🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Goodbye LoRA (Part 17) 👋 Diffusion Self-Distillation can generate high-quality images of specific subjects in new settings by preserving identity. Also supports relighting 👌

「ICLR 2025」We introduce a new method for noise warping that reduces the time cost of existing SOTA by 42x and memory cost by 7x, without compromising quality. Our method can “paste” noise textures onto screen-space or 3D objects to enhance temporal or multi-view consistency.

We propose Long Context Tuning (LCT) for scene-level video generation to bridge the gap between current single-shot generation and real-world narrative video productions. Homepage: guoyww.github.io/projects/long-… Report: arxiv.org/abs/2503.10589

Towards scene-level video generation! See our latest work: Long Context Tuning for Video Generation Homepage: guoyww.github.io/projects/long-… Report: arxiv.org/pdf/2503.10589

Long Context Tuning for Video Generation

Introducing Diffusion Self-Distillation (DSD): A new method from Shengqu Cai and Stanford University researchers that fine-tunes text-to-image models to enable "identity-preserving generation" for characters, objects and scenes with a single input image. It works for any style,

Thanks for sharing! Heather Cooper

Check out our latest work CameraCtrl II. By carefully collecting and processing data and introducing as little inductive bias as we can, users are allowed to explore the generated world with appealing dynamics and consistency. Together with extension and distillation, CameraCtrl

To further understand the consistency of generated ones, we visualize the 3D reconstruction results here.

Can we reconstruct relightable human hair appearance from real-world visual observations? We introduce GroomLight, a hybrid inverse rendering method for relightable human hair appearance modeling. syntec-research.github.io/GroomLight/

🏡Building realistic 3D scenes just got smarter! Introducing our #CVPR2025 work, 🔥FirePlace, a framework that enables Multimodal LLMs to automatically generate realistic and geometrically valid placements for objects into complex 3D scenes. How does it work?🧵👇

Introduce CoT-VLA – Visual Chain-of-Thought reasoning for Robot Foundation Models! 🤖 By leveraging next-frame prediction as visual chain-of-thought reasoning, CoT-VLA uses future prediction to guide action generation and unlock large-scale video data for training. #CVPR2025

Excited to share our work: Gaussian Mixture Flow Matching Models (GMFlow) github.com/lakonik/gmflow GMFlow generalizes diffusion models by predicting Gaussian mixture denoising distributions, enabling precise few-step sampling and high-quality generation.

🥳Excited to share our latest work, WildGS-SLAM: Monocular Gaussian Splatting SLAM in Dynamic Environments, accepted to #CVPR2025 🌐 We present a robust monocular RGB SLAM system that uses uncertainty-aware tracking and mapping to handle dynamic scenes.

Glad to share Seaweed-7B, a cost-effective foundation model for video generation. Our tech report highlights the key designs that significantly improve compute efficiency and performance given limited resources, achieving comparable quality against other industry-level models. To

🔔 [SIGGRAPH ’25] Want to redesign your apartment and control the style of every piece of furniture? (virtual try-on for 3D scenes). 🎨 Introducing ReStyle3D, a method that transforms your apartment into the design styles as you want! #stylization #interiordesign

Our new ACM SIGGRAPH paper led by the amazing Liyuan Zhu! We take the "virtual try-on" task, but for 3D and interior designs!

ByteDance presents APT2 Autoregressive Adversarial Post-Training for Real-Time Interactive Video Generation

Seedance 1.0 Exploring the Boundaries of Video Generation Models

(1/n) Time to unify your favorite visual generative models, VLMs, and simulators for controllable visual generation—Introducing a Product of Experts (PoE) framework for inference-time knowledge composition from heterogeneous models.