Meta launches Muse Image to support AI ads and subscriptions

Meta released Muse Image on Tuesday, adding a proprietary AI image-generation model to its consumer apps and advertising products. CNBC market data showed Meta shares up $20.60, or 3.43%, as the company positioned the tool as part of a broader effort to turn AI infrastructure spending into revenue from advertisers, creators and subscribers.

The model is the first image-creation system from Meta Superintelligence Labs, the unit led by Alexandr Wang. It is the group’s second major launch after Muse Spark, a large language model introduced in April that followed Meta’s earlier Llama model family.

Muse Image had previously carried the internal code name Mango, according to CNBC. Meta said consumers will be able to use it free through the Meta AI app and website, WhatsApp direct messages and Instagram Stories.

How Meta plans to charge users

Meta is placing limits around free use. The company said users who reach those limits can either wait for them to reset or pay for a Meta One subscription. Higher-volume users and creators will need one of Meta’s monthly subscription plans, introduced in May, to generate larger numbers of AI images and unlock some features.

The structure gives Meta a direct consumer-payment channel around AI tools, in addition to its core advertising business. CNBC previously reported that Meta had been testing AI subscription services, including a lowest-priced plan at $7.99 a month.

Muse Image is also slated to appear on Facebook and Messenger, and in more parts of Instagram and WhatsApp, later this year, Meta said.

Advertisers get image variants

For advertisers, Muse Image will be built into Meta’s AI-powered Advantage Plus service. That product helps brands create ad materials and automate parts of campaign development.

Meta said it has been working with businesses and advertisers in connection with the release. In a business blog post, the company said Muse Image can apply reasoning to creative work by adjusting elements, changing styles and producing variations from an advertiser’s existing materials. Meta said agencies and advertisers should see image variants powered by the model in the coming weeks.

The commercial mechanism is straightforward: a model can take text or visual prompts, produce alternative images, and support iterative edits inside campaign tools. If effective, that can reduce the time brands spend producing multiple versions of creative assets for different audiences or placements, though Meta’s claims about quality and workflow gains come from the company.

Race with OpenAI and Google

Meta enters a field where OpenAI and Alphabet’s Google have already released competing image-generation models. CNBC reported that Google’s Nano Banana became popular with consumers after its release last fall.

Meta published internal benchmark results showing Muse Image behind OpenAI’s GPT Image 2 model, while ahead of Google’s Nano Banana 2 on tasks including editing single and multiple images. The tests were Meta’s own benchmarks, rather than independent market assessments.

The launch also reduces Meta’s dependence on outside image and video systems. The company has previously used third-party models from Midjourney and Black Forest Labs to support image and video generation in the Meta AI app and website, according to CNBC. Meta said it intends to use Muse Image to lessen reliance on similar outside technology.

The company also plans to release Muse Video, an AI video-generation model, at a later date. In a technical blog post, Meta said that system offers competitive performance on prompt adherence, visual fidelity and temporal consistency.

This story draws on original reporting from CNBC.