Researchers just open-sourced UniMate, a foundation model that animates virtually any rigged 3D asset from a single text prompt. Unlike previous animators locked to human rigs, its topology-aware diffusion transformer drives birds, serpents, and mechanical objects zero-shot with no per-skeleton retraining. The full codebase, training pipeline, and weights are publicly available.
Google just connected Google Photos directly into Gemini’s Spark power agent for US subscribers. Users can now run complex automations across their photo library with single prompts, such as batch-erasing photobombers, turning event flyers into calendar invites, or auto-generating weekly family recap albums. It turns your photo archive into an active assistant that actually does chores for you.
inclusionAI just open-sourced LLaDA-Image, a 6B unified diffusion model family that handles both text-to-image and instruction-guided editing from a single checkpoint. It delivers photorealistic rendering alongside bilingual Chinese-English text, while a 4-step distilled Turbo version enables near-instant generation. Checkpoints and ComfyUI integrations are already live.
Google just launched Pics, a new AI image creation and editing tool powered by its Nano Banana model. Users can isolate objects for targeted tweaks, edit or translate in-image text without breaking fonts, and collaborate in real time. It is live now at pics.new and baked directly into Google Docs and Slides, with Drive support rolling out soon.
Shanghai AI Laboratory just open-sourced InternLumina-U2, a 16.9B diffusion MoE model with only ~1.4B active parameters. It combines text QA, image generation, editing, video, and 3D understanding into a single framework using an 8-codebook discrete visual tokenizer.
World Labs just unveiled Atlas, a multimodal world model built for spatial intelligence. It combines text, video, and 3D data into a shared spatial context to generate up to 1-minute 1440p clips with pixel-perfect camera control, while reconstructing spaces into 3D Gaussian splats from just a few photos.
ByteDance researchers just introduced Lucida, a pipeline that turns indoor video captures into interactive 3D simulations. Instead of outputting a single frozen mesh, it builds clean 3D assets for each detected object and uses a vision-language agent to place them precisely back into position. Why it matters: It gives robotics and embodied AI fully manipulable digital twins of real-world spaces.
Viggle just open-sourced Viggle-Animate, an AI pipeline built on MiniMax-H3 that ditches traditional pose skeletons, masks, and face trackers. You simply edit one frame of a video to introduce the new character, and the model propagates it across the entire shot in three forward passes, rendering 124 frames in 26 seconds on a single B200 GPU.
Midjourney just launched its first V8.2 image edit model for community testing. The model introduces instruction-based edits, inpainting, canvas expansion, and support for up to four image references at once to replace omni-reference. Why it matters: it brings precision editing and complex style blending directly into standard workflows without messy workarounds.
Adobe just rolled out a beta "AI Assisted Editor" for Photoshop, grouping its generative tools into a single streamlined toolbar. The biggest upgrade is a new markup tool that lets you draw arrows or brush rough shapes directly on an image to guide edits without typing a prompt, backed by Firefly Image 5 understanding full scene context.
NVIDIA just rolled out DLSS 4.5 Ray Reconstruction, powered by a second-generation transformer model to dramatically sharpen ray-traced and path-traced graphics. The neural rendering update is already live across 30 games via the NVIDIA app, cutting down ghosting artifacts. NVIDIA also previewed RTX Mega Geometry, bringing full-detail Nanite ray tracing to upcoming titles.
Researchers just open-sourced FixAnything, an AI tool that repairs visual glitches across 3D representations. When novel views in 3D Gaussian Splatting, NeRF, or meshes suffer from holes and floaters, it leverages a fine-tuned Wan2.1 video diffusion model to turn degraded walkthroughs into clean, 3D-consistent clips.
Bria AI just released Fibo 1.5, an 8B open-source image model built around structured JSON captions rather than messy text prompts. Thanks to two-stage distillation, it runs in just 4 to 6 steps without CFG, allowing artists to tweak lighting or camera angles without breaking the rest of the scene.
Midjourney just rolled out a massive patch for its alpha web interface, fixing dozens of headaches two weeks after its initial overhaul. The update brings collapsible project folders to the sidebar, restores missing upscale and vary buttons for V8.2 images, and wipes out typing lag in the prompt bar.
OpenAI just rolled out transparent background generation for its GPT-Image-2 API in preview. Developers can now generate reusable visual assets directly without manual cutouts, making it easy to drop generated objects straight into product shots, UI mockups, and ad layouts. Why it matters: It cuts out a tedious compositing step and makes gen-AI images instantly production-ready.
4DAnyone just dropped as an open-source framework that turns a single monocular video of someone into an interactive 4D Gaussian Splatting model. By combining skeleton tracking with multi-view diffusion, it eliminates drift and maintains clean consistency across dozens of camera angles—and the entire 12GB model is free to run locally.
Tripo just dropped P2.0 Preview, a 3D asset generator built specifically for games and real-time pipelines. It features native quad topology supporting up to 50,000 triangles or 25,000 quads, delivering clean edge loops around joints and smooth flat surfaces. Every user gets two free generations to test it out.
Researchers just introduced GeoWeaver, a framework that reconstructs complete and consistent 3D scenes from standard long videos. By breaking footage into chunks and stitching them with long-range geometric matching, it stops camera drift and sets the lowest average error on benchmarks. The project is currently available as a research paper.
ByteDance just struck the first-ever AI copyright agreement with Hollywood's Motion Picture Association, agreeing to tighten guardrails on its Seedance video and Seedream image tools. The pact follows cease-and-desist threats over unlicensed SpongeBob and Stranger Things generations, though specific guardrail terms remain confidential.
ByteDance just open-sourced Bernini 2, an omnimodal video pipeline that lets you edit existing footage using natural language, from swapping backgrounds to inserting new characters. It pairs an MLLM planner with a diffusion renderer for fine-grained control, but weighing in at 180GB for the model alone, running it locally will be a heavy lift.
