Jun-Yan Zhu

@junyanz89

Building the Frontier of Multimodal Inference @NunchuxAI, Associate Professor at the Generative Intelligence Lab @CarnegieMellon.

Pittsburgh, PA

nunchux.ai

Replies Reposts Media

Pinned
Jun-Yan Zhu
@junyanz89
Sep 3
Nunchux AI (nunchux.ai) comes out of stealth today. Ten years ago at Berkeley, we spent whole nights on a Titan X in Theano, trying to make iGAN, pix2pix, and CycleGAN real-time so creators could see the results the moment they asked for them. Image models now
Nunchux AI
@NunchuxAI
Sep 3
Nunchux is live at nunchux.ai! We’re building the frontier of multimodal generative AI inference: fast, affordable, high-quality serving for image, video, and world models. Our Modelverse brings 30+ image, video, and avatar models behind one API. Use
30
Jun-Yan Zhu
@junyanz89
Sep 9
Congrats @nupurkmr9! We'll miss you. Nupur wrote the first Ph.D. thesis on customizing generative image models. Check out her thesis: publications.ri.cmu.edu/customizing-ge… Looking forward to what your team builds next.
Nupur Kumari
@nupurkmr9
Sep 9
Life update: defended my PhD earlier this year 🎉Grateful to everyone who made it possible, especially @junyanz89 for being my advisor! Joined @reve and now at @OpenAI 😁 Excited to work on the next frontier of multimodal models and image generation!
2
Jun-Yan Zhu
@junyanz89
Aug 20
Try out Sheng-Yu's latest work with pip install tpips. TPIPS is a text-prompted image perceptual similarity metric and a revisit of the classic LPIPS metric.
Sheng-Yu Wang
@ShengYuWang6
Aug 18
How similar are two images? Prior metrics (e.g., LPIPS, DreamSim) give just a single score. But actually, there are multiple *senses* of similarity (color, pose, etc.) We introduce TPIPS -- Text-Prompted Image Perceptual Similarity "pip install tpips" peterwang512.github.io/TPIPS 🧵
00:00
Jun-Yan Zhu
@junyanz89
Oct 18, 2025
Check out our new unpaired learning method for instruction-following image editing models.
Nupur Kumari
@nupurkmr9
Oct 17, 2025
🚀 New preprint! We present NP-Edit, a framework for training an image editing diffusion model without paired supervision. We use differentiable feedback from Vision-Language Models (VLMs) combined with distribution-matching loss (DMD) to learn editing directly. webpage:
GIF
2
Jun-Yan Zhu
@junyanz89
May 9, 2025
We've released the code for LegoGPT. This autoregressive model generates physically stable and buildable designs from text prompts, by integrating physics laws and assembly constraints into LLM training and inference. This work is led by PhD students @AvaLovelace0, @kangle_deng,
00:00
9