Researchers just released TeleStyle, an open-source model that restyles both images and videos using reference pictures or text prompts. Built on Qwen-Image-Edit, it pairs curriculum learning with a dedicated video-to-video module to keep content and motion consistent across frames.
Alibaba's Tongyi Lab just open-sourced Z-Image, the undistilled foundation model behind its fast Turbo edition. Built specifically for customization, it fixes the Turbo version's repetitive faces and compositions while adding negative prompt support. It needs 50 steps instead of 28, and ComfyUI already supports it out of the box.
European Space Agency astronomers just deployed a neural network called AnomalyMatch across 35 years of Hubble archives. In just two and a half days, it plowed through nearly 100 million image cutouts to spot over 800 previously undocumented cosmic anomalies, from gravitational lenses to completely unclassifiable objects. Why it matters: AI is turning decades of dormant sky surveys into fresh scientific discoveries without requiring years of manual sorting.
Krea AI just kicked off beta testing for Realtime Edit, rolling out immediate access to the first 10,000 waitlist signups and all Max subscribers. The tool brings live image editing to the canvas, but everyone else will have to stay in line for now.
Google DeepMind just debuted 'Dear Upstairs Neighbors' at the Sundance Film Festival, an animated short co-created with animation veterans. Instead of relying on text prompts, researchers built custom video-to-video workflows and local refinement tools powered by Veo and Imagen. Why it matters: It shows a practical path where professional animators keep precise creative control instead of wrestling with AI randomness.
Tencent just released HunyuanImage 3.0 Instruct, an open-weights image-to-image model that uses chain-of-thought reasoning to plan edit steps and preserve untouched regions before rendering. It handles lighting, perspective shifts, and multi-image merging, though self-hosting its 169 GB weights requires massive VRAM.
Google DeepMind just introduced D4RT, a model that reconstructs and tracks dynamic 3D scenes directly from standard 2D video. By turning 4D perception into parallel spatial queries, it achieves up to 300x higher trajectory throughput than previous methods. Why it matters: it gives robots and AR glasses a path to real-time spatial intuition in moving environments.
Researchers just released RF-DETR segmentation under an Apache 2.0 license, introducing a new real-time vision baseline. It arrives in six model sizes, ranging from a Nano variant clocking 3.4 ms per image (40.3 mAP) to a 2XLarge variant reaching 49.9 mAP at 21.8 ms. Why it matters: Developers get an unconstrained, ultra-fast segmentation backbone built for edge devices and latency-critical vision tasks.
YouTube just announced that creators will soon be able to generate Shorts using their own AI likeness, alongside text-to-game creation tools. To prevent AI slop and identity theft, the platform is pairing the feature with strict likeness-management controls. It's rolling out later this year.
Researchers just introduced VideoMaMa, a model that uses generative video diffusion priors to turn coarse segmentation masks into pixel-accurate alpha mattes. Even with only synthetic training, it handles in-the-wild video effortlessly and powers a massive 50K-video dataset. All models and data are slated for public open-source release.
Researchers just open-sourced Motion 3-to-4, a framework that synthesizes coherent 4D dynamic objects from a single monocular video. By decoupling static 3D shapes from motion latent representations, it recovers per-frame vertex trajectories and can even transfer character motion into entirely new videos.
Lightricks just rolled out an audio-to-video feature that lets creators generate video directly from speech, music, or sound effects. The model understands beats and rhythm to keep visuals in sync with the track, making it easy to generate music clips, b-roll, and digital avatars. Basically, audio is becoming the new visual prompt.
Overworld just dropped Waypoint-1, an interactive video diffusion model you can explore in real time using a keyboard and mouse. Powered by its new WorldEngine inference library, the 2.3B model hits up to 60 FPS on an RTX 5090 with near-zero control lag. Visual consistency and hallucinations are still rough around the edges, but you can already test the demo or run it locally.
Meta just open-sourced ShapeR, a conditional 3D generation framework that reconstructs casual video captures into individual, metrically accurate 3D meshes. By combining SLAM points with rectified flow transformers, it avoids baking an entire scene into a single uneditable blob. Why it matters: Creators and roboticists can now pull individual, real-scale 3D assets straight from phone-recorded video.
Researchers just introduced VIGA, a multimodal agent that turns 2D images into fully editable 3D scene programs inside Blender. By combining interleaved multimodal reasoning with graphics engine feedback, it can assemble scenes from basic geometric primitives or plug in existing 3D tools, beating standard baselines by over 124% on BlenderBench.
Researchers just open-sourced CoDance, a new animation model that can animate multiple subjects from a single, potentially misaligned pose sequence. It breaks rigid spatial constraints and accurately rebinds motion to target characters using text prompts and masks. All code and model weights are now publicly available.
Black Forest Labs just launched FLUX.2 klein, delivering sub-second image generation and editing on consumer GPUs with only 13GB of VRAM. It comes in an Apache 2.0-licensed 4B version and an open-weights 9B model, available via API, local runtimes, and a free web demo.
Black Forest Labs just launched FLUX.2 [klein], a speedy image generation family claiming sub-second inference that beats rival models by over 30%. It arrives in 4B and 9B variants, with the lightweight 4B edition released under a permissive Apache 2.0 license runnable on 8.4GB of VRAM.
Researchers just introduced RigMo, a framework that learns both skeletons and motion directly from raw 3D mesh sequences without human rigging annotations. It handles humans, animals, and even objects like fans with fewer artifacts than previous methods. Why it matters: It turns static or deforming 3D meshes into game- and film-ready animatable assets automatically.
Researchers from Tencent just introduced VerseCrafter, a 4D video world model that turns a single image into fully controllable video. By piping 3D Gaussian paths and custom camera rigs straight from Blender into a frozen Wan2.1 diffusion model, creators can finally dictate exact scene choreography without visual degradation.
