01
Muse Image is live, but not everywhere.
Meta says it is available in the Meta AI app, on meta.ai, Instagram Stories in the US, and WhatsApp in limited countries.
Field report - Meta Muse
Meta Muse Image and Muse Video: What Creators Need to Know
Meta Muse Image is not just another prompt-to-picture launch. It is Meta's agentic image workflow entering the apps where creators already publish, with Muse Video waiting behind it as a native-audio preview.
02
The real angle is agentic image work.
Muse Image can use search and coding tools, self-refine, compose references, and edit through iterative instructions.
03
Muse Video is still a preview.
Meta previewed native-audio video generation, but it is not a public creator tool yet.
04
Benchmarks are strong and dated.
Arena had Muse Image at No. 2 in three image tasks and Muse Video at No. 3 in text-to-video as of July 5, 2026.
05
No public API or open weights are verified.
This page does not claim developer access, downloadable weights, or exact subscription pricing.
Official image gallery
All the Muse launch images in one grid.
Browse the official Muse launch images and click any one to open a larger preview.
Official video gallery
All the Muse launch videos in one grid.
Watch selected Muse Image and Muse Video demos from Meta's launch.
Official demo displays
The launch demos worth actually seeing.
These are selected visual demos from Meta's launch page, pulled into cleaner panels so you can see the editing and composition behavior without the long prompt-thread clutter.
Official Meta demo source
A multi-turn visual world
The cat-and-dog cafe sequence is the clearest demo of Muse Image maintaining an idea across multiple creative turns: picnic scene, framed cafe photo, storefront, and menu design.
Reference-aware interiors
This is the cleaner interior example: one clear before-and-after view that shows Muse carrying a room through an edit without burying the point in extra outputs.
What Muse Image is
A social-native image workflow, not a standalone lab demo.
Meta says Muse Image is the first image generation model from Meta Superintelligence Labs and is now being placed inside Meta AI, meta.ai, Instagram Stories in the US, and WhatsApp in limited countries. The practical point for creators is distribution: this is designed for the Meta ecosystem, not just a model card.
The interesting part is how it works. Muse Image can reason through a prompt, call search, write code for precise visual tasks, self-refine, edit images over multiple turns, and compose a result from several references.
Agentic capabilities
The useful difference is how much of the work it can plan.
A normal image generator maps a prompt to an image. Meta describes Muse Image more like an agent: it can gather context, use tools, refine its own drafts, and keep iterating toward the result.
Search-grounded prompts
Meta says search helps Muse Image ground current, factual, brand, landmark, and identity-heavy prompts with better visual references.
Coding tool use
Muse Image can write and execute code for outputs that need precision, such as plots, QR codes, rendered figures, and structured visuals.
Self-refinement
The model can reflect on an output, correct local errors, regenerate larger misses, or choose another tactic before it presents the result.
Multi-reference composition
Muse Image can combine people, objects, clothing, styles, and environments from multiple image references in a single prompt.
Editing and composition examples
This is where creators should pay attention.
The strongest use cases are not just "make me an image." They are reference blending, room or product edits, social mockups, accurate charts, and repeated small changes without losing the original idea.
Search improves factual prompts
Official Meta chart. Use this as Meta-reported internal ablation context, not an independent benchmark.
Precise room editing
Official Muse Image demo showing the kind of practical edit creators can ask for directly.
Multi-reference composition
Official Muse Image demo showing identity, wardrobe, and setting references combined into one output.
Muse Video preview
Native audio is the hook. Public access is not here yet.
Meta previewed Muse Video alongside Muse Image and says it is built on the same pretraining base, with native audio support, prompt adherence, visual fidelity, and temporal consistency. The careful read: Meta also says audio-video sync and physically accurate fast motion remain areas of investment.
Status: preview / in development
Benchmarks
Strong numbers, dated carefully.
These are Arena leaderboard snapshots, not a blanket claim that Muse is better than every frontier image or video model. Use them as current signal and keep the date attached.
| Arena | Date | Leader | Leader score | Muse rank | Muse score | Muse votes | Source |
|---|---|---|---|---|---|---|---|
| Text-to-image | Jul 5, 2026 | gpt-image-2 (medium) | 1385+/-5 | No. 2 | 1280+/-7 | 7,715 | Open |
| Single-image edit | Jul 5, 2026 | gpt-image-2 (medium) | 1466+/-4 | No. 2 | 1405+/-6 | 12,751 | Open |
| Multi-image edit | Jul 5, 2026 | gpt-image-2 (medium) | 1454+/-5 | No. 2 | 1399+/-6 | 13,851 | Open |
| Text-to-video | Jul 5, 2026 | Gemini Omni Flash | 1527+/-13 | No. 3 | 1459+/-15 | 2,152 | Open |
Benchmark caveat
Artificial Analysis text-to-image, image-editing, and text-to-video leaderboards were checked before publication. Muse did not surface in the accessible top rows, so this page does not assign Muse an Artificial Analysis rank.
Available now
Limited rollout surfaces.
- Meta AI app
- meta.ai
- Instagram Stories in the US
- WhatsApp in limited countries
Coming later
Useful, but future-tense.
- Messenger
- More countries and surfaces
- Advantage+ creative workflows
- Muse Video for creators and Meta AI
Creator workflows
What I would actually test first.
Muse makes the most sense where the image lives inside the social context: Stories, product ideas, reference blending, branded mockups, and fast iteration before a post or ad goes live.
TestTurn a rough Instagram Story idea into multiple visual directions before you commit to a post.
TestBlend brand, product, wardrobe, and environment references into one social-native creative concept.
TestUse search-grounded image generation for visual prompts that depend on current facts, places, or public context.
TestBuild better social ads and product mockups before the Advantage+ creative rollout arrives.
TestUse iterative editing for small creative changes instead of restarting the whole prompt from scratch.
Official media gallery
Real launch demos, not generic AI art.
These selected visuals come from Meta's official launch pages, so the article shows the actual Muse demos instead of filler imagery.
Search improves factual prompts
Official Meta chart. Use this as Meta-reported internal ablation context, not an independent benchmark.
Precise room editing
Official Muse Image demo showing the kind of practical edit creators can ask for directly.
Multi-reference composition
Official Muse Image demo showing identity, wardrobe, and setting references combined into one output.
Code-assisted visuals
Official Muse Image demo connected to the coding-tool angle for charts, plots, and structured images.
Muse demo clip
Official Meta demo video.
Caveats
The claims this page is not making.
- Muse Image is not described as globally available across every Meta product.
- Muse Video is previewed, not publicly launched.
- No public Muse Image or Muse Video API, open weights, model download, or developer endpoint is verified here.
- Meta mentions subscription plans for more creation, but this page does not publish an exact price.
- Benchmark claims are dated leaderboard snapshots, not a universal claim that Muse beats every frontier model.
Sources
Source list and update trail.
Factual claims on this page are anchored to official Meta pages, Arena leaderboard snapshots, and Artificial Analysis context pages.
01Meta AI research launchPrimary source for Muse Image, Muse Video, agentic capabilities, Arena claims, Content Seal, and rollout caveats.02About Meta product launchPrimary product source for Meta AI, Instagram, WhatsApp, creator access, social references, and official media.03Arena text-to-image leaderboardDated July 5, 2026 leaderboard snapshot used for the Muse Image text-to-image row.04Arena image-edit leaderboardDated July 5, 2026 leaderboard snapshot used for single-image editing.05Arena multi-image edit leaderboardDated July 5, 2026 leaderboard snapshot used for multi-image editing.06Arena text-to-video leaderboardDated July 5, 2026 leaderboard snapshot used for Muse Video.07Artificial Analysis text-to-image leaderboardIndependent context source checked before publish; Muse did not surface in the top rows from the accessible page text.08Artificial Analysis image editing leaderboardIndependent editing context source checked before publish; Muse did not surface in the accessible top rows.09Artificial Analysis text-to-video leaderboardIndependent video context source checked before publish; Muse did not surface in the accessible top rows.
Last public page update: July 8, 2026. Route: /meta-muse-image-video
Work with Matt
Need help turning AI launches into usable workflows?
Work with Matt for practical AI consulting, creator workflow strategy, training, sponsored AI videos, and AI tool marketing that turns product news into clear action.