Meta Platforms (META) officially launched a new artificial intelligence image generation model, Muse Image, on Tuesday, aiming to further expand its generative AI tool ecosystem, attract content creators and advertisers, and accelerate the commercial application of AI capabilities in its advertising business. Boosted by the news, Meta’s stock closed up 2.55% against the market trend at $615.58 that day.
Muse Image, initially developed under the codename Mango, is the second major model released by Meta’s Super Intelligence Lab. The lab, led by Alexandr Wang, previously released the Muse Spark large language model in April this year to succeed the earlier Llama series. Meta stated that it will gradually reduce its reliance on third-party AI technologies such as Midjourney and Black Forest Labs through its self-developed models.
At the application level, general users can access Muse Image for free through the Meta AI app and website, WhatsApp direct messages, and Instagram Stories. For heavy users and content creators who need to generate more images or access advanced features, a subscription to Meta One, a monthly paid plan launched by Meta in May this year, is required. After free quotas are exhausted, users can choose to purchase a subscription or wait for quota resets.
In addition to consumer-facing applications, Muse Image will also be integrated into Meta’s AI advertising tools for advertisers, becoming part of the Advantage Plus service. The model can adjust image elements, replace styles, and generate multiple variations based on advertisers’ existing materials, thereby producing higher-quality ad creatives that better align with brand identity with fewer modification iterations. Meta stated that advertisers and agencies will begin to see image variation features powered by Muse Image within the coming weeks. Analysts believe this move demonstrates that Meta is deeply embedding AI capabilities into its core advertising business while continuing to increase investment in AI infrastructure and exploring new commercialization pathways.
Currently, OpenAI and Google (GOOGL) have taken early leads in the image generation field, with Google’s Nano Banana, launched last autumn, having sparked significant consumer interest. According to internal tests released by Meta, Muse Image underperforms OpenAI’s latest GPT Image 2 model in some image editing tasks but outperforms Google’s Nano Banana 2 model, particularly showing clear advantages in single-image and multi-image editing scenarios.
Meta also revealed plans to launch an AI video generation model, Muse Video, in the future. In a technical blog post, the company stated that the model demonstrates strong competitiveness in prompt comprehension, visual fidelity, and temporal consistency. Additionally, Muse Image will be expanded to Facebook, Messenger, and more functional scenarios within Instagram and WhatsApp later this year.
