Google just added Maps Imagery Grounding to Google Flow, letting users generate photos and videos anchored to real-world Google Maps Street View data. Instead of conjuring generic scenes, the agent pulls actual street-level details whenever you name a specific location or landmark. It basically gives generative video actual GPS coordinates, though it is currently limited to U.S. locations.
HKUST just open-sourced Domain Shuttle, a subject-driven video generation tool that preserves reference character consistency even across mixed artistic styles. It lets creators combine live-action, 2D anime, and 3D characters in a single scene while transforming styles using text prompts. The code and 35GB model based on Wan 2.2 are available on GitHub, requiring a high-end GPU.
Kuaishou's Kling research team just previewed Omni Director, a camera-control system that extracts full camera motions—from crane shots to cuts and bullet-time effects—directly from a reference clip and applies them to a new image. It moves video generation past clunky text prompts, though code and weights aren't out yet.
Google DeepMind just announced a first-of-its-kind research partnership with indie movie powerhouse A24, embedding its tech directly into filmmakers' creative workflows. Rather than handing creators off-the-shelf models, the deal lets directors and artists guide DeepMind's tool development from the ground up—backed by a fresh investment from Google. Why it matters: It is a rare push toward tech labs building creative AI hand-in-hand with top Hollywood storytellers.
xAI just rolled out Grok Imagine Video 1.5 across grok.com and its API, claiming the top spot on the Image-to-Video Arena leaderboard. Built on the autoregressive Aurora engine, it costs just $0.14 per second at 720p with native synchronized audio—roughly 86% cheaper than Sora 2 Pro. The catch is that output is currently capped at 720p.
Adobe just expanded its Firefly AI assistant to Premiere, Illustrator, InDesign, and Frame.io, offloading grunt work like sorting clips into bins, batch-renaming assets, and fixing missing fonts. Firefly is also adding Elements and Projects to save brand assets across teams, with upcoming support for Google Gemini and Slack.
Ai2 just open-sourced MolmoMotion, a multimodal model that predicts how 3D points on objects will move based on video frames and text prompts. Built on Molmo 2, it boosts simulated robotic pick-and-place success from 56% to 76.3% and steers controllable video generation. Ai2 also released an open dataset of 1.16M video trajectories and a new benchmark.
Palmier just launched an open-source, Mac-native video editor designed specifically for Claude, letting creators edit, organize, and generate footage directly inside the timeline. It includes free on-device semantic search and transcription, letting users bring their own Anthropic API key or connect via MCP. Paid add-ons also unlock third-party models like Kling and ElevenLabs.
Alibaba Cloud just introduced HappyOyster 1.0, an open world model that builds interactive digital environments in real time from text prompts. Unlike video models that make you wait for renders, it continuously responds to user input, allowing people to steer live storylines or walk around generated worlds using WASD keyboard controls.
Researchers just unveiled FlexAvatar, a novel 3D Gaussian Splatting method that reconstructs an animatable 360-degree head from a single photo in only 2 minutes. It disentangles driving signals from viewpoints across mixed datasets, hitting 20 FPS on an RTX 3090 with just 1.7GB of VRAM.
ByteDance just launched Seedance 2.0 Mini on its BytePlus platform, bringing faster video creation from image, video, and audio references. Geared toward daily production pipelines, the lightweight model cuts standard Seedance 2.0 pricing by roughly 50% for budget-conscious creators.
PermaVid just open-sourced a video editing framework that solves one of generative video's biggest headaches: consistency. By splitting context into two separate memory banks for RGB appearance and 3D depth geometry, it keeps scene edits stable across time without distorting structure. Why it matters: It stops AI from forgetting edits, making iterative long-form video editing actually viable.
Meta just previewed an upcoming AI assistant and a desktop version for its video-editing app Edits to challenge CapCut. The assistant taps into creators' Instagram analytics to brainstorm video ideas and recommend trending audio, while desktop support brings larger screens and cross-device sync. A new Beta tab for experimental features and expanded audience insights also roll out today.
Runway just expanded its partnership with Hollywood giant Lionsgate with a joint development program to create original IP together. Instead of just licensing video models as production tools, the AI startup will now co-create content directly with the studio. It marks one of the deepest integrations between generative AI and legacy Hollywood filmmaking yet.
MoVerse just introduced a real-time video world model that turns a single regular image into an interactively navigable 360° world. By decoupling 3D Gaussian world memory from video rendering, it pulls off 8 FPS interactive roaming on a single consumer RTX 4090 GPU. The code and weights are slated for release following compliance review.
ElevenLabs just launched Avatars in ElevenCreative, letting creators turn scripts into lip-synced talking-head videos inside a single interface. Users can create persistent visual identities from photos or prompts, keep characters consistent across scenes, and batch-produce videos through Flows.
Researchers from KAIST AI and AIPARK just unveiled Lip Forcing, the first autoregressive diffusion model for video-to-video lip sync. By distilling a 14B teacher into a 2-step causal student, the 1.3B model hits 31 FPS on a single H100, easily clearing the real-time threshold.
HeyGen just launched HyperFrames, an official MCP connector for Claude and Claude Code that turns text prompts into animated videos. Instead of skimming dense walls of text, users can now generate code-first motion graphics, voiceovers, and subtitles directly inside Claude, exporting ready-to-share MP4s without touching a video editor.
ZAI just open-sourced SCAIL-2, an end-to-end character animation model that transfers complex motion without relying on ambiguous skeleton maps. It handles intense multi-character combat and can even map human dancing onto flamingos or stylized cartoons. The catch: the full weights clock in at 81GB, so consumer GPUs will likely need to wait for quantized builds.
Researchers just unveiled Anchor World, an embodied simulator controlled by actual 3D human motion instead of text prompts or key presses. By feeding in body movements alongside scene anchor views, it renders egocentric video of people navigating spaces and handling tools like sinks or knives. It could give humanoid robots a massive new source of first-person training data.
