Meta $META introduced Muse Image on Tuesday, the company's first image generation model built in-house, making it available to consumers through the Meta AI app and to advertisers through its Advantage+ ad platform.
The model, developed by Meta Superintelligence Labs, powers image creation across Meta AI and a set of new features within Instagram Stories and WhatsApp direct messages, including more than 30 new AI-powered effects, the company said. Muse Image will expand to Facebook and Messenger later in the year.
Muse Image is free for general use. Consumers who want higher usage limits can access the model through one of Meta's subscription plans, which debuted in May, the company said. In the coming weeks, advertisers and agencies will be able to use Muse Image through Advantage+, Meta's AI-powered ad tool, to generate and vary image-based ad creative.
The model was originally codenamed Mango, according to CNBC. Alexandr Wang, who heads Meta Superintelligence Labs, previously oversaw the April debut of Muse Spark — a large language model that replaced the company's Llama family — making Muse Image the lab's second significant model launch.
Among Muse Image's features, users can edit photos by annotating directly on an image, tag public Instagram accounts to incorporate their photos into generated images, and snap a photo of a room to generate a redesigned version using real products from the web or Facebook Marketplace, the company said. According to Axios, the tagging feature is on by default, though Instagram users have the option to opt out.
Meta released internal benchmark data placing Muse Image below OpenAI's GPT Image 2 in overall performance while putting it ahead of Google $GOOGL's Nano Banana 2 when tested on image-editing scenarios involving one or more photos. Image generation inside Meta AI has until now been handled by outside vendors, including Midjourney and Black Forest Labs; with Muse Image available, the company intends to pull that work in-house, according to CNBC.
Meta also said a video generation model called Muse Video is in development, though no release date was given.
The launch reflects Meta's broader effort to bring AI capabilities in-house. Meta lost access to a portion of its contracted Gemini capacity from Google earlier this year due to compute constraints, which set back some internal AI projects and prompted the company to ask employees to limit AI token usage. Muse Spark has since taken on more of the work previously handled by Gemini.
Meta is also exploring a cloud infrastructure venture that would sell outside developers access to its AI models and computing capacity, putting it in potential competition with Amazon $AMZN Web Services, Microsoft $MSFT Azure, and Google Cloud.