Meta has introduced two new artificial intelligence models, Muse Image and Muse Video, marking its latest push to strengthen its generative AI capabilities with tools designed to create and edit images and videos more accurately while integrating deeply across its ecosystem of apps.
The company announced on Tuesday that Muse Image, developed by its Meta Superintelligence Labs, is now available through the Meta AI app and on Meta.ai, as well as in Instagram Stories in the United States and WhatsApp in selected countries. Support for Facebook will be added later, while Muse Video is expected to roll out to creators and Meta AI users in the coming months.
According to Meta, Muse Image is its most advanced image generation model to date, capable of understanding complex prompts, making precise edits, combining multiple reference images and incorporating social context from Instagram to produce more personalised results. The model was also designed to work alongside Muse Spark, enabling the two AI systems to jointly plan and generate multimedia content.
Unlike conventional text-to-image models, Meta said Muse Image operates as an “agentic” AI system that can reason through tasks before generating an image. To improve accuracy, it can search the web for up-to-date information, write and execute code to create elements such as charts and QR codes, and refine its own outputs through multiple iterations.
The company said these capabilities allow the model to generate technically accurate visual content, including scientific illustrations, infographics, interactive websites, animations and image-based games. It added that giving the model access to search tools significantly improves the factual accuracy of images generated from prompts involving current events or real-world information.
Meta also highlighted the model’s image-editing capabilities, saying Muse Image can make targeted modifications without unnecessarily altering other parts of an image. It supports iterative editing, enabling users to refine images over multiple prompts while maintaining consistency across each version.
The model can also combine multiple reference images in a single prompt, allowing users to merge people, objects, clothing, artistic styles and environments into one composition.
Beyond image generation, Meta previewed Muse Video, which is built on the same underlying architecture as Muse Image. The company said the forthcoming model delivers strong prompt adherence, high visual quality and native audio support, while work continues on improving audio-video synchronisation and the realism of fast-moving scenes.
Meta said independent Arena benchmark rankings place Muse Image among the leading AI image generation models for text-to-image creation and image editing, while Muse Video currently ranks among the top-performing text-to-video models based on human preference evaluations.
As part of its responsible AI efforts, the company introduced Content Seal, an invisible watermarking technology embedded into images created with Muse Image through Meta AI. According to Meta, the watermark is designed to remain intact even after images are cropped, resized, compressed or captured via screenshots. The company said it plans to extend the technology to AI-generated videos and will provide a detection tool that enables users to verify whether an image was produced using Meta AI.
Meta said the new models will be integrated across its family of applications, enabling creators, businesses and everyday users to generate marketing materials, redesign photos, create social media content and collaborate on AI-generated visuals directly within its platforms.
Almajiranci at the Crossroads