01
Faithful instruction following
Translate detailed creative direction into coherent images while keeping requested subjects, styling, layout, and visual relationships in focus.
Muse Image AI Generator
Turn detailed instructions and multiple visual references into polished images. Explore precise editing, faithful composition, and an agentic creative workflow powered by Muse Image.
No. 2Arena: text-to-image
No. 2Arena: single-image editing
No. 2Arena: multi-image editing
MSLMeta Superintelligence Labs
From instruction to finished image, with more room to think
Muse Image is Meta's most advanced image generation model, designed to follow instructions faithfully, edit with precision, and compose scenes from multiple references.
What makes Muse Image different
Its agentic workflow can reason, use tools, inspect drafts, and choose a better approach before returning the result.
02
Precise, iterative image editing
Change exactly what the instruction asks for while maintaining coherence through multiple editing turns and open-ended exploration.
03
Compose from multiple references
Combine people, objects, clothing, styles, and environments from many reference images, interleaving text and visuals in one prompt.
04
Reason, use tools, and self-refine
Muse Image can use search and coding tools, inspect its own draft, make a local correction, regenerate, or choose a more accurate strategy.
Grounded in official model evidence
These facts come from Meta Superintelligence Labs' July 2026 introduction of Muse Image and Muse Video.
Human-preference rankings
Meta reports the No. 2 Arena position for text-to-image, single-image editing, and multi-image editing as of July 5, 2026.
Test-time scaling
Official ablations show quality improves as the model spends more inference-time compute on reasoning, tool calls, and refinement.
Content Seal provenance
Meta AI outputs made with Muse Image include an invisible signal designed to survive cropping, compression, resizing, and screenshots.
A single model for generating, editing, and composing
Build a new image, revise an existing one, or combine people, products, clothing, styles, and environments from several references.
Generate from a detailed brief
Turn art direction, copy requirements, style, and composition into a new visual.
Edit without losing coherence
Refine a specific object, material, color, or layout across an ongoing creative thread.
Blend multiple visual references
Bring a subject, wardrobe, setting, and style from separate inputs into one composition.
How to create with Muse Image
Muse Video joins the creative workspace
Muse Video shares Muse Image's pretraining base and is previewed with native audio, strong prompt adherence, visual fidelity, and temporal consistency.
Native audio support
Muse Video is built to generate visuals and audio from the same media-generation foundation.
Visual and temporal consistency
The preview emphasizes prompt adherence, visual fidelity, and consistency across time.
No. 3 on text-to-video Arena
Meta reports the No. 3 human-preference Elo position as of July 5, 2026. Muse Video is still coming soon.
FAQs about Muse Image and Muse Video
What is Muse Image?
Muse Image is Meta Superintelligence Labs' advanced image generation model for instruction following, precise editing, multi-reference composition, tool use, and self-refinement.
Can Muse Image edit an existing image?
Yes. Meta describes precise editing with coherence across multiple turns, supporting iterative refinement toward a target result.
How does multi-reference composition work?
The model can combine elements from multiple input images, including people, objects, clothing, styles, and environments, alongside inline text instructions.
What does agentic image generation mean?
Instead of mapping a prompt directly to one image, Muse Image can reason, use search or coding tools, inspect its output, and refine its approach.
Is Muse Video available now?
Meta describes Muse Video as an early preview that is coming soon to creators and Meta AI. The MuseImagine workspace labels it as Preview and does not submit a real generation request.
Does MuseImagine add Meta Content Seal?
No claim is made that this prototype adds Content Seal. Meta states that images created by Muse Image in Meta AI products carry the invisible provenance signal.
Start your next image with Muse
Describe the result, add references, and continue refining in the MuseImagine workspace.