Meta's first in-house image model reasons before it draws, grounds images with web search and edits single or multiple pictures. Three tiers from 1 to 4 credits, five aspect ratios, no watermark, on every Aura AI plan.
Meta Muse is the name Aura AI uses for Muse Image, the image generation model that Meta Superintelligence Labs released on July 7, 2026, alongside Muse Video. It is the first media model from the lab Meta set up under Alexandr Wang. Muse Image powers image creation in the Meta AI app, on meta.ai, in Instagram Stories in the US and in WhatsApp in a limited set of countries, and Meta has also made it available through third-party APIs.
Meta describes Muse Image as an agentic system rather than a plain text-to-image model. It builds on Muse Spark, Meta's language model, and the two share tools and plan jointly. In practice Muse Image can reason through a request before generating, call web search to ground a picture in current information or real visual references, and improve with more test-time compute on hard prompts. It performs single-image and multi-image editing and keeps results coherent across several editing turns. Meta reported second place on Arena for text-to-image, single-image editing and multi-image editing by human preference Elo as of July 5, 2026.
On Aura AI, Meta Muse comes in three text-to-image tiers: Muse Image at 1 credit, the premium Muse Image tier at 2 credits and the Muse flagship at 4 credits per image, each with a matching edit mode at the same price. All tiers run in five aspect ratios on every plan, without watermark. Its closest rivals in the studio are Nano Banana, GPT Image 2 and FLUX 2, all one click away in the same model selector.
Six things that make Meta's first image model different
Muse Image plans the picture the way a language model plans an answer. Meta reports that quality rises with more test-time compute, which is what the higher tiers on Aura AI are for.
The model learned to search online when a prompt depends on real-world knowledge, so a request about a specific landmark, product category or current event can be grounded in actual references.
Upload a picture and describe the change. Meta Muse Edit applies it with precision and keeps everything you did not mention intact. Three edit tiers on Aura AI at 1, 2 and 4 credits.
Combine several photos into one result: a product into a scene, two people into one frame, a logo onto packaging. Meta reported second place on Arena's multi-image editing board at launch.
Edit, then edit again. The model maintains coherence across multiple editing rounds, so a five-step retouch does not drift away from the original subject.
Aura AI exposes Meta Muse at 1, 2 and 4 credits so you can match the spend to the job: quick drafts at 1 credit, client work at 2, and the flagship for final hero images and hard composites.
Four steps from prompt to a downloadable image
Three tiers for generation, three matching tiers for editing, flat per image
| Variant | Mode | Credits per image | Best for |
|---|---|---|---|
| Muse Image | Text to image | 1 credit | Drafts, social posts, quick variations |
| Muse Image (premium tier) | Text to image | 2 credits | Client-facing images, detailed prompts |
| Muse (flagship) | Text to image | 4 credits | Hero images, knowledge-heavy prompts, hard scenes |
| Muse Image Edit | Image edit | 1 credit | Simple fixes and swaps |
| Muse Image Edit (premium tier) | Image edit | 2 credits | Everyday retouching and relighting |
| Muse Edit (flagship) | Image edit | 4 credits | Multi-image composites, precise edits |
How Meta's model compares to the other big-lab image generators on Aura AI
| Feature | Meta Muse | Nano Banana Pro | GPT Image 2 | FLUX 2 |
|---|---|---|---|---|
| Maker | Meta | OpenAI | Black Forest Labs | |
| Text-to-image credits | 1, 2 or 4 | 3 | 4 | 2 (Turbo 1, Max 3) |
| Image edit credits | 1, 2 or 4 | 2 | 4 | 2 (Max Edit 3) |
| Reasoning and tool use | ✓ Web search, planning | ✓ Gemini reasoning | ✓ Yes | No |
| Multi-image editing | ✓ Yes | ✓ Yes | ✓ Yes | ✓ Multi-reference |
| Arena position at launch | 2nd (Meta, July 5, 2026) | Top tier | Top tier | Not the focus |
| Price tiers on Aura AI | 3 | 2 | 1 | 4 |
| Plan | All plans | All plans | All plans | All plans |
Where reasoning, grounding and multi-image editing pay off
Drop a real product photo into a generated scene, or merge two photos into one shot. Multi-image editing keeps the product's shape, label and color intact.
Images that depend on how a real thing looks: a specific architectural style, a regional dish, a sports kit. Web grounding gives the flagship tier a better chance of getting details right.
The 1 credit tier is built for the daily feed: story backgrounds, quote cards, quick illustrations. Generate ten, keep three, upgrade the winner on the 2 credit tier.
Fix a portrait in several passes: background first, then wardrobe, then lighting. Coherence across editing turns keeps the subject the same person from the first step to the last.
Common questions about Meta Muse on Aura AI
Meta's agentic image model in three tiers from 1 to 4 credits, editing included, five aspect ratios, no watermark. Available on every plan, worldwide.