He Zhang

Several new goodies dropped in Photoshop today, including the new Reference Image with Object Stitch! 🙌

Congratulations Boss!

Honored to be named an IEEE Fellow for contributions to image processing, computer vision & biometrics. Also grateful to be an AAAI Senior Member and a 2025 Clarivate Highly Cited Researcher. Huge thanks to my students, mentors & collaborators! @jhuclsp @HopkinsEngineer

Please send email to hezhan@adobe.com

Adobe Research Intern Hiring for Summer 2026: If you are interested in working with me on a research project in Unified Model for understanding and generation. Feel free to contact me. #adoberesearch #computervision #CVPR

@purnomo_wakhid @drtypography Latest beta will have it

@drtypography what version of psd? love it😍

And the 2nd one: DIVE: Taming DINO for Subject-Driven Video Editing, which leverages powerful DINO features to achieve consistent, identity-preserving video edits guided by text or reference images.

The first one is DiffusionGS, a single-stage diffusion model that bridges 2D generation and 3D Gaussian splatting for fast, high-quality 3D scene reconstruction from just one image. caiyuanhao1998.github.io/project/Diffus…

I will be at #ICCV2025 presenting two papers in the main conference! 🚀 Details are in the thread below. In the meantime, we’re hiring interns and full-time research scientists passionate about generative modeling — feel free to reach out if you’d like to join us or chat at ICCV!

Nice work ! Glad to see more research on treating image editing as a video generation task like we did with @chenxi36824648 on UniReal (CVPR'25) arxiv.org/abs/2412.07774

@WenhuChen Nice work! Congratulations!

What’s preventing us from training open-source image editing models like Nano-Banana or Seedream? The main barrier is the lack of high-quality training data for image editing. Most existing image editing datasets are synthesized using weak reward models or poor quality filters—for example, by prompting GPT-4o or other VLMs. To address this problem, we built the most powerful image editing reward model available. We first curated a large-scale image editing preference dataset and then trained EditReward on top of it. Our best EditReward model, trained from Mimo-7B, achieves the highest agreement with human experts. We also applied EditReward to filter existing noisy datasets and demonstrated significant improvements. Paper: arxiv.org/abs/2509.26346 Website: tiger-ai-lab.github.io/EditReward/ Code: github.com/TIGER-AI-Lab/E… HF: huggingface.co/collections/TI…

First unveiled as early research at #AdobeMAX last year, @Photoshop’s new Harmonize feature can turn hours of editing into minutes! Kudos to the @AdobeResearch and Photoshop Engineering teams! adobe.ly/4lxiXXs

It seems that people like our feature @mengweir

@Densdesign8 @Photoshop @Adobe @creativecloud @AdobeDesign Hi, thanks for testing. I am just curious what kinds of issues you think it is more eager to solve? Like more high quality details on the face or?

Tried Photoshop’s new “Harmonize” feature Honestly, the results look good. There are still some issues, but let’s not forget it’s in beta.

Photoshop’s new harmonize feature looks genuinely useful — effectively making complex compositing tasks just one click. Seems Adobe has productized Project Perfect Blend from their sneaks presentation.

@jnack @paultrani It is in public beta now, John!

@paultrani Looks insane, and I can't wait to try it! Any idea why the Pre-release section of my Mac CC app doesn't list PS, just Illustrator?

#Photoshop beta new feature! Seamlessly composite images with Harmonize!

Huge thanks to our incredible collaborators: Qing Liu, Zhixin Shu JaeShin Yoon, Kangning Liu Zijun Wei, Xingtai Li— and many more who contributed across PM, UI, Engineering, and Service and etc. This launch wouldn’t have been possible without our teamwork.

It’s finally happening — after years of hard work, our project is going live in production! Not just a demo from @mengweir presented last year, but the real thing that everyone can play with.

Mengwei Ren and I have been pushing hard on the final details, and we’re so proud to see this become a reality. Of course, this has never been a two-researchers effort.