muse-imageMuse Image is Meta's image generation model, Muse Image 1.0 from Meta Superintelligence Labs. Meta describes it as reasoning over a prompt before it renders, with precise composition and text rendering. On APIMODELS it is one model id, muse-image (muse-image-1.0 also works), on the unified image endpoint, for both text-to-image and image editing. Send a prompt of up to 4,000 characters and you get one image of about 2.5 megapixels. Add one to four reference images (image for one, images for several; a public URL or a base64 data URI, PNG, JPEG or WebP, up to 10 MB each and 24 MB in total) and the same request edits them instead: restyle a photo, change the background, or combine things from several pictures. The first image is the main one, and the prompt can point at the others ("put the cup from the second image on the table"); say what should stay and what should change. Reference images cost nothing extra. Leave aspect_ratio out, or set it to auto, and the model picks the shape from the prompt; in our test an unsized request came back at 1920×1280. To fix the shape, set aspect_ratio to 1:1, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, 2:1, 1:2, 21:9 or 9:21, or pass an OpenAI-style size such as 2048x1152. The size sets the proportions and the model renders at its own resolution (about 1600×1600 for 1:1, 2048×1152 for 16:9), so the pixel count is not exact. Sizes above about 2.6 megapixels, such as 4K or 2048x2048, are rejected with a 400 rather than delivered smaller. Each request returns exactly one image; n above 1 is rejected, so a batch is several parallel requests. Masks (inpainting), more than four reference images, a transparent background, negative_prompt and seed are not supported and are rejected with a 400 instead of being silently ignored. Not every request succeeds: the model can refuse content or hit a temporary error, and a failed request is not charged. Pricing is a flat $0.02 per image, charged only when an image is delivered.
Generated image will appear here
Enter a prompt and click Generate
Muse Image 1.0: a prompt alone, or a prompt plus 1–4 reference images to restyle, change the background or combine pictures; same price either way
1600×1600 at 1:1, 2048×1152 at 16:9; the model sets the exact pixels and larger sizes are rejected
aspect_ratio auto lets the prompt decide; 13 fixed ratios from 21:9 to 9:21, or an OpenAI-style size
Masks, more than 4 reference images, n above 1, sizes above 2.6 MP, transparent backgrounds, negative_prompt and seed return a 400 instead of being silently ignored
Not every request succeeds; a request that fails returns the reason and is not charged
One flat price at every aspect ratio
Muse Image is Meta's image model, Muse Image 1.0. Send a prompt and get one image of about 2.5 megapixels, or add up to four reference images to edit a photo, restyle it or combine pictures. The shape follows the prompt or one of 13 fixed aspect ratios.
On APIMODELS it is one model id, muse-image, on POST /v1/images/generations, at $0.02 per image with or without reference images. It has no masks, no 4K and no transparent backgrounds, and not every request succeeds; the limits are listed below.
muse-imageWhat Muse Image accepts and what it does not. Unsupported inputs and sizes are rejected with a 400 instead of being silently ignored, so you never pay for an image that ignored your request.
| Supported | Not supported | |
|---|---|---|
| Input | A text prompt of up to 4,000 characters, alone or with 1 to 4 reference images for editing, restyling or combining pictures (the first image is the main one) | Masks (inpainting) |
| Reference images | 1 to 4, PNG, JPEG or WebP, up to 10 MB each and 24 MB in total, as public URLs or base64 data URIs | More than 4 (rejected with 400) |
| Images per request | 1 | n above 1 (send parallel requests for a batch) |
| Resolution | About 2.5 MP at the model's native size for each ratio: 1:1 1600×1600, 3:2 1920×1280, 16:9 2048×1152, 9:16 1152×2048, 21:9 2352×1008 | 4K, 2048×2048 or any size above about 2.6 MP (rejected with 400); exact custom pixel sizes, because size only sets the proportions |
| Aspect ratio | auto (the prompt decides), 1:1, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, 2:1, 1:2, 21:9, 9:21, or an OpenAI-style size up to 16:1 | Ratios beyond 16:1 |
| Output format | JPEG by default; PNG or WebP with output_format | Transparent background (background: transparent is rejected) |
| Other controls | None beyond the above | negative_prompt and seed (both rejected; results are not repeatable); quality tiers (quality has no effect) |
Like every image model, Muse Image does not succeed 100% of the time. You are charged only when an image is delivered.
A request can fail after it is accepted: the model can refuse the content, or a generation can hit a temporary error. A failed request returns the reason and costs nothing.
All seven of our test requests on 30 October 2026 returned an image, which is far too few to promise a success rate, so we do not publish one yet. Build in retries for temporary errors, rewrite refused prompts, and allow 35 to 70 seconds per image (our tests took 34 to 67).
One image of about 2.5 MP per prompt, in the ratio you need, at a flat $0.02. Meta describes careful composition and text rendering.
Add up to four reference images to change a background, restyle a photo or put an object from one picture into another. The first image is the main one; say in the prompt what stays and what changes. The price stays $0.02 per image.
Each request is one image at one price, so a batch of 100 is 100 parallel requests and $2.00.
There is no mask and no inpainting, so you cannot mark the exact area to change. Describe the change in the prompt instead; the result is a new image of about 2.5 MP.
Output is about 2.5 MP at the model's native size for the ratio, with an opaque background. Larger sizes and transparent backgrounds are rejected.
There is no seed, so the same prompt gives a different image each time.
Muse Image is an Image Generation API provided by Meta. Muse Image is Meta's image generation model, Muse Image 1.0 from Meta Superintelligence Labs. Meta describes it as reasoning over a prompt before it renders, with precise composition and text rendering. On APIMODELS it is one model id, muse-image (muse-image-1.0 also works), on the unified image endpoint, for both text-to-image and image editing. Send a prompt of up to 4,000 characters and you get one image of about 2.5 megapixels. Add one to four reference images (image for one, images for several; a public URL or a base64 data URI, PNG, JPEG or WebP, up to 10 MB each and 24 MB in total) and the same request edits them instead: restyle a photo, change the background, or combine things from several pictures. The first image is the main one, and the prompt can point at the others ("put the cup from the second image on the table"); say what should stay and what should change. Reference images cost nothing extra. Leave aspect_ratio out, or set it to auto, and the model picks the shape from the prompt; in our test an unsized request came back at 1920×1280. To fix the shape, set aspect_ratio to 1:1, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, 2:1, 1:2, 21:9 or 9:21, or pass an OpenAI-style size such as 2048x1152. The size sets the proportions and the model renders at its own resolution (about 1600×1600 for 1:1, 2048×1152 for 16:9), so the pixel count is not exact. Sizes above about 2.6 megapixels, such as 4K or 2048x2048, are rejected with a 400 rather than delivered smaller. Each request returns exactly one image; n above 1 is rejected, so a batch is several parallel requests. Masks (inpainting), more than four reference images, a transparent background, negative_prompt and seed are not supported and are rejected with a 400 instead of being silently ignored. Not every request succeeds: the model can refuse content or hit a temporary error, and a failed request is not charged. Pricing is a flat $0.02 per image, charged only when an image is delivered. Through APIMODELS platform, you can access this model via a unified API with transparent pay-as-you-go pricing. Current pricing: per image: $0.02.
Generate high-quality product visuals for online stores, ads, and marketing materials.
Create eye-catching visual content for social platforms to boost engagement and brand visibility.
Produce concept art for characters, scenes, and props to accelerate game development.
Design posters, banners, and promotional graphics at a fraction of traditional design costs.
Muse Image is available through APIMODELS at: per image: $0.02. Billing is pay-as-you-go — you only pay for what you generate.
Sign up at APIMODELS, get your API key, and call our unified API endpoint. We provide detailed API documentation with code examples in cURL, Python, and Node.js.
APIMODELS offers the same Muse Image model through our aggregation platform. We provide a unified API interface so you do not need separate accounts for each provider - one API key to access all models.
Muse Image is Meta's image generation model, Muse Image 1.0 from Meta Superintelligence Labs. Meta describes it as reasoning over a prompt before it renders, with precise composition and text rendering. On apimodels.app it is one model id, muse-image (muse-image-1.0 also works), on POST /v1/images/generations: send a prompt of up to 4,000 characters for text-to-image, or add 1 to 4 reference images to edit a picture or combine several, and you get one image of about 2.5 megapixels. You are charged only when an image is delivered, a flat $0.02 per image.
All of Muse Image's limits: text-to-image, or image-to-image with 1 to 4 reference images, the first being the main image (PNG, JPEG or WebP, each up to 10 MB and up to 24 MB together; no masks or inpainting); one image per request (n must be 1); about 2.5 megapixels at a native size for each ratio (1:1 1600×1600, 3:2 1920×1280, 16:9 2048×1152, 9:16 1152×2048, 21:9 2352×1008), with no 4K, no sizes above about 2.6 megapixels and no exact custom pixel dimensions; aspect_ratio auto or one of 13 ratios (1:1, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, 2:1, 1:2, 21:9, 9:21), or an OpenAI-style size up to 16:1; JPEG by default, PNG or WebP through output_format, no transparent background; no negative_prompt, no seed (results are not repeatable) and no quality tiers; prompts up to 4,000 characters. Unsupported inputs, such as a mask, more than four reference images, n above 1, a size above about 2.6 megapixels, resolution 4K, a transparent background, negative_prompt or seed, are rejected with a 400 instead of being silently ignored. Even within these limits, not every request succeeds: the model can refuse content or hit a temporary error, and a failed request is not charged.
aspect_ratio defaults to auto, which lets the model pick the shape from the prompt; in our test an unsized request came back at 1920×1280. To fix the shape, choose one of 13 ratios: 1:1, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, 2:1, 1:2, 21:9 or 9:21. You can also pass an OpenAI-style size such as 2048x1152; it sets the proportions, up to 16:1, and takes priority over aspect_ratio. Any other ratio, a size not written as WIDTHxHEIGHT, or a size above about 2.6 megapixels (such as 3840x2160) returns a 400. The price is the same at every shape.
About 2.5 megapixels, at a fixed native size for each shape. In our tests on 30 October 2026 (7 requests), 1:1 came back at 1600×1600 (a 1024x1024 request also came back at 1600×1600), 16:9 at 2048×1152, 9:16 at 1152×2048 and 21:9 at 2352×1008, and one auto request with no size came back at 1920×1280. There is no 4K: size only sets the proportions, and a size above about 2.6 megapixels, such as 3840x2160 or 2048x2048, is rejected with a 400 instead of being delivered smaller. Read the width and height of the image you get back instead of assuming the numbers you sent.
No. There is no 4K: output is about 2.5 megapixels, at the model's native size for each shape. A size above about 2.6 megapixels, such as 3840x2160 or 2048x2048, is rejected with a 400 before anything is created instead of being delivered smaller, and resolution set to 4K is rejected the same way; a rejected request is not charged. Smaller sizes are accepted and only set the proportions: in our tests a 1024x1024 request came back at 1600×1600. If you need exact pixel dimensions, resize the result on your side; if you need 4K, pick an image model on apimodels.app that offers a 4K output. Images come back as JPEG by default; set output_format to png or webp to get that format instead.
Yes. The same model id, muse-image, also does image-to-image: pass 1 to 4 reference images to edit a picture, change its style or combine several pictures into one. The first image is the main image, the one being changed, and the prompt can point to the others by position, such as "the second image". Send one image as image (or image_url) and several as an array in images (or image_urls); each can be a public URL or a base64 data URI, and raw base64 also works through image_base64 with image_mime_type. Images must be PNG, JPEG or WebP, each up to 10 MB and up to 24 MB together. You still get one image of about 2.5 megapixels, shaped by aspect_ratio or size, at the same $0.02 however many reference images you send. Masks and inpainting are not supported: a mask returns a 400 ("muse-image does not support masks (mask): describe what to change in prompt instead"), so describe the change in the prompt instead. A fifth image returns a 400 too ("muse-image takes at most 4 reference images (got 5)"). Both are rejected before anything is created and are not charged. We have not run our own image-to-image tests yet, so we do not quote measured editing results.
Put the picture you want to change first in images, then say in the prompt what stays and what changes, for example "keep the person and their position, change the background to a beach, oil-painting style". With several images, say what each one is for: "use the first image as the base and put the cup from the second image on the table". Two things to expect: small text inside a reference image may not be preserved, and long passages of text in the output are decorative only and can contain typos, so add exact long copy yourself afterwards. Every edit is one request that returns one image at $0.02, charged only when it is delivered. Not every request succeeds: the model can refuse content or hit a temporary error, so retry temporary errors and rewrite a refused prompt; a failed request is not charged.
One. n must be 1, and n=2 or more returns a 400 ("muse-image generates one image per request (got n=2); submit one request per image") before anything is created. Reference images do not change this: one to four images in, one image out. To make several images, send several requests in parallel; each one is a separate task, billed $0.02 only if its image is delivered.
About 35 to 70 seconds per image in our tests: 7 text-to-image test requests on 30 October 2026 took between 34 and 67 seconds each; we have not timed image-to-image requests ourselves yet. The call is asynchronous, so create the task, then poll GET /v1/images/generations?task_id= every few seconds or pass callback_url. If you use the one-call OpenAI-style mode (size without callback_url), set your client timeout to at least 120 seconds.
Not every Muse Image request succeeds. The model can refuse the content, or a generation can hit a temporary error. A failed request returns the reason and is not charged. All 7 of our text-to-image test requests on 30 October 2026 returned an image, but seven tests are far too few to promise a success rate, so we do not publish one, and we have not tested image-to-image ourselves yet. Retry temporary errors, rewrite a refused prompt, and do not build a pipeline that assumes the first attempt always delivers. Requests outside Muse Image's limits, such as a mask or more than four reference images, are rejected with a 400 before anything is created, also without charge.
A flat $0.02 per image, charged only when an image is delivered. The price is the same for text-to-image and for editing with one to four reference images, and at every aspect ratio and size; since each request returns one image, it is also $0.02 per request, and 100 images cost $2.00. Requests rejected with a 400 and generations that fail are not charged.
Create an account at apimodels.app and generate an API key in the console; one key and one USD balance cover every model. Then send: curl -X POST https://api.apimodels.app/v1/images/generations -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"model": "muse-image", "prompt": "a bakery window poster with the headline FRESH BREAD DAILY, warm morning light", "aspect_ratio": "3:4"}'. To edit or combine pictures, send the same call with images, one to four public URLs or base64 data URIs with the main image first: -d '{"model": "muse-image", "prompt": "keep the person and their position, change the background to a beach, oil-painting style", "images": ["https://example.com/portrait.jpg"]}'. The response holds a taskId: poll GET /v1/images/generations?task_id= until the state is completed, or pass callback_url, and take the image from resultUrls within 30 days. If you send an OpenAI-style size such as 2048x1152 and no callback_url, the same call returns the image directly in the OpenAI response format.
Yes, for a first look. Eligible new accounts get $0.10 of trial credit, which covers five Muse Image generations at $0.02 each, so you can check the output before paying anything. After that, top-ups start at $10 and billing is pay-as-you-go with no subscription. An image is charged only when it is delivered.
On APIMODELS, Muse Image runs alongside 60+ models on one API key and one balance, so choosing is about fit, not lock-in. It supports Text to Image, Image Editing · 1–4 Reference Images, About 2.5MP, Auto or 13 Aspect Ratios, One Image per Request, $0.02/image, and you can weigh it on price and capability against other Image Generation models, then switch by changing a single model-name string — no new account or integration. Browse every Image Generation option with live pricing at apimodels.app/models.
Muse Image supports: Text to Image, Image Editing · 1–4 Reference Images, About 2.5MP, Auto or 13 Aspect Ratios, One Image per Request, $0.02/image. See the APIMODELS docs for full parameters and call examples.
Yes. APIMODELS exposes Muse Image through a single unified API and one key — no separate provider accounts, and no need to handle each provider's regional network access yourself.
We support Stripe (Visa, Mastercard, and other international cards) and Alipay. Credits are available instantly after payment.