Video generation, understanding, editing and temporal modeling.
LatestTencent Hunyuan Releases Hy-Nomas Technical Report and Open-Sources Hy-Nomas-Bench
Why read thisThe original text provides the technical report, the open-source benchmark, and a comparison video with Genie 4, so readers can judge the reproducibility of this world model.
Google DeepMind has released the world model Genie 4, which can generate interactive worlds in real time and, for the first time, natively generate ambient sound and character dialogue. The model builds on capabilities from the Gemini series; Gemini 3.8 Flash breaks down goals in a loop and checks frames to maintain causal coherence across multi-step events, and Gemini 4 Argon has also been used for joint simulation tasks with Gemini Robotics. The company says it remains an early experimental model: over long runs, object permanence and complex physical interactions still drift, all generated content carries a SynthID watermark, and it is currently available in preview to trusted testers first.
Why read thisThe official disclosure covers three capabilities—native audio, real-time language control, and Gemini-guided reasoning—which can be used to assess where world models stand in robotics and content previsualization.
Why read thisHiggsfield disclosed data including its Series B extension financing and annualized revenue surpassing $1 billion, offering a look at the commercialization progress of AI video companies.
Why read thisKling 4.0 launches on fal and opens its API; the original text lists specific capability changes such as duration, reference materials, and keyframe control.
Why read thisKling 4.0 is now live on idea.ai, generating 30-second scenes from a single prompt while keeping characters consistent, so readers can judge how video generation workflows are changing.
Why read thisThe original text provides comparison data on generation speed and third-party Elo, allowing readers to judge the trade-off between speed and quality for the video model.
Why read thisThe original article lists several specific capability changes following Kling 4.0's launch on Higgsfield, which readers can use to assess how video generation workflows will be affected.
Why read thisThe official Elo score and generation speed of FLUX 3 Max in third-party human preference evaluations are provided, which can be used to judge its position among frontier image-to-video models.
NVIDIA has released Cosmos3-H-Dreams, specialized from the 16-billion-parameter Cosmos3-H-Embodied-Nano for action-conditioned video generation of cholecystectomy. The model is trained with 28 action dimensions, 832 x 480 video, and 10 Hz action alignment, and distillation yields two student checkpoints, four-step and two-step; the two-step version is trained separately, offering faster inference but lower image quality. NVIDIA also provides the FlashDreams streaming runtime and methods for adapting to custom data and controllers, and states that the model is a research simulator, not a validated clinical system.
Why read thisThe original text provides a complete path from distilling the teacher model to four-step and two-step student models, so readers can understand how real-time surgical video generation balances latency and image quality.
Why read thisThe original text gives FLUX 3 Max's generation speed and image-to-video ranking, which can be used to judge the trade-offs of the customized model in video generation.
Why read thisfal has partnered with Black Forest Labs to launch a video model, along with latency and image-quality benchmark positions that can be compared against the speed and cost of existing frontier video models.
Why read thisViggle AI moves the world model from pixel space to 3D space, allowing readers to understand how this technical approach differs from mainstream methods.
Why read thisThe official announcement provides the available entry points and future roadmap for Kling 4.0, allowing readers to judge how video creation workflows will be integrated.
Why read thisGoogle Gemini officially demonstrated how to turn camera footage into a playable synthesizer in Canvas, so readers can learn about its new multimodal interaction use cases.