Speech & Audio @ Hugging Face đ€
Joined August 2023
- Tweets283
- Following385
- Followers842
- Likes511
Eustache Le Bihan retweeted
đ€ transformers serve now speaks đ
We just added an OpenAI-compatible `/v1/audio/speech` endpoint.
And we couldn't have picked a better model to launch it
with than Phonon, @GradiumAI brand-new on-device TTS: tiny, fast and multilingual.
Run with:
transformers serve gradium/phonon-40m-en
I told you it was not finished!
Banger release from the goated french speech team, aka @GradiumAI
day0 transformers đ€ support, have fun building with it!
A new frontier for on-device speech.
We're open-sourcing Phonon, a suite of TTS models under 50m parameters, setting a new standard for naturalness and accuracy across languages, with the broadest inference support to date.
Phonon-31m is available under Apache 2.0.
gradium.ai/phononđ§”
find the model weights on their hub org đ€
huggingface.co/gradium
Eustache Le Bihan retweeted
Harness Lens - what harness for your use-case?
Compare 19 (and counting!) harnesses across tools, design, and how opinionated they are.
Learn about how each compacts, their system prompt, and what makes them đ©đđđą.
Impressive tool by @yb2698: huggingface.co/spaces/ybelkaâŠ
Eustache Le Bihan retweeted
added Fish Audioâs Drama-3 to speech-to-speech đ
cloned my voice. now I can ask it to whisper, laugh, sigh or get worried, just by talking.
cheers to @andimarafioti for the demo + sampling tips!
PR + setup: github.com/huggingface/speecâŠ
Eustache Le Bihan retweeted
Fun and crazy napkin math: Meta, NVIDIA, Alibaba, DeepSeek, Moonshot, Mistral and ~30 other labs have shared close to $1B worth of training compute on the Hugging Face Hub
Free to download
End of 2023 it was under $40M
Wikipedia was written in volunteer hours. This AI age commons is written in GPU-hours and data
(it's a floor btw: about half comes from official compute numbers rest is educated guess and it still miss the ~900k community fine-tunes)
Space with data to explore: thomwolf-hub-dollars.static.âŠ
Eustache Le Bihan retweeted
There are 505 "agent traces" datasets on the @huggingface Hub (17.6 GB).
I parsed them to answer: what models, which harnesses, what tasks, how long are sessions?
Real user usage is ~15% of sessions. Here it is broken down:
𧔠đ
Eustache Le Bihan retweeted
Crazy week for open-source AI, right? Now itâs our turn!
Beam, Reflectionâs first open model, is out:
-501B MoE, 23B active
-1M context
-Built for coding, reasoning & agentic workloads
-Apache 2.0
đ§”
đš JUST SPOTTED chilling in the HF Paris sauna
Muse Spark Open landing DAY 0 in đ€ transformers!!
What a week for open source đ
đ€ Made with AI
Today weâre releasing Muse Spark Open, the most capable US open-weight model to date.
Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.
Key Model Specs:
â 816B total / 42B active parameters
â 1M token context window
â Natively multimodal across text, image, video, audio
Eustache Le Bihan retweeted
Jensen Huang le sait : la France est le pays le plus attractif dâEurope pour lâintelligence artificielle, grĂące Ă nos talents, nos infrastructures et nos entreprises qui nous permettent de rester Ă lâavant-garde.
Depuis Choose France, notre coopĂ©ration avec NVIDIA sâaccĂ©lĂšre, des startups aux grands acteurs industriels, en passant par les infrastructures de calcul et la recherche.
Avec une nouvelle Ă©tape aujourdâhui : NVIDIA investit dans des entreprises françaises et sâengage Ă soutenir notre initiative franco-allemande pour dĂ©velopper la recherche ouverte en Europe.
Cela a dâautant plus de valeur que nous partageons et portons la mĂȘme exigence dâune intelligence artificielle ouverte, sĂ»re et au service de notre souverainetĂ©.
Merci cher Jensen pour votre confiance et votre soutien.
Eustache Le Bihan retweeted
Yesterday and today I worked on the new "decision" pipeline to support all `jev`-like models in `transformers`.
This will bring good standards in the mayhem that the ecosystem is now đ
Meaning support for all jev/clef/von/etc via the simple and efficient `transformers serve`
crazy year for OS text-to-speech, and itâs clearly not the end đ
BIG release by Fish Audio: drama-3
- big boy: 35B, 3B active, day0 support in sglang-omni
- would have loved a license allowing distillation though đ
plain-language direction, mid-line shifts, world-level tags
83 languages
features are cool but the moat isnât just the weights. Itâs that reward model (and the human A/B data behind it) + voice library, not released
$52M seed round. 15M+ users. But this is what we're most proud of yet.
Introducing Drama 3: the most emotive and controllable TTS model in the world
And yes, it's open-weight.
You can now control how even a single word should sound, change its tone, and shift your voice mid-sentence.
Uncle Roger was impressed too.
Try it here:
s.fish.audio/rqh0l8
Comment âDRAMAâ within the next 48 hours for 1 month free.
If you're an enterprise your first month is on us:
s.fish.audio/qz9r0t
