Meta launched Muse Image on Tuesday, its first in-house artificial intelligence image model, letting users pull public Instagram photos into AI-generated pictures by tagging other people's accounts.
Key Points:
- Meta rolled out Muse Image, its first image model from Superintelligence Labs, across Meta AI, Instagram and WhatsApp.
- Users can tag public Instagram accounts in prompts, and the tool then draws on those photos to build new images.
- Public profiles are included by default, and account holders can switch the reuse setting off.
Muse Image Powers Meta's Apps
The company said the model now runs the image tools inside its Meta AI app, on Instagram and on WhatsApp, products used by billions each day, with Facebook and Messenger set to follow. It is the first image generator from Meta Superintelligence Labs, the division Mark Zuckerberg built last year to close the gap with rivals such as OpenAI and Google.
Users can tag a public account inside a prompt, and the model draws on that profile's photos to make a fresh picture. Meta AI can also blend a selfie with a vacation shot, redecorate a room using furniture listed on Facebook Marketplace, and render legible text or working QR codes. The model additionally drives more than 30 new AI effects for Instagram Stories in the United States.
Also Read: SpaceXAI Cursor-Powered Claude Killer Before The $60B Deal Even Closes
Instagram Privacy Draws Scrutiny
Public profiles are switched on by default, and reviewers warned that anyone can feed those photos into the model without asking permission first. Meta's help pages state that users will not be told when its AI features use their content.
Account holders can turn the option off inside their sharing settings, though the switch only blocks future creations, any pictures already generated will stay in place, and some users had not yet seen the updated controls.
Muse Image behaves as an agent, reasoning through a prompt and searching the web before it builds a picture, and engineers noted that it refines each result on its own. Every image carries Content Seal, an invisible watermark meant to flag AI origin. Internal tests place the model ahead of Google's Nano Banana 2 on editing, yet behind OpenAI's GPT Image 2 on overall quality.
Meta AI Race Heats Up
Meta debuted Muse Spark in April as the first model in a lineup meant to replace its open Llama family. Muse Image extends that system into pictures, everyday creation stays free while heavy users need a paid plan, and advertisers gain access through the Advantage Plus tools within weeks.
The Muse rollout follows Meta's roughly $14 billion move to bring Alexandr Wang and his Scale AI team aboard in 2025. Zuckerberg set up Superintelligence Labs after the company trailed rivals in the AI race. Wang has since teased a Muse Video model, now in preview, that Meta says rivals leading systems on visual fidelity and temporal consistency.
Read Next: Cheaper Fable 5 Rivals: Don't Overpay For Daily AI Work