Meta Ships Its Own AI Image Generator, and It Wants Advertisers to Pay for It

Meta released a new AI image-generation model called Muse Image on Tuesday, and while consumers get to play with it for free, the real target audience is advertisers and paying subscribers.

The model, once known internally by the codename Mango, is the second big release from Meta Superintelligence Labs, the AI unit run by Alexandr Wang. The first was Muse Spark, a large language model that debuted in April and effectively replaced Meta's Llama family as the company's flagship AI brain. Muse Image is the visual counterpart, and Meta is already teasing a third piece, a video-generation model called Muse Video, for later release.

For everyday users, Muse Image will show up in familiar places: the Meta AI app and website, WhatsApp direct messages, and Instagram Stories. Anyone can generate images for free up to a point. But if you want to create a high volume of images or unlock certain premium features, Meta is steering you toward the monthly subscription plans it introduced in May under the Meta One name. Hit your free limit, and the choice is simple — pay up or wait for the limit to reset.

That subscription push is only half the strategy. The other half is advertising. Muse Image will power new tools inside Meta's Advantage Plus system, the AI-driven suite that helps brands build and automate ad campaigns. Meta says it has already been testing the model with businesses and advertisers ahead of this launch.

Built for the ad machine

The pitch to marketers is speed and consistency. According to a Meta blog post aimed at businesses, Muse Image applies what the company calls "native reasoning" to the creative process, letting advertisers adjust elements, swap visual styles, and generate variations based on existing ad creative. The goal is fewer rounds of back-and-forth revisions to get an on-brand result. Meta says advertisers and agencies should start seeing image variants powered by Muse Image within the coming weeks.

That framing matters because it shows exactly what Meta wants out of this technology. This isn't just a toy for generating funny pictures in a chat app — it's infrastructure meant to plug directly into the company's core business of selling ads, and to eventually justify the enormous sums Meta has poured into AI computing capacity. A tool that can spin out multiple polished ad variants automatically, with minimal designer input, is a tangible way to make that infrastructure pay for itself.

Playing catch-up on image quality

Meta is not first to this race, and it isn't pretending otherwise. OpenAI and Google both beat Meta to market with image-generation models that found real audiences — Google's Nano Banana, released last fall, became a genuine consumer hit.

Meta published its own internal benchmark comparisons alongside the launch, and the results are mixed. Muse Image trails OpenAI's newer GPT Image 2 model overall, according to the company's own testing. But it reportedly outperforms Google's Nano Banana 2 on specific tasks, including editing both single images and groups of images. That's a narrow win, not a claim of overall superiority, and Meta's willingness to publish it suggests the company is trying to set realistic expectations rather than overpromise.

There's also a practical, cost-driven reason for building this in-house. Meta has relied on third-party AI models — including tools from Midjourney and Black Forest Labs — to handle image and video generation features inside its apps. Meta says it intends to lean on Muse Image to cut down on that outside dependence, which likely means lower licensing costs and tighter control over how the technology evolves.

What's next

Muse Image is rolling out gradually. It's live now in the Meta AI app and site, WhatsApp DMs, and Instagram Stories, with expansion planned to Facebook, Messenger, and additional corners of Instagram and WhatsApp later this year. Muse Video, the company's answer to AI video generation, is still to come — Meta describes it in a technical blog post as offering competitive performance on prompt adherence, visual fidelity, and keeping motion consistent over time, but hasn't given it a firm release date.

The bigger picture here is Meta trying to turn generative AI from a cool feature into an actual revenue line, whether through subscriptions from power users or through advertisers paying, directly or indirectly, for faster and cheaper ad production. Whether Muse Image can close the quality gap with OpenAI's and Google's tools will matter to consumers. Whether it can make Meta's ad business more efficient is the question that matters more to Meta's bottom line.