Meta has officially unveiled its new flagship image generation model — Muse Image, and also demonstrated an early version of the video generator Muse Video for the first time. Both solutions were developed by the Meta Superintelligence Labs division and mark a significant step forward in the field of AI-powered content creation.
Muse Image is now available to users in the US through the Meta AI app, the web version meta.ai, Instagram Stories, and partially in WhatsApp. According to the company, this is the most advanced image generation model to date. The key difference lies in its high accuracy in following prompts, the ability to edit already created images, and scene assembly from multiple references. Special emphasis is placed on integrating the social context of Instagram, allowing the model to generate more relevant and personalized content.
The architecture of Muse Image is built on an agent-based model principle: it can independently search the internet, use tools for writing and executing code, and refine the image during the generation process. This makes it not just a generator, but a full-fledged assistant for creative tasks.
Additionally, Muse Image is integrated with another model — Muse Spark. The combination of these two AIs allows for the creation of animated GIFs, full-fledged websites with embedded images, and even interactive visual games, leveraging both media generation and code writing.
Muse Video: A Preview of the Future
In parallel, Meta showed an early preview of Muse Video. The model is built on the same pre-trained base as Muse Image, ensuring high visual quality and native audio support. The company acknowledges that the final version is still under development: current efforts are primarily focused on synchronizing audio and video, as well as physically accurate rendering of fast movements.
Copyright Protection and Market Position
Meta has implemented an invisible watermark system called Content Seal in Muse Image. According to the company, the mark persists even after cropping, compression, resizing, or taking a screenshot. In the future, it plans to extend this protection to video as well.
According to Arena metrics, Muse Image ranks second in the text-to-image, single-image editing, and multi-image editing categories, trailing only GPT Image 2. This confirms the strong competitiveness of Meta's solution in the generative AI market.
Cryptalist Expert Opinion: The launch of Muse Image and the announcement of Muse Video are not just routine updates, but a strategic move by Meta to capture market share in the creative AI tools segment. Of particular interest is the agent-based architecture of Muse Image, which could become the foundation for future autonomous AI assistants. However, the key challenge is bringing Muse Video to a competitive level, especially in the area of physical motion simulation, where models from OpenAI and Runway currently maintain an advantage.