This GitHub repository provides RBench, a benchmark for evaluating robotics video generation, and RoVid-X, a dataset for training models with RGB, depth, and optical flow videos. The authors highlight limitations in existing video models and aim to enhance embodied AI research.

Disney has signed a three-year deal with OpenAI to integrate its characters into the Sora AI video generator. Users will be able to create videos featuring over 200 Disney, Marvel, Pixar, and Star Wars characters, while Disney makes a $1 billion investment in OpenAI. The agreement excludes talent likenesses or voices.

STARFlow and STARFlow-V are open-source models designed for generating high-quality images and videos from text prompts. They combine autoregressive models with normalizing flows to achieve impressive results in both text-to-image and text-to-video tasks. Users can easily set up the models and start generating content with provided scripts and configurations.

Monet AI combines top video, image, and audio generation models into a single platform. Users can easily switch between different AI tools for various creative projects without the hassle of multiple registrations or interfaces. The platform offers a streamlined workflow and cost-effective solutions for creators.

Runway has introduced GWM-1, its first world model, expanding beyond video generation. This set of autoregressive models allows users to create and explore digital environments in real time, useful for game design, virtual reality, and training AI agents. The second model, GWM Robotics, generates synthetic data for robotics training.

Dreamlux AI allows users to create videos from text or images quickly and easily, without any watermarks. The platform features various templates and an intuitive interface, making video creation accessible to everyone.

Runway Gen-4.5 is a cutting-edge video generation model that enhances visual fidelity and control. It excels in creating realistic motion and intricate scenes, maintaining high performance while accommodating various creative styles. The model is now available to creators across all subscription plans.

Saber is a zero-shot framework for reference-to-video generation that relies solely on video-text pairs instead of costly reference image-video-text triplets. It uses masked training with dynamic substitutes to enhance subject integration and generalization across diverse scenarios. The model shows improved performance in generating videos that maintain subject identity while following text prompts.

This article discusses the progression of video generation techniques towards creating comprehensive world models that simulate real-world dynamics. It outlines a four-generation taxonomy, highlighting how each generation enhances capabilities like realism, interaction, planning, and stochasticity. The authors emphasize the importance of integrating physical and mental world models for applications in robotics and AI.

Domer AI allows users to create videos and images quickly by typing descriptions or uploading photos. It offers various models for different quality and speed needs, making it suitable for social media content without the hassle of filming or editing. Users can start with 10 free credits and have full rights to their creations.

Runway has introduced its first world model, GWM-1, which predicts frame-by-frame simulations to understand physics and dynamics. This model aims to enhance video generation and training for robotics and other applications. Alongside, Runway updated its Gen 4.5 video model to include native audio and multi-shot capabilities.

Whisper Thunder uses the Runway Gen-4.5 model to turn your images into 5 or 10-second HD videos. It supports multiple image formats and offers realistic motion and detail. New users get a free trial credit to start creating.

MAGI-1 is an autoregressive video generation model that creates videos by predicting sequences of fixed-length video chunks, achieving high temporal consistency and scalability. It incorporates innovations such as a transformer-based variational autoencoder and a unique denoising algorithm, enabling efficient and controllable video generation from text or images. The model has shown state-of-the-art performance in both instruction following and physical behavior prediction compared to existing models.

A Model Context Protocol (MCP) server is presented, which integrates with OpenAI's Sora 2 API to facilitate video creation and remixing from text prompts. It allows users to generate videos, check job statuses, and manage video files through various compatible clients and transport methods. The setup includes Node.js requirements, configuration instructions, and usage examples for generating and managing videos efficiently.

VaViM and VaVAM introduce a novel approach to autonomous driving using large-scale generative video models. VaViM predicts video frames through autoregressive modeling, while VaVAM generates driving trajectories via imitation learning, showcasing emergent behaviors in complex driving scenarios. The paper analyzes the model's performance, including its strengths and limitations in various driving situations.

Google Photos has introduced the Create tab, enhancing its features for users to creatively transform their images using the new Veo 3 video generation model. Users can turn still photos into dynamic clips, remix images, create collages, generate highlight videos, and produce animations, all from this central hub for creativity.

AvatarFX by Character.AI introduces advanced video generation capabilities, enabling users to create photorealistic videos with expressive movements and audio from pre-existing images. The technology employs flow-based diffusion models and a sophisticated data pipeline to achieve high-quality, diverse video outputs while prioritizing safety measures against misuse. CAI+ subscribers will get early access to these features as they are integrated into the Character.AI platform.

A collection of video generation demos showcasing the capabilities of the Goku model is presented, featuring various imaginative scenes created from original prompts. The demos include animations of diverse subjects, ranging from realistic animals to whimsical scenarios, highlighting the model's versatility in rendering vivid visuals.

Luma AI has launched Ray3, an advanced text-to-video AI model that incorporates built-in reasoning for enhanced cinematic video production. The model allows users to generate high-quality videos by sketching scenes and following detailed instructions, making it a significant upgrade over its predecessor, Ray2. Partnerships with Adobe and Dentsu Digital highlight its potential impact in professional creative workflows.

Researchers at Mandiant have discovered a new malware strain dubbed "UNC6032," which utilizes AI-generated video content to deceive victims. The malware operates primarily through phishing campaigns, leveraging convincing videos to trick users into downloading malicious software. This highlights a growing trend in cyber threats where AI technology is exploited for malicious purposes.

Character.AI, a prominent chatbot platform, has introduced a new feature that allows users to generate videos and share them on social feeds. This innovation aims to enhance user engagement and creativity by integrating video generation capabilities alongside its existing chatbot functionalities.

Tencent has released HunyuanWorld-Voyager, an AI model that generates 3D-consistent video sequences from a single image, allowing users to explore virtual scenes by defining camera paths. While it offers impressive spatial consistency and depth information, it still relies on pattern matching rather than true 3D modeling, limiting its potential for real-time interactive experiences. The model requires significant computing power and has specific licensing restrictions for commercial use.

HuMo showcases a series of video generation methods that create high-quality, text-aligned, and subject-consistent videos from text, images, and audio prompts. The article includes detailed descriptions of various scenes depicted in the demo videos, highlighting the capabilities of the technology in producing immersive visual content.

Test-Time Training (TTT) layers enhance pre-trained Transformers' ability to generate one-minute videos from text narratives, yielding improved coherence and aesthetics compared to existing methods. Despite notable artifacts and limitations in the current implementation, TTT-MLP shows significant advancements in temporal consistency and motion smoothness, particularly when tested on a dataset of Tom and Jerry cartoons. Future work aims to extend this approach to longer videos and more complex storytelling.

Google has announced significant updates to Veo 3 and Veo 3 Fast, including support for vertical format outputs, 1080p HD resolution, and reduced pricing, making video generation more accessible. The new pricing is $0.40 per second for Veo 3 and $0.15 per second for Veo 3 Fast, allowing users to create high-quality videos tailored for mobile and social media. Additionally, integrations with tools like Mosaic and MediaSim demonstrate the potential for innovative multimedia applications using these updates.

OpenAI has introduced Sora 2, an advanced video and audio generation model that offers enhanced realism and controllability, including synchronized dialogue and sound effects. The app emphasizes user creativity over consumption, with features designed to promote well-being and community engagement. Safety measures, especially for teen users, are also a priority.

Amazon has introduced Amazon Nova Reel 1.1, an enhanced video generation model that allows users to create multi-shot videos up to 2 minutes long from text prompts and optional reference images. The update improves video quality and reduces generation latency, making it ideal for marketing and creative projects through Amazon Bedrock. Users can choose between automated and manual modes for greater control over video composition.

Gemini Advanced users can now generate high-resolution videos using the Veo 2 model, which translates text prompts into dynamic video content. This feature, available through Google Labs' Whisk, allows users to create and share engaging videos easily across various platforms, while ensuring safety with embedded digital watermarks. The video generation capability is rolling out to subscribers globally.

Google has enhanced its Veo 3 platform by introducing a feature that allows users to generate videos from images, significantly expanding creative possibilities for content creators. This capability aims to streamline video production processes and boost engagement across various digital platforms.

Wan2.2 is a significant upgrade to large-scale video generative models, introducing innovations like an effective Mixture-of-Experts architecture, cinematic-level aesthetics, and enhanced motion generation capabilities. The model supports both text-to-video and image-to-video generation at high definitions and is optimized for efficiency, making it accessible for both academic and industrial applications. Various tools and integrations are provided for users to implement these models effectively.

Veo 2, Google's advanced video generation model, is now available for developers, enabling the creation of dynamic eight-second videos from text and image prompts. Users can experiment with its features in Google AI Studio and integrate it into applications via the Gemini API, allowing for innovative content creation in various styles and formats.

Midjourney has launched its new V1 video generation model, capable of producing videos up to 21 seconds long. This model allows users to animate AI-generated images and customize the video length and style, competing with other models like Google’s Veo 3 and OpenAI’s Sora. V1 is part of Midjourney's broader strategy to develop interactive 3D simulations by creating a foundation of moving visuals.

OpenAI's new video generation app, Sora, has quickly climbed to the top of Apple's App Store, allowing users to create and remix AI-generated videos. Despite being invite-only and exclusive to iOS, Sora's innovative features and the backing of OpenAI's advanced technology have generated significant interest, though concerns about potential misuse have also been raised.

Google has announced the global rollout of its new Veo 3 video generation model, which enhances the capabilities of creating video content using advanced AI technology. This model aims to improve user experience by automating video production and providing more creative tools for content creators.

Sora 2 is an advanced video-audio generation system that creates realistic soundscapes and characters, enabling users to inject real-world elements into generated environments. The app prioritizes user control and well-being, featuring tools for customization and safety, particularly for teens, while fostering a community-driven creative experience.

Wan-S2V is an advanced AI model designed for generating high-quality videos from static images and audio, particularly suited for film and television. It can create realistic character actions and expressions, synchronize audio with video, and support various professional content creation needs. The model demonstrates superior performance in key metrics compared to other state-of-the-art methods.

Sora Extend allows users to create extended-duration videos using OpenAI's Sora 2 model by intelligently breaking down prompts into coherent segments. It processes each segment sequentially, maintaining visual and thematic continuity, and automatically concatenates the clips into a single seamless video output. This tool enhances the video generation experience beyond the existing 12-second limit.

Vidu is an advanced AI video generator that rapidly transforms text and images into high-quality videos, offering features like Image to Video and Reference to Video for seamless animation creation. Designed for creators and businesses, it enables efficient production of engaging content while ensuring user data security and privacy. Users can enjoy unlimited free video creation in Off-Peak Mode and leverage Vidu's templates for viral video formats.

Veo 3, Google's new video generation model, combines high-fidelity visuals and synchronized audio, enabling developers to create immersive content efficiently. With features like realistic physics and cinematic quality, it supports various applications from 3D animation to in-game video production, and is available through the Gemini API and Google AI Studio for a fee.

Character.AI introduces TalkingMachines, an autoregressive diffusion model that allows real-time video generation driven by audio, enabling characters to interact dynamically. This technology enhances the potential for immersive audiovisual experiences, paving the way for interactive storytelling and character-driven entertainment. The model utilizes advanced techniques to ensure high-quality, synchronized animations based on audio input.

xAI is set to enhance its Grok app with the introduction of a new character, Valentin, and a feature called Imagine that enables infinite image and video generation with sound. These updates aim to attract creative users, particularly women, by offering customizable experiences and a focus on user-generated content. The launch is anticipated to coincide with the release of GPT-5, positioning Grok as a competitive player in the generative AI landscape.