Meta Launches New AI Image Generator to Compete with OpenAI and Google

Meta has introduced a new AI model called Muse Spark Image designed to generate images from complex prompts, aiming to attract creators and advertisers to its AI products and close the gap with competitors like OpenAI and Google. This release follows the April launch of the Muse Spark language model, which replaced Meta's previous Llama models and is part of the Meta Superintelligence Labs, a division led by Alexander Wong and funded by CEO Mark Zuckerberg with billions of dollars.

The Muse Spark Image tool will be available for free through Meta AI’s app and website, as well as via WhatsApp direct messages and Instagram Stories. However, active users and content creators will need to subscribe to one of Meta’s new monthly plans, launched in May, to generate large volumes of AI images and access certain features. Users who exceed their free quota can either purchase a subscription or wait for their quota to reset.

Meta also plans to integrate the new image model into its advertising tools to help create tailored ads across its platforms. Additionally, the company is developing a new video AI model called Muse Video. Until now, Meta has relied on external AI models like Midjourney for image and video generation, but the new models aim to reduce this dependency.

While the Muse Spark language model is an improvement over previous Meta models, Zuckerberg and Wong acknowledged it still lags behind competitors in programming tasks. The new image generator is a smaller, closed model, preventing external developers from customizing it. The tool has faced criticism after The Verge revealed it allows users to alter images of other Instagram users if their accounts are public, raising privacy concerns. One user on X warned that "pulling real users into generative images without explicit consent is a privacy landmine waiting to explode."