Luma Labs just introduced Uni-1, an AI model that reasons and generates pixels simultaneously. Instead of treating reasoning and visual generation as separate steps, it combines both into a single unified process that Luma calls less artificial and more intelligent.
Pascal Editor just launched as a 100% free and open-source 3D architecture editor running entirely in the browser. Powered by WebGPU and React Three Fiber, it supports real-time editing of walls, slabs, and zones with GPU-accelerated rendering. It brings desktop-grade architectural modeling to the web without the typical enterprise software price tag.
Microsoft just launched MAI-Image-2, pushing its in-house model to #3 on the Arena.ai text-to-image leaderboard. The update focuses on lifelike photorealism, cleaner in-image typography for slides and posters, and richer complex scenes. It is already rolling out across Copilot and Bing Image Creator, giving designers a direct alternative to existing image generators.
Adobe just launched Firefly Custom Models in public beta, allowing creators and brands to train custom AI image generators on their own artwork to keep visual styles and character designs consistent across projects. Models are private by default, and Firefly automatically blocks uploads marked as opted out of AI training via Content Authenticity Initiative credentials.
Researchers just introduced 3DreamBooth, a framework that brings true 3D spatial awareness to subject-driven video generation. Instead of treating subjects like flat 2D images, it uses a 1-frame optimization trick to decouple geometry from motion and pairs with 3Dapter to route multi-view cues. This basically prevents models from hallucinating mismatched angles in novel views.
Character.ai just launched Imagine Gallery on its mobile app, bringing every AI visual generated across your chats into a single organized grid. You can now filter creations by Persona, save favorites, and—with the new Imagine Message feature—tap any character prompt to spin up art on the spot. Plus, c.ai+ subscribers can turn images directly into custom chat backgrounds.
Google Labs just transformed its Stitch tool into an AI-native design canvas designed for "vibe design." Users can now talk directly to the canvas to critique layouts, generate high-fidelity UIs from business goals, and instantly turn screens into interactive flows. It bridges the gap from raw sketches to production-ready code via an MCP server integration.
SegviGen just open-sourced a 3D part segmentation framework built on TRELLIS.2 that splits meshes into individual components via single clicks or 2D guide images. It outperforms previous SOTA models by 40% in interactive segmentation while requiring far less supervision, though local inference demands at least 24GB of VRAM.
Researchers just introduced World Reconstruction from Inconsistent Views, a new method that turns AI-generated videos into consistent, explorable 3D environments. It uses non-rigid geometry alignment to clean up visual glitches and contradictions across frames, working with almost any video model from Wan-2.2 to Genie 3. Basically, it turns off-the-shelf video generators into functional 3D world builders without retraining.
Midjourney just launched alpha testing for its V8 image model, delivering roughly 5x faster generation than V7. The update significantly sharpens prompt-following, text rendering, and image coherence while adding native 2K resolution via a new --hd mode. Relax mode isn't supported yet, but older personalization profiles carry over seamlessly.
Researchers just introduced Seoul World Model (SWM), the first city-scale simulation grounded in a real metropolis. By using RAG to anchor video generation to real street-view photos, it produces stable, multi-kilometer trajectories without visual degradation—and even lets you spawn Godzilla into the skyline via text prompts.
Nvidia just introduced DLSS 5 at GTC, calling it the company's biggest graphics breakthrough since real-time ray tracing in 2018. DLSS 5 introduces a real-time neural rendering model that infuses scenes with photoreal lighting and materials at up to 4K resolution. Why it matters: DLSS is moving beyond upscaling and frame generation into generative AI that turns 3D frames into cinematic visuals. It arrives this fall.
Retime GS just showed how to build smooth, full-body 3D animations from regular 2D video frames. Unlike conventional 4D reconstruction tools that glitch when frames drop, it uses continuous-time 3D representations and improved 4D Gaussian Splatting to seamlessly fill in missing motion.
Canva just launched Magic Layers, a new capability powered by its Canva Design Model that turns flat, static images and AI visuals into fully editable, multi-layered designs. It automatically recognizes text boxes, separates individual elements, and restores layout structure so you can tweak details without reprompting. It's now in public beta across the US, UK, Canada, and Australia.
Researchers just open-sourced ID Laura (ID-LoRA), a unified digital human model that dumps the clunky two-step workflow of cloning audio and then syncing lips. Built on LTX-2, it generates matching video, cloned voice, and realistic ambient sound in a single pass from just an image and reference audio. Running it locally is heavy, demanding at least 24GB of VRAM.
NVIDIA just rolled out major updates for ComfyUI, adding a simplified App View that hides complex node graphs behind a clean prompt box. Creators also get an RTX Video Super Resolution node delivering 30x faster 4K upscaling, alongside NVFP4 precision on RTX 50 Series GPUs that slashes VRAM use by 60%. Why it matters: Running production-grade AI video locally on consumer hardware is becoming dramatically more practical.
UCL scientists just figured out how to reconstruct 10-second video clips solely by reading single-neuron activity in a mouse's visual cortex. By factoring in physical states like pupil dilation and movement, their dynamic neural model produced high-quality reconstructions that get sharper with more tracked neurons. Why it matters: It pushes neural decoding beyond fuzzy fMRI scans toward cell-level movie playback.
Adobe just launched a public beta of AI Assistant in Photoshop across web and mobile, letting you edit photos simply by typing or speaking natural instructions. The update adds AI Markup to sketch precisely where edits happen, while Firefly Image Editor gains Generative Fill tools and access to over 25 third-party models.
Researchers just released GlyphPrinter, an open-source framework designed to stop AI from botching text in generated images. Built on FLUX.1-dev with region-level preference optimization, it accurately renders notoriously difficult Chinese, Japanese, Korean, and Thai glyphs, along with realistic carvings and symbols.
Rednote just open-sourced FireRed-Image-Edit 1.1, a semantic editor that modifies poses and backgrounds while keeping human faces and outfits locked in. It can blend multiple reference photos and even clone specific poster font styles onto new images. The catch is its roughly 60GB footprint, meaning consumer GPUs will need quantized versions to run smoothly.
