Carbon-X brings DNA and language together: it reads sequences and answers questions about them π§¬
To accompany the release article, I put together Carbon-X Playground: five experiments exploring sequence completion, single-letter changes, regulatory roles, DNA origins and motif finding.
Try curated samples or your own sequences, ask questions and compare answers with references when available.
Introducing Live Subtitles, a Windows tray app that captions anything your PC is playing and isn't butt ugly. live and fully local. It runs NVIDIA's Parakeet-unified 0.6B on your GPU, puts words on screen about 0.36 s after they're spoken, and uses less than one CPU core while it does it.
I built a benchmark for it out of real TV, podcasts, earnings calls and a short film. Parakeet-unified scores 12.1% WER there, close to Nemotron streaming's 11.4%, but with much better punctuation and lower worst-case delay. Once a caption is on screen it never gets rewritten or jumps lines, and the overlay disappears when nobody's talking.
No NVIDIA GPU? It falls back to Nemotron on the CPU, which also covers 30 other languages. It can also write SRT/ASS/VTT files for videos on disk.
Windows only. The GPU engine needs an NVIDIA card with ~3.2 GB of free VRAM. MIT.
Edit: the first version of this post said 13.8% vs 17.0%. The replies caught a scoring bug (numbers weren't normalized), so the benchmark now uses Whisper's English normalizer and the numbers above are re-scored.
π A gamebook you can play in your own words, powered by local Decision Models.
Traditional gamebooks offer explicit branches: βTo enter the forest, turn to page 42.β In Choose your own adventure, you can describe an action, question a character, or ask what you already know before deciding. Clickable choices remain available.
The storyβs branches and consequences are authored. Natural language adds a flexible way to interact with them, with conversations between actions.
I built this project to explore a concrete role for Decision Models in an interactive application:
β’ Kev-4B makes structured decisions: matching intentions to available actions and distinguishing acting from discussing. β’ Ink manages story progression, discoveries, inventory and endings. β’ Qwen3-8B selects dialogue responses from authored content available in the current state.
Both models run locally inside the Space through llama.cpp, using quantized GGUFs. You can inspect the model calls, scores and routing decisions directly in the interface.
The current release includes three adventures in French and English, with a pack format for adding new stories without changing the engine.
Itβs still experimental, and interpretations can go wrong. Iβd love feedback on how it feels to playβand on this use of Decision Models to connect natural language with explicit game logic.
If an agent can build the obvious demo, the obvious demo probably isnβt worth building anymore. For years, turning a research repo into something people could actually try was valuable by itself. That part is becoming automated β and thatβs a good thing.
Which means the interesting work moves elsewhere: finding the weird use case, the right interaction, the unexpected model combination β or simply knowing which paper is worth anyoneβs attention.
The demo used to be the product. Now it needs a point of view.
So far I've been pointing it at Markdown Minimap, an Obsidian plugin that adds a scrollable IDE-style minimap to your notes. This week I've been clearing a backlog of user-reported issues on it, with Claude often handling them end to end.
Introducing Inflect-v2, two exceptionally small, open-weight English TTS models at just 3.9M and 9.3M parameters. Both generate speech multiple times faster than real-time on CPU. Despite their size, Inflect-v2 delivers quality that is competitive with much larger lightweight TTS systems, including KittenTTS, Piper, and Supertonic-3.
CPU, CUDA, PyTorch, and ONNX are supported. Apache 2.0.
β±οΈ Built a small Space for Visual Chronometer / Pulse of Motion.
Upload a video and estimate its Physical FPS: the frame rate implied by visual motion, independent of metadata. Useful to inspect βchronometric hallucinationβ in generated videos: clips that look smooth, but move with the wrong physical time scale.
A few weeks ago, @victor opened the door: coding agents can now ship Hugging Face Spaces autonomously.
I pulled on that thread.
As someone who builds and ships Gradio demos regularly, I didnβt just want to reproduce the loop. I wanted to see what happens when that loop is plugged into the whole Hugging Face stack.
The interesting part is not only that an agent can ship a Space.
Itβs what happens when Space generation becomes a first-class Hugging Face workflow.
Sharing how I built the LongCat-Video-Avatar 1.5 Space (+500k views on X) in one agent session. Gave a coding agent its own AI lab on ZeroGPU, framed the goal, walked away. It designed, deployed, tested against the live API, fixed, shipped.
A live community radio for AI-generated songs, powered by tracks created with ACE-Step.
You can tune in, discover community-made songs in many languages, vote on what sounds good, and mark your real favorites as Bangers.
The more people listen, vote, and create, the better the station gets.
Under the hood, it connects a few Hugging Face pieces together:
Spaces for the live app, HF buckets for community tracks, OAuth for signed-in listeners, server-side streaming with ffmpeg, hourly playlist refreshes, moderation, jingles, and community feedback loops.
Itβs not just a playlist.
Itβs a shared taste experiment: new songs get a shot every hour, and the community helps decide what deserves another spin.
Come listen. Find weird gems. Support the Bangers. Shape the radio.
Great technical guide by Nico Martin on the Hugging Face blog, showing how to use Transformers.js inside a Chrome extension and run ONNX models from the Hub locally with WebGPU inside a Manifest V3 extension.
The interesting part: this is not just a chatbot in a side panel.
The article walks through the architecture behind a browser agent that can read open tabs, query webpages, search history, and highlight elements directly on the page β with models downloaded from the Hugging Face Hub, cached under the extension origin, and executed locally instead of being called through a remote API for every prompt.
A strong blueprint for building local-first web copilots, reading assistants, and AI-powered browsing workflows.
The paper asks a simple but important question: what if the chatbot interface is not just a neutral wrapper around AI models, but part of the problem?
A chatbot can make a system feel more capable, more certain, and more βhumanβ than it really is. That matters, because interfaces shape how we trust, use, and delegate to AI systems.
When everything becomes: ask β answer we can lose sight of the actual workflow: - parameters - alternatives - uncertainty - intermediate steps - failure modes - human control
For creative AI especially β image, video, editing, animation β Iβm not sure βchatβ should always be the default interface.
Sometimes we need a conversation. But often we need a canvas, a timeline, sliders, masks, previews, comparisons, and visible pipelines.
This is also why I find many open ML demos interesting: Spaces, Gradio apps, visual tools, small focused interfaces.
They often explore another direction β not just better assistants, but better tools. π€
Quietly baking Image β Music π΅ v3 β now running on SOTA open-source models. π fffiloni/image-2-music-v3 | Feel free to test it and share feedback.
Want to share my enthusiasm for zai-org/GLM-5.1 here too π₯
I think we have it: our open source Claude Code = GLM-5.1 + Pi (https://pi.dev/) - Built a Three.js racing game to eval and it's extremely impressive. Thoughts:
- One-shot car physics with real drift mechanics (this is hard)
- My fav part: Awesome at self iterating (with no vision!) created 20+ Bun.WebView debugging tools to drive the car programmatically and read game state. Proved a winding bug with vector math without ever seeing the screen
- 531-line racing AI in a single write: 4 personalities, curvature map, racing lines, tactical drifting. Built telemetry tools to compare player vs AI speed curves and data-tuned parameters
- All assets from scratch: 3D models, procedural textures, sky shader, engine sounds, spatial AI audio!
- Can do hard math: proved road normals pointed DOWN via vector cross products, computed track curvature normalized by arc length to tune AI cornering speed
You are going to hear about this model a lot in the next months - open source let's go - and thanks z-aiππ
PASD isnβt recent, but still delivers strong results β worth restoring rather than replacing.
Getting it to run again wasnβt a simple dependency issue. It relied on parts of diffusers that no longer exist, while moving to Gradio 6 forced a much newer HF stack β and I couldnβt modify the original source directly.
Recreating the old environment wasnβt practical. So I patched the downloaded code at runtime before import and made it compatible with todayβs stack.
That ended up being the only approach that held without forking or freezing everything to outdated versions.
If youβve used it before (or are curious), feel free to give it another try.
My TIGER app is now fully working again, with fixes and full compatibility with Gradio 6 π
It lets you: - ποΈ Separate multiple speakers from an audio file - π¬ Extract each speaker directly from a video - π§ Split audio into dialog, music, and sound effects (DnR) - π₯ Apply DnR separation directly on videos
All powered by lightweight TIGER models for fast and efficient speech separation.
