Muse ImageMuse Image: Meta's Agentic AI Image Generator
Meta's first image model plans its compositions, checks real-world context as it works and refines its own drafts before you ever see them. It has no public API yet — Molyin adds Muse Image the day Meta opens access, and every live image model is ready for your prompts today.
- Failed generations auto-refunded
- Credit packs never expire
- Every model's price is public
- Cancel anytime in one click
- Text to image · Image to image
- Generation modes
- 10 presets
- Aspect ratios
- #2
- Arena rank at launch
- 30+
- Suggested prompts
Muse Image example prompts
Five official launch images, each paired with the exact prompt that produced it. They follow a pattern you can reuse with any model: a clear subject, a few concrete style words, then lighting, mood and finishing tags.
- Muse Image
- Text to image
- 3:2
Floating island with waterfall and cherry blossoms, soft pink sky, glowing lanterns, magical and serene, detailed, dreamy art.
- Muse Image
- Text to image
- 1:1
A candid photo of a grandmother teaching her granddaughter how to make pasta in a warm, sunlit kitchen.
- Muse Image
- Text to image
- 1:1
Elaborate gold foil fox logo, luxury vector style, sitting side view, intricate geometric linework, negative space details, line-weight variation, premium packaging texture, highlights, black background
- Muse Image
- Text to image
- 3:2
Sunset over a calm lake with mountains, golden light, wildflowers, soft clouds, peaceful mood, highly detailed, 8k, photorealistic
- Muse Image
- Text to image
- 3:2
Cozy cabin in snowy forest at twilight, warm lights in windows, soft snowfall, starry sky, peaceful, detailed, cinematic, 4k
What is Muse Image?
Muse Image is the first image model from Meta Superintelligence Labs, launched on July 7, 2026. It is built as an agentic generator rather than a one-shot renderer: it plans a composition before drawing, can write and run code for accurate charts and diagrams, looks up real-time web context to get facts and references right, and reviews its own drafts — making local edits or regenerating until the result matches the instruction. At launch it ranked #2 on the Arena leaderboard for text-to-image, single-image editing and multi-image editing.
Beyond generation, Muse Image edits across multiple conversational turns, understands markup and sketch annotations on a photo, restores old pictures, and composes from multiple references — people, objects, clothing, styles and environments in one scene. It renders legible in-image text, and every output carries Content Seal, an invisible watermark that survives cropping, compression and screenshots. It currently lives inside Meta's own apps, with no developer API announced yet.
Molyin will add Muse Image the day Meta opens access — the model is already registered, so nothing about this page will change except the generate button going live. Until then, the same AI image generator runs GPT Image 2, Nano Banana, Seedream 5.0 and FLUX.2 with one credit balance on a single subscription.
Muse Image vs Nano Banana Pro
Muse Image is Meta's answer to Google's Nano Banana Pro — two consumer-scale image models with very different homes. Here is how Meta's launch positioning lines up against what Nano Banana Pro does on Molyin today.
| Capability | Nano Banana Pro | Muse Image |
|---|---|---|
| Positioning | Google's top image tier with 1K / 2K / 4K outputs | Meta's agentic model — plans, searches and self-refines |
| Editing | Image-to-image editing with reference images | Multi-turn conversational editing, reads markup annotations |
| Reference images | Up to 8 reference images | Multi-reference composition — people, objects, styles, scenes |
| Text rendering | Strong in-image text | Legible text, plus code-rendered charts and QR codes |
| Cost | 24–32 credits per image on Molyin | Free in Meta's apps; Molyin credit price set at API launch |
| Availability | Available on Molyin | Meta apps only — no developer API yet |
Comparison based on Meta's July 2026 launch announcement; Nano Banana Pro column reflects its capabilities on Molyin.
Muse Image vs image models available on Molyin
| Model | Positioning | Release date | On Molyin |
|---|---|---|---|
| Muse Image | Meta's agentic image model — search-grounded, self-refining | Jul 7, 2026 (Meta apps) | Coming soon |
| GPT Image 2 | OpenAI top-tier image model | Apr 21, 2026 (API) | Live — generate above |
| Nano Banana | Google fast / standard / pro image family | 2026 | Live — generate above |
| Seedream 5.0 | ByteDance image model, 1K / 2K / 3K / 4K | Jul 8, 2026 | Live — generate above |
| FLUX.2 | Black Forest Labs Dev / Pro / Max | 2025 | Live — generate above |
One subscription, one credit balance — every model available above is included.
Muse Image full specifications
| Parameter | Value | Notes |
|---|---|---|
| Generation modes | Text to image · Image to image | Generate from plain-language text, or upload an image and edit it conversationally. |
| Aspect ratios | 1:1 · 2:3 · 3:2 · 3:4 · 4:3 · 4:5 · 5:4 · 9:16 · 16:9 · 21:9 | Preset ratios Muse Image will support in the Molyin creator. |
| Editing | Multi-turn conversational | Keeps context across turns; understands markup and sketch annotations drawn on a photo. |
| Reference images | Multi-reference composition | Blends people, objects, clothing, styles and environments from several references into one scene. |
| Text rendering | Legible in-image text | Renders readable text, and can write and run code for accurate charts, plots and QR codes. |
| Web grounding | Built-in | Looks up real-time web context so knowledge-heavy prompts get facts and references right. |
| Self-refinement | Agentic | Reviews its own drafts and makes local edits, regenerates or changes tactics before returning a result. |
| Provenance | Content Seal (invisible) | An invisible watermark that survives cropping, compression, resizing and screenshots, with a public detection tool. |
| Where it runs today | Meta AI app · meta.ai · Instagram Stories · WhatsApp | Free for everyday creation; heavier use is part of Meta's subscription plans. Facebook and Messenger are next. |
| Developer API | Not yet available | Meta has announced advertiser tools first; a public API has not been dated. Molyin integrates the day it opens. |
Specifications reflect Meta's July 2026 launch announcement; Molyin values will be finalized when the developer API opens.
How to use Muse Image on Molyin
Once Meta opens developer access, Muse Image will work like every other image model on Molyin.
Pick the model
Choose any live image model in the creator — Muse Image joins the same picker the day Meta opens its API.
Describe your image
Write a plain-language prompt, or start from one of the official examples on this page and make it yours.
Generate and iterate
Generate, refine with follow-up edits or reference images, and download in the ratio your project needs.
Muse Image FAQ
What people ask about Meta's image model while it makes its way to Molyin.
Start Creating with Molyin Today
Sign up free and turn your first idea into a cinematic video in minutes.