New models and features in AI image, video, voice and music generation, from Midjourney, Sora, Veo, Suno and more.
Runway just dropped GWM Worlds 2, a real-time world model that continuously generates 720p video at 24 fps with synced audio. Users can define scene physics and characters with WorldPrompt, then steer actions and camera movement on the fly with keybinds or text prompts. Why it matters: It shifts AI video from static clips into interactive simulation for games and robotics, though fast camera turns still cause visual artifacts.
Fal just launched fal.live, an interactive AI livestreaming platform where viewers shape what happens in real time. Users pick a channel and type text prompts to steer the video stream as it generates. It shifts audiences from passive viewers to active directors in endless generative broadcasts.
Hao AI Lab and MiniMax just released FastH3 v1, an open-weight video generation model that creates a 15-second 768p video in only 13 seconds. Clocking up to a 14x speedup on NVIDIA Blackwell GPUs with a single flexible checkpoint, it brings production-grade speed to the open-source community. Full weights and inference code are out now.
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.
Google just released Gemini Omni 1.1 Flash, making its generative video model production-ready for developers. The update can analyze 10 seconds of prior context to extend clips up to 40 seconds, supports first-and-last frame interpolation, and adds a 360p draft mode that renders 60% faster at one-third the cost alongside 4K upscaling.
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.
Runway just launched Ruby, a new model that converts standard SDR video into up to 16-bit HDR, exporting in ProRes and EXR sequence formats. It works with any uploaded clip or previously generated Runway output up to 30 seconds long. It gives filmmakers an instant HDR upgrade without the headache of reshooting.
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.
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.
Stability AI just launched beta tools for Stable Audio 3.0, including a native DAW plugin and an overhauled web studio. The macOS AU/VST3 plugin syncs directly to your session BPM in Logic Pro and Ableton Live to generate up to six minutes of audio, while the web app adds multi-track mixing. Outputs are commercially safe and user-owned.
Alibaba just rolled out HappyShrimp 1.0 into beta, an AI music generation model that crafts complete tracks—melody, arrangement, lyrics, and vocals—from a single text description. It maps out musical structure and emotional tone across genres like pop, rock, and Chinese style, with China's Taihe Music already on board for artist co-creation.
Pika just launched Pika Audio, a family of four foundation models covering video soundtracks, music, sound effects, and text-to-speech. The company claims its training and inference stack makes them up to 20x cheaper than rival models, with speech generation 9x cheaper than ElevenLabs v3. They are available exclusively through the Pika API Club.
Suno just launched Studio 2.0, upgrading its browser-based workstation with MIDI editing, automation curves, and advanced stem separation. Musicians can now use a chat interface to design custom audio plugins from scratch or use MIDI inputs as generation prompts. The update is currently available to Premier subscribers.
MiniMax just launched Music 3.0, an open-weights AI model that generates complete songs up to five minutes long in a single pass. Powered by a hybrid architecture based on Qwen3.5-8B alongside flow matching, it preserves long-range song structure while capturing realistic instrument dynamics. Why it matters: It brings fully produced, extended AI tracks into the open-source fold without sacrificing acoustic fidelity.
Microsoft just launched MAI-Image-2.6, debuting at number two on the Arena text-to-image leaderboard ahead of models from Google, Meta, and xAI. The new release gains +79 Elo overall over MAI-Image-2.5, with text rendering alone jumping by +91 Elo alongside noticeably cleaner portraits and commercial graphics. It puts Microsoft firmly in the top tier of frontier image generation.
Tencent just introduced WorldClaw, an agentic system that turns a single text prompt into an explorable, editable 3D open world. It maps large-scale terrain, populates regions with textured meshes, and deploys render agents to tweak contacts and object poses—all while keeping every single asset reusable.
A mysterious "mona-lisa-1" model just surfaced on LM Arena, fueling speculation that OpenAI is testing a new GPT image generation model. Community reports note the outputs contain SynthID watermarks and visibly outperform Image 2, though OpenAI has yet to officially confirm the model.
Alibaba Cloud just launched Wan3.0, extending single-pass video generation to 30 seconds. For the first time, the model can ingest office files like PDFs, slides, and spreadsheets alongside text and images to produce narrative videos. It also sharpens facial fidelity and adds built-in video editing, with API pricing starting at $0.05 per second.
xAI just rolled out Imagine Image 2.0 as the new Quality Mode across grok.com and its mobile apps. Built for practical design work, the model brings precise local tweaks via a magic wand tool, background removal, and multi-reference support for up to 5 images. It also jumped to second place globally on Arena's text-to-image and editing leaderboards.
