🚨 Paper Alert 🚨 ➡️Paper Title: Kontinuous Kontext: Continuous Strength Control for Instruction-based Image Editing 🌟Few pointers from the paper 🎯Instruction-based image editing offers a powerful and intuitive way to manipulate images through natural language. 🎯 Yet,

truly the greatest day ever🎗️

T2I models excel at realism, but true creativity means generating what doesn't exist yet. How do you prompt for something you can't describe? 🎨 We introduce VLM-Guided Adaptive Negative Prompting: inference time method that promotes creative image generation. 1/6

Generative models have made huge progress, but creativity still lags behind. We show that these models already have the capacity for it, we just need to guide them properly!

The initial noise in diffusion models is surprisingly correlated with the final image. Our NoisePrints paper exploits this to provide a lightweight, distortion-free, cryptographically secure watermark for proving authorship of generated images & videos, requiring no model access.

✨ I’ll be presenting our work on depth-aware image editing at #ICCV2025 in Hawaii 🌴 next week! 📅 Oct 22 | 📍 Exhibit Hall I | 🧩 Poster #82 🌍 Project: rishubhpar.github.io/DAEdit/ 🤝 Working on image generation or editing? I’d love to chat at ICCV! Vision and AI Lab, IISc

Can’t wait to join this incredible lineup of speakers at the Personalization in Generative AI Workshop! See you all at #ICCV2025 next week 🙌✨

🚀 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:

Inbar Huberman-Spiegelglas #ICCV2025 Vova Kulikov matan kleiner Happy to share that FlowEdit received the Best Student Paper Award at #ICCV2025 🎉 Huge congrats to the team: Vova Kulikov, matan kleiner, Inbar Huberman-Spiegelglas

Hey #ICCV2025 goers, I’ll be presenting this cool point cloud editing work at 14:45-16:45 in exhibit hall 1 poster #1771. Come say hi!

We present DyPE, a framework for ultra high resolution image generation. DyPE adjusts positional embeddings to evolve dynamically with the spectral progression of diffusion. This lets pre-trained DiTs create images with 16M+ pixels without retraining or extra inference cost. 🧵👇

“What big teeth you have!” said Red.👩‍🦰 “All because my model suffers from semantic leakage,” said the Wolf.🐺 When Text-to-Image models blur boundaries, identities collapse. Meet 𝐃𝐞𝐋𝐞𝐚𝐤𝐞𝐫, a lightweight inference-time fix that mitigates semantic leakage! 👇

Visual Diffusion Models are Geometric Solvers We cast geometry as images: a plain diffusion model denoises into valid solutions. It is simple, general and effective. Shown on Inscribed Square, Steiner Tree, and Maximum Area Polygonization - all classic hard problems.

Generating an image from 1,000 words. Very excited to release Fibo 😃, the first ever open-source model trained exclusively on long, structured captions. Fibo sets a new standard for controllability and disentanglement in image generation [1/6] 🧵

Wow, it opens so many new research directions. Can't wait to play with it

🚀 LTX-2 is now the first AI model to generate 20 seconds of continuous, synchronized audio and video! That’s double the previous 10-second max. With 20s of audio + video, we finally have time to tell much more interesting - and touching - stories. 🎥 Open-sourcing soon! 🧵👇

Wow, really impressive. Looking forward to the open sourcing of the video-audio model! Big congrats to the LTX team! ⭐ Yoav HaCohen

🚀 We're seeking interns for 2026! Join Snap's Creative Vision Team and help advance the frontiers of generative AI across images, videos, 3D, and 4D. We’re looking for PhD students with strong research backgrounds in these areas. Apply here: 👉 snap-research.github.io/cv-call-for-in…