Video Rebirth just released HyperFlow H3 Flash, an open ~14B distillation of MiniMax-H3 that generates synchronized 768p video and audio in 4 to 8 steps. The Q4 GGUF weight drops down to 9.1 GB to run on 24GB RTX GPUs, while 4x Blackwell B200s achieve playback-speed real-time generation. Why it matters: Cutting-edge, synchronized audiovisual models are finally breaking out of hyperscaler clusters and onto local creator hardware.
Meta just rolled out realtime voice and expressive avatars to its Muse app across mobile, web, and Mac. The avatars react with subtle expressions matching your conversational tone, support natural mid-sentence interruptions, and feature hundreds of voice personalities. Currently live for US users, with support for AI Glasses coming next.
Luma AI just revealed its new Ray 2 video model, extending generation limits from 5 seconds to up to a full minute, with outputs ready in as little as 10 seconds. The startup is also bringing its visual models to Amazon Bedrock and tapping AWS custom silicon to slash compute costs.
Runway just halved the hardware footprint for its 57B-parameter Characters model from eight H100s to four while sustaining 24fps video. By moving FP8 quantization and RoPE to cache write-time and fusing communication kernels into matrix multiplications, it brought diffusion steps to 147.4ms. That makes serving live, interactive video avatars far more viable at scale.
Google just launched Gemini Omni 1.2 Flash, boosting native video generation to 15 seconds while slashing pricing by 30% to $0.07 per second. The update adds multi-turn conversational editing so you can tweak scenes without regenerating from scratch, alongside tighter reference consistency and realistic lip-sync. Google account users can now generate three 15-second clips every month in Google Vids for free.
Selendis just launched Archon, a video foundation model that automatically turns raw footage into finished edits in minutes. Built with musician Steve Aoki, it learns a creator's taste from past videos and reference clips to get sharper with every project. Early access is currently open with three months free.
NVIDIA just introduced Cosmos Framework and Cosmos NIM microservices to build and deploy physical AI world models for robotics, AVs, and surgical simulation. It features new Cosmos 3 Nano checkpoints capable of generating synchronized multi-camera video and LiDAR data. With speculative decoding, the NIM containers deliver 2–3x faster inference on H200 GPUs.
Demos circulating on X show an unreleased Claude checkpoint dubbed "Fable 5.5" generating animations without any video or image models. In one test, it autonomously wrote Python code for over two hours to produce a standalone HTML animated film. Anthropic has not confirmed the model exists.
Runway just launched Dev plugins for Claude Code, Codex, and Cursor, connecting coding agents directly to its generative media platform via MCP. Agents can check credit balances, pick models, and trigger video workflows right inside the editor without storing API keys. It basically turns your coding agent into an end-to-end multimedia builder.
Pika just open-sourced Pika H3 Flex, a post-trained model built on MiniMax-H3 that generates videos with synchronized stereo audio in only six sampling steps. It handles 5-to-15-second clips across 11 aspect ratios, accepts up to 12 multimodal references, and runs on a single H100 GPU. Why it matters: professional creators get rapid, high-control iteration without waiting minutes for a render.
Nvidia just launched Open H3 Regenerate 2K, an open-source tool that redraws low-resolution video drafts at 2K and up to 4K in just three diffusion steps. Built primarily for MiniMax H3 but compatible with models like LTX-2, it gives creators full local control over high-res video refinement, though it may slightly alter fine facial details.
NVIDIA just open-sourced MotionBricks, a real-time motion generation model for digital characters and humanoid robots. By feeding it a handful of target poses, it instantly connects the dots to generate smooth transitions and object interactions without retraining. Pretrained models for human skeletons and the Unitree G1 robot are ready to run on GitHub.
DeepMind just unveiled Genie 4, a real-time world simulation model that pairs interactive environments with native sound effects, music, and spoken dialogue. Backed by Gemini models for physical realism and continuous logical reasoning, it lets users steer simulated scenes on the fly using standard keyboard controls or text prompts. It is currently rolling out as a research preview to trusted testers.
Designer AmirMushich just open-sourced a clever technique to create interactive, rotatable web product pages without 3D modeling. The method generates a spinning video from reference images using LTX-2.5 Pro, splits it into about 30 frames, and controls playback via JavaScript based on cursor movement. Why it matters: E-commerce brands can build sleek pseudo-3D product showcases with zero 3D modeling overhead or runtime AI costs.
Researchers just introduced MatAnyone, a video matting model capable of isolating complex elements like individual hair strands and semi-transparent edges. It runs on around 3GB of VRAM and hits real-time 4K matting on a single H200 GPU by selectively allocating compute to tricky regions. Open-source code and ComfyUI node support are planned soon.
ByteDance’s video editor CapCut just announced a ChatGPT plugin that turns raw footage into a rough first cut through simple prompts. Creators can keep refining the timeline in chat—tweaking shot lengths and subtitles—or adjust things manually. It’s set to roll out for testing in the coming months.
SenseTime just open-sourced SenseNova 6.8 Code2Video, a model that builds animations by generating executable code rather than diffusion sampling. Because every element is defined in code, you can tweak text, timing, or trajectories without re-rolling the entire clip. The 27B model weighs 55GB and requires a high-end GPU to run.
ElevenLabs just launched Chimera, a new platform that ties its signature voice cloning and dubbing tech directly into video generation workflows. The tool integrates video generation models like Seedance into the ElevenLabs app, letting creators and enterprises build end-to-end multimedia pipelines in one place.
Researchers just introduced HiP3R, a hierarchical pose graph framework for long-horizon 3D reconstruction. It aligns video chunks first and then refines frame-by-frame camera poses to kill accumulation drift. Why it matters: it plugs directly into existing pipelines without retraining foundation models, turning lengthy video walks into consistent 3D scenes.
Tencent just open-sourced Hy-Nomas, an interactive world generation model that turns natural language and camera trajectories into 720p virtual environments. By generating short chunks with persistent memory, it keeps scenes consistent for over a minute, and a 9GB distilled version runs locally on a single consumer GPU under Apache 2.0. Why it matters: It brings real-time interactive video down from datacenter clusters to individual creators and robotics researchers.
