Ting Chen Liang

564 posts

Ting Chen Liang banner
Ting Chen Liang

Ting Chen Liang

@ting_

unhear the world @MosiAI_Official @Open_MOSS 🦋 ex @novita_labs

Katılım Şubat 2025
150 Takip Edilen554 Takipçiler
Sabitlenmiş Tweet
Ting Chen Liang
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. 👇

English
8
10
111
13.2K
Ting Chen Liang retweetledi
Daniel van Strien
Daniel van Strien@vanstriendaniel·
This week @Open_MOSS released MOSS-Transcribe-Diarize, a 0.9B open model (Apache 2.0) that transcribes, diarizes, and timestamps in a single pass. I used it to make 174 hours of Apollo 11 mission audio searchable by who said what, when. Total cost: $9.46. These are the real NASA tapes from July 1969, hosted by the @internetarchive. All 103 run through one @huggingface Job: a100-large, 3.8h, 47x realtime, @sgl_project serving the model inside the job. 45,355 timestamped speaker segments. Search the mission and hear any moment from the original tape! Space: huggingface.co/spaces/davanst…
Daniel van Strien tweet media
English
6
9
46
6.8K
ALTIC
ALTIC@ALTIC_DEV·
Thanks for making it open source. Beautiful technical report! Would love to collaborate with you to ship this with FluidVoice ( 7K stars, 100K+ downloads and more! ) to push MOSS into 1000s of devices across the globe who love open source! Our DMs are open if you’re interested :))
English
1
0
7
133
MOSI
MOSI@MosiAI_Official·
🤗 MOSS-Transcribe-Diarize-0.9B is now open source on @huggingface. Built with an end-to-end audio-to-structured-transcript paradigm: >0.9B open-source ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword biasing for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 Thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support! 🚀 Github: github.com/OpenMOSS/MOSS-… Huggingface: huggingface.co/spaces/OpenMOS… API: shorturl.at/DWwe3 Live demo: shorturl.at/wRZ3j Technical Report：arxiv.org/abs/2601.01554 HF Space: huggingface.co/spaces/OpenMOS… AtomGit：ai.atomgit.com/OpenMOSS/MOSS-… SGLang-Omni: github.com/sgl-project/sg… vLLM: github.com/vllm-project/v… MLX-audio: github.com/Blaizzy/mlx-au… Discord:discord.gg/SmVQHGffZU
English
27
92
563
130K
Feiteng
Feiteng@FeitengLi·
OpenMOSS 开源 MOSS-Transcribe-Diarize-0.9B，多人长音频转写一次搞定「说了什么、谁说的、何时说的」。 0.9B 参数，128K 上下文，最长约 90 分钟音频一次喂进去，不切片不拼接。 4090 单卡，RTF 0.017，5–10 分钟录音 30 秒转完。 架构是 Whisper-Medium 编码器 + Qwen3-0.6B 风格 decoder，端到端一个自回归任务全包，不走 ASR＋说话人分离＋时间对齐那套级联。 带热词增强，人名/术语/型号能预配。 Apache 2.0，能商用。 模型 🤗 huggingface.co/OpenMOSS-Team/… 代码 github.com/OpenMOSS/MOSS-…
中文
10
22
72
7.3K
Chayenne Zhao
Chayenne Zhao@GenAI_is_real·
huge thanks to the MOSS team for the collaboration on this. MOSS-Transcribe-Diarize is the model that taught us the most about what long-sequence speech inference actually demands. 128k context with multi-speaker diarization in a single generation pass is a workload profile that breaks assumptions baked into most LLM serving systems. it pushed SGLang Omni to its absolute limits and exposed kernel issues we didnt know we had. on our side we built the first long-sequence multi-speaker ASR dataset for CI and got the system to a stable and satisfactory state - but theres still a lot of headroom left on the performance side. this is some of the hardest inference optimization work out there and were actively looking for people to help push it further @MosiAI_Official @sgl_project
MOSI@MosiAI_Official

🤗 MOSS-Transcribe-Diarize-0.9B is now open source on @huggingface. Built with an end-to-end audio-to-structured-transcript paradigm: >0.9B open-source ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword biasing for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 Thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support! 🚀 Github: github.com/OpenMOSS/MOSS-… Huggingface: huggingface.co/spaces/OpenMOS… API: shorturl.at/DWwe3 Live demo: shorturl.at/wRZ3j Technical Report：arxiv.org/abs/2601.01554 HF Space: huggingface.co/spaces/OpenMOS… AtomGit：ai.atomgit.com/OpenMOSS/MOSS-… SGLang-Omni: github.com/sgl-project/sg… vLLM: github.com/vllm-project/v… MLX-audio: github.com/Blaizzy/mlx-au… Discord:discord.gg/SmVQHGffZU

English
3
1
21
2.8K
Prince Canuma
Prince Canuma@Prince_Canuma·
🎉 Congrats to @MosiAI_Official + @Open_MOSS on the release of MOSS-Transcribe-Diarize-0.9B — a genuinely impressive end-to-end ASR model for real-world, multi-speaker conversations. We're proud to partner with them for day-0 support on mlx-audio (Python & Swift). Now running local-first on Apple Silicon: 🎙️ transcript + speaker labels + timestamps in one pass 🗣️ unlimited-speaker diarization ⏱️ up to ~90 min of audio per input 🌍 90+ languages 🧠 128K context • Whisper-Medium encoder + Qwen3-0.6B decoder 🎯 hotword boosting for names, products & domain terms Perfect for meetings, podcasts, interviews, and long-form call analysis — no cloud, no data leaving your Mac. Get started now: 🐍 Python > uv pip install -U mlx-audio 🍎 Swift .package(url: "github.com/Blaizzy/mlx-au…", from: "0.1.3") Go build. 🚀
MOSI@MosiAI_Official

🤗 MOSS-Transcribe-Diarize-0.9B is now open source on @huggingface. Built with an end-to-end audio-to-structured-transcript paradigm: >0.9B open-source ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword biasing for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 Thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support! 🚀 Github: github.com/OpenMOSS/MOSS-… Huggingface: huggingface.co/spaces/OpenMOS… API: shorturl.at/DWwe3 Live demo: shorturl.at/wRZ3j Technical Report：arxiv.org/abs/2601.01554 HF Space: huggingface.co/spaces/OpenMOS… AtomGit：ai.atomgit.com/OpenMOSS/MOSS-… SGLang-Omni: github.com/sgl-project/sg… vLLM: github.com/vllm-project/v… MLX-audio: github.com/Blaizzy/mlx-au… Discord:discord.gg/SmVQHGffZU

English
9
17
86
10.6K
Ting Chen Liang retweetledi
lucas
lucas@olucasandrad·
What a time to be alive
Ting Chen Liang@ting_

🤗 we just opensourced our tiny beast on @huggingface >0.9B ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword customization for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support 💖 try it live👇

English
1
1
3
707
Adina Yakup
Adina Yakup@AdinaYakup·
OpenMOSS has been shipping cool stuff 🔥 Weights: huggingface.co/OpenMOSS-Team/… Demo: huggingface.co/spaces/OpenMOS…
MOSI@MosiAI_Official

🤗 MOSS-Transcribe-Diarize-0.9B is now open source on @huggingface. Built with an end-to-end audio-to-structured-transcript paradigm: >0.9B open-source ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword biasing for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 Thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support! 🚀 Github: github.com/OpenMOSS/MOSS-… Huggingface: huggingface.co/spaces/OpenMOS… API: shorturl.at/DWwe3 Live demo: shorturl.at/wRZ3j Technical Report：arxiv.org/abs/2601.01554 HF Space: huggingface.co/spaces/OpenMOS… AtomGit：ai.atomgit.com/OpenMOSS/MOSS-… SGLang-Omni: github.com/sgl-project/sg… vLLM: github.com/vllm-project/v… MLX-audio: github.com/Blaizzy/mlx-au… Discord:discord.gg/SmVQHGffZU

English
4
5
24
1.7K
Ting Chen Liang retweetledi
Lucas Vieira
Lucas Vieira@luksamuk·
GENTE, OLHEM ISSO. ISSO AQUI RODA DENTRO DO SEU COMPUTADOR.
Ting Chen Liang@ting_

🤗 we just opensourced our tiny beast on @huggingface >0.9B ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword customization for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support 💖 try it live👇

Português
3
7
149
11.7K
MOSI
MOSI@MosiAI_Official·
🤗 MOSS Transcribe series update.@Open_MOSS Alongside MOSS-Transcribe-Diarize-0.9B, we are also updating two models: MOSS-Transcribe-Diarize Pro >Flagship multi-speaker transcription for complex meetings, long audio, multilingual scenarios, and enterprise API use. MOSS-Transcribe >Open-source ASR for challenging English speech, including standard English, diverse accents, quiet speech, and whispers. Built for enterprise meetings, call-center QA, podcast transcription, interview cleanup, and content production.
MOSI tweet media
English
11
20
125
8.9K
Yichi Zhang
Yichi Zhang@YichiZ03·
MOSS-Transcribe-Diarize-0.9B is out — congrats to @MosiAI_Official 🎉 Two things stand out: 90 min of audio in a single pass, and multi-speaker transcription. Both push hard on the serving stack. Day-0 support in SGLang-Omni — run it today on cookbook: sgl-project.github.io/sglang-omni/co…
MOSI@MosiAI_Official

🤗 MOSS-Transcribe-Diarize-0.9B is now open source on @huggingface. Built with an end-to-end audio-to-structured-transcript paradigm: >0.9B open-source ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword biasing for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 Thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support! 🚀 Github: github.com/OpenMOSS/MOSS-… Huggingface: huggingface.co/spaces/OpenMOS… API: shorturl.at/DWwe3 Live demo: shorturl.at/wRZ3j Technical Report：arxiv.org/abs/2601.01554 HF Space: huggingface.co/spaces/OpenMOS… AtomGit：ai.atomgit.com/OpenMOSS/MOSS-… SGLang-Omni: github.com/sgl-project/sg… vLLM: github.com/vllm-project/v… MLX-audio: github.com/Blaizzy/mlx-au… Discord:discord.gg/SmVQHGffZU

English
3
5
28
4.2K
Ting Chen Liang
Ting Chen Liang@ting_·
🤗 we just opensourced our tiny beast on @huggingface >0.9B ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword customization for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support 💖 try it live👇
MOSI@MosiAI_Official

🤗 MOSS-Transcribe-Diarize-0.9B is now open source on @huggingface. Built with an end-to-end audio-to-structured-transcript paradigm: >0.9B open-source ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Multi-speaker diarization for meetings, interruptions, and overlapping voices >Hotword biasing for names, terms, and domain-specific vocabulary >~100 token/s on NVIDIA RTX 4090, RTF ~0.017 Thank you @sgl_project @vllm_project @Prince_Canuma @lllucas for day-0 support! 🚀 Github: github.com/OpenMOSS/MOSS-… Huggingface: huggingface.co/spaces/OpenMOS… API: shorturl.at/DWwe3 Live demo: shorturl.at/wRZ3j Technical Report：arxiv.org/abs/2601.01554 HF Space: huggingface.co/spaces/OpenMOS… AtomGit：ai.atomgit.com/OpenMOSS/MOSS-… SGLang-Omni: github.com/sgl-project/sg… vLLM: github.com/vllm-project/v… MLX-audio: github.com/Blaizzy/mlx-au… Discord:discord.gg/SmVQHGffZU

English
4
21
169
31.5K
Ting Chen Liang retweetledi
Yichi Zhang
Yichi Zhang@YichiZ03·
People may think serving and optimizing TTS (Text to Speech) model is similar to traditional LLM. It isn't. We made Higgs and MOSS-TTS 2–3.4× faster in SGLang-Omni — and most of the wins landed outside the LLM backbone. Delay patterns, D2H stalls, a vocoder tail-latency killer... In this post, we share everything we learned — every optimization, and every trade-off behind it. We hope it proves a valuable resource for developers building in this space. Full breakdown: x.com/YichiZ03/statu… @sgl_project
Yichi Zhang@YichiZ03

x.com/i/article/2074…

English
10
13
89
21.7K

Keşfet

@Open_MOSS @internetarchive @huggingface @sgl_project @ALTIC_DEV @MosiAI_Official @vllm_project @Prince_Canuma