toebee

5 ms·

1. 1.▲
   Show HN: Nari Qwen3-TTS and Qwen3-ASR – High accuracy, low latency and cost(narilabs.com)

   90 points by toebee 4d ago|31 comments
2. 2.▲
   Expanding the Pareto Frontier for Realtime Transcription(narilabs.com)

   1 points by toebee 7d ago|1 comments
3. 3.▲
   by toebee 7d ago

   Qwen3-ASR 1.7B is an excellent OSS Speech-to-Text model. But it is slow out of the box. We built an inference engine around it, and now it is the fastest realtime transcription model (as of Sep 2026) at 40 ms time-to-final-segment (TTFS). A
4. 4.▲
   Show HN: 10x Cheaper TTS at 50ms Time-to-First-Audio(narilabs.com)

   1 points by toebee 8d ago|1 comments
5. 5.▲
   by toebee 8d ago

   Hey HN, Toby from Nari Labs here. Our TTS endpoints for Qwen3-TTS 1.7B are live: > "Fast" endpoint: client-side p90 time-to-first-audio (TTFA) at 50ms. > "Standard" endpoint: dirt cheap at $5 per 1M characters, sti
6. 6.▲
   by toebee 26d ago

   will look into this and get it fixed
7. 7.▲
   by toebee 27d ago

   will try to record a video soon, in the meantime you can grab a h100 from somewhere like runpod or modal and test it out: docker run --rm --gpus all \ -p 8000:8000 \ -e HF_TOKEN \ -e QWEN3_TTS_PROFILE=ttfa \ -v nari-qwen3-tts-cache:
8. 8.▲
   by toebee 27d ago

   will try to record something - in the meantime you can spin up a machine on runpod or modal to quickly test it out. docker run --rm --gpus all \ -p 8000:8000 \ -e HF_TOKEN \ -e QWEN3_TTS_PROFILE=ttfa \ -v nari-qwen3-tts-cache:/
9. 9.▲
   by toebee 27d ago

   we haven't tried so can't say for sure, but if optimized for a scenario where the batch size is 1 and max concurrency is 1, it seems possible to get something pretty fast. i'm guessing mobile hardware specific optimizations w
10. 10.▲
    by toebee 27d ago

    we continuously compared output to qwen's original implementation and do not see differences in output quality. let us know if it works well for you!
11. 11.▲
    by toebee 27d ago

    Yep, might need some changes.
12. 12.▲
    by toebee 27d ago

    We do graph capture etc at startup (same as vLLM) but this model variant doesn’t require prefix caching - the prefix is just 10 tokens.
13. 13.▲
    by toebee 27d ago

    Qwen3 TTS has input streaming mode: you can stream LLM output into the speech model. So don’t need to wait for a full sentence. We also implement this websocket variant, and it also runs at sub 50 ms. LLM TTFT is still a big issue, and we m
14. 14.▲
    by toebee 27d ago

    We got a rtx 4090 handling around 10 concurrent requests at 50 ms TTFA after some config changes / adjustment as it doesn’t have FP8. So this 50 ms TTFA thing is very much possible on consumer hardware.
15. 15.▲
    by toebee 27d ago

    Hey thank you for your kind words! Yes, we’ve shifted to inference but will also continue doing finetuning etc. on top of open models. Don’t have plans to do pretraining though.
16. 16.▲
    How we made a text-to-speech model respond in sub-50 ms(nari-labs.com)

    171 points by toebee 28d ago|44 comments
17. 17.▲
    by toebee 28d ago

    time-to-first-audio (TTFA) is critical for realtime voice applications. open source implementations (e.g. vLLM-Omni, SGLang-Omni) are often too slow for production and can have issues with realtime playback if you push for lower latency. we
18. 18.▲
    Show HN: Dia2, open-weights TTS model for realtime speech to speech(github.com)

    3 points by toebee 10mo ago|2 comments
19. 19.▲
    by toebee 1y ago

    You're absolutely right. We used Jordan's Whisper-D, and he was generous enough to offer some guidance along the way. It's also a valid criticism that we haven’t yet audited the dataset for existing list of tags. That’s somet
20. 20.▲
    by toebee 1y ago

    not part of the license!
21. 21.▲
    by toebee 1y ago

    Thank you for the kind words! We only support English at the moment.. Hope to add more languages in the future.
22. 22.▲
    by toebee 1y ago

    Thanks you!! We personally used Quickpod and Runpod the most. But you can try it now on HF Spaces without spinning up GPUs yourself! https://huggingface.co/spaces/nari-labs/Dia-1.6B
23. 23.▲
    by toebee 1y ago

    Thanks for the interest! We also enjoyed using E5-F2 :) You can try it now on HF Spaces: https://huggingface.co/spaces/nari-labs/Dia-1.6B
24. 24.▲
    by toebee 1y ago

    Thank you so much for the kind words :) We only support English at the moment, hopefully can do more languages in the future. We are planning to release a technical report on some of the details, so stay tuned for that!
25. 25.▲
    by toebee 1y ago

    We will try to make it work, but not sure if will be an easy task. For now, you can try with https://huggingface.co/spaces/nari-labs/Dia-1.6B
26. 26.▲
    by toebee 1y ago

    Thank you! You can add audio prompts of calm voices to make them a bit smoother. https://huggingface.co/spaces/nari-labs/Dia-1.6B you can try it here!
27. 27.▲
    by toebee 1y ago

    Thank you!! Works for English only unfortunately :((
28. 28.▲
    by toebee 1y ago

    Thanks for the kind words! We're just following our interests and staying upwind.
29. 29.▲
    by toebee 1y ago

    Thanks for the kind words! You can try it now on https://huggingface.co/spaces/nari-labs/Dia-1.6B Also, we'll try to update the Demo Page to something lighter when we have time. Thanks for the feedback :))
30. 30.▲
    by toebee 1y ago

    Yes! But you would need to put together a LLM system that created scripts from the book content. There is an open source project called OpenNotebookLM ( https://github.com/gabrielchua/open-notebooklm ) that does somethin