Krea AI just introduced prompt-to-workflow in Krea Nodes, allowing creators to build entire node graphs using plain text instructions. Instead of manually dragging and linking blocks, you can now describe your desired pipeline and have the system wire it up instantly.
Alibaba's Qwen team just launched Qwen-Image-2.0, a major upgrade to its image generation model. It brings native 2K resolution, seamless text rendering without glitchy letters, and handles up to 1K-token prompts for slides and posters. A lighter architecture also cuts inference times while unifying generation and editing.
Hyper3D just upgraded its 3D generator with Rodin Gen-2, adding targeted "3D Inpainting" capabilities. Creators can now box-select specific regions of a 3D model and tweak geometry or textures using natural language prompts, without rebuilding the mesh. The production-ready quad assets can be exported directly to STL for 3D printing.
ByteDance just launched Seedream 5.0 Preview, packing live web search, precise photo editing, and logical reasoning into a single image model. It can fetch real-time info like local weather to ground visual details, replicate editing styles from reference images, and render 2K visuals. Best of all, it is currently free to use.
Google just dropped a new series of ads ahead of football's Big Game, putting Gemini front and center. The flagship spot, "New Home," shows a family visualizing interior decor for empty rooms in real time using Nano Banana-powered image editing. Why it matters: Google is moving past technical benchmarks to show mainstream audiences how AI fits into relatable, everyday chores like home design, car repairs, and haircuts.
Researchers just unveiled SkinTokens and TokenRig, an autoregressive framework that automatically generates skeletons and skinning weights for 3D meshes. Beyond human models, it handles complex creatures like birds, crabs, and beetles with up to 133% higher skinning accuracy. Why it matters: it solves one of 3D animation's nastiest bottlenecks, turning static generative assets into ready-to-animate characters.
Kling AI just launched Kling 3.0, an all-in-one multimodal creative engine. The new model can generate realistic 15-second clips with customizable multi-shots, 4K stills, and native multi-character audio across multiple languages and accents. Early access is currently live on the web for Ultra subscribers.
Roblox just launched a beta for 4D generation powered by its Cube foundation model, adding real-time interactivity to 3D objects. Using modular schemas, players can type a prompt to spawn a multi-part car and immediately drive it around. Early tests in community games saw player time jump 64%.
Kling just unveiled 3D Move (3DiMo), an implicit motion control technique that transfers movement from a driving video to a new character while granting full camera control. Unlike prior tools locked to the original driving angle, it keeps 3D consistency through orbit and zoom shots. It is currently detailed in a paper without a public release.
World Labs just revealed that its world model can output persistent 3D scenes that users can actively build on top of. Instead of collapsing after a few moments, these environments allow users to stay and explore for as long as they like, even past the 60-second mark. Why it matters: persistence is the missing bridge needed to turn fleeting generative clips into interactive, durable virtual worlds.
See-Through just made 2D animation way faster by automatically splitting flat anime illustrations into up to 23 distinct, fully inpainted layers. It outputs a layered PSD complete with inferred depth maps, basically turning static art into ready-to-rig assets for Live2D.
Omnimat Zero just open-sourced a video tool that erases objects along with their tricky reflections and ground shadows. It can also isolate foreground subjects into transparent layers for background swaps, making clean video cleanup far less tedious.
Adobe just unlocked unlimited image and video generations for Firefly subscribers. The update includes Adobe's own commercially safe models alongside third-party heavyweights like Google Nano Banana Pro, GPT Image Generation, and Runway Gen-4 Image. Users can generate up to 2K resolution without counting credits, removing the pressure of trial and error before exporting to Photoshop or Premiere.
Tencent just unveiled InteractAvatar, a framework that lets talking digital humans realistically interact with objects around them, like picking up a phone or putting on headphones. Unlike traditional avatars limited to stiff talking-head motions, it supports continuous action prompts with precise timestamps. Why it matters: It turns static talking avatars into truly interactive virtual characters.
Nvidia just introduced DuoGen, a multimodal framework capable of generating coherent, step-by-step interleaved image and text sequences like cooking recipes or robot action plans. It keeps visual consistency across consecutive steps, though it is currently only available as a research paper.
Researchers from Peking University and Google Cloud AI just built PaperBanana, a multi-agent system that turns raw scientific text into publication-ready illustrations. Five specialized agents handle retrieval, planning, styling, rendering, and critique, beating standard baselines on NeurIPS benchmarks and even polishing rough human sketches. Why it matters: It tackles one of the most tedious bottlenecks in scientific research, though the models still occasionally mix up complex node connections.
A new tool called Pencil just brought an infinite design canvas directly into your IDE, powered by Claude Code. It lets you import Figma files, tweak colors and copy visually on the canvas, and sync clean frontend code in real time. Basically, no more painful handoffs between designers and engineers.
Krea AI just launched Realtime Edit to all users, letting creators tweak and transform images with complex instructions on the fly. Rather than waiting around for generation queues, users can now adjust prompts and see visual edits reflected instantaneously.
ByteDance and Alibaba are reportedly prepping next-generation flagship models for release during the February Lunar New Year holiday. ByteDance is lining up Doubao 2.0 alongside major image and video upgrades, while Alibaba is readying Qwen 3.5 with targeted reasoning boosts. With DeepSeek V4 also rumored for mid-February, China's frontier AI race is entering an intense new clash.
robbyant just open-sourced LingBot-World, a real-time interactive world model built on Wan2.2. It generates controllable 480P and 720P environments from text or images at 16 FPS with sub-second latency, while retaining consistent scene memory over minutes. Why it matters: it gives developers a practical open alternative for interactive worlds and robot learning.
