said what, and when, from a single file.

100% open source

[Ting Chen Liang @ting_]

@Psalteric thank you for sharing our model!!

[Ting Chen Liang @ting_]

@alisaqqt @atlas_cloud_ai this is goood

[Ting Chen Liang @ting_]

@YangGao07 🔥

[Ting Chen Liang @ting_]

@victormustar dope

[narise @xxxxWxxxxxQ]

こんな感じの効果音作れた

[Ting Chen Liang @ting_]

@xxxxWxxxxxQ 🤩 omg thank you for testing our SFX model!! btw love your video as well ahaha, may I ask how did you make the soundwave animations 🤣?

[Ting Chen Liang @ting_]

@alisaqqt @MosiAI_Official @Open_MOSS 情人眼里出西施

[Ting Chen Liang @ting_]

come visit us at WAIC 2026 @MosiAI_Official @Open_MOSS

special gifts & free token giveaway🎁

come get it 👇

📍世博展览馆 H1-C1234

[Ting Chen Liang @ting_]

@AdinaYakup Hiii Adina, if you have time could swing by to MOSI/Openmoss booth, we'd love to meet you!

[Ting Chen Liang @ting_]

@victormustar 😂😂😂

[Ting Chen Liang @ting_]

@Prince_Canuma coding at 3??🤣

[MOSI @MosiAI_Official]

We're at WAIC 2026!

@MosiAI_Official & @Open_MOSS is bringing MOSS Context Intelligence Park to Shanghai!

📍世博展览馆 Booth H1-C1234｜📅 July 17–20

Experience how AI listens, sees, understands context, and creates in the real world:

🔊 Voice Babel Tower - light up a 5m AI voice installation with your own voice

🎙️ Mossland Creator Booth - immersive AIGC creation with AI voice

❓ VBTI - discover your voice personality

🤖 Live demos of MOSS-TTS, MOSS-Transcribe-Diarize & MOSS-VL-Realtime

Come meet us at WAIC 2026! 🌏 #waic2026

[Ting Chen Liang @ting_]

@ivanfioravanti @HarakiriInu @MiniMax_AI @Zai_org @MosiAI_Official @Open_MOSS @MarcoIannello thank you so much Ivan!😍

[Ting Chen Liang @ting_]

@HarakiriInu @ivanfioravanti @MiniMax_AI @Zai_org @MosiAI_Official @Open_MOSS sounds cool! would be awesome to see what you did with our model! ill DM you

[Daniel van Strien @vanstriendaniel]

Last week: 175 hours of Apollo 11 mission audio, transcribed + diarized with an open model for $9.46 (@Open_MOSS)

This week: semantic search over all 45,355 utterances via one more @huggingface Job, 261 seconds, $0.03.

"trouble with the radio" now finds "your transmission is breaking up" and "My antenna's out" — words the crew never says.

[Ting Chen Liang @ting_]

@mervenoyann good testing angle, will try the same with our models 🤣

[Ting Chen Liang @ting_]

@Presidentlin 😂 mostalgic