What inputs does Meta Muse Image Text-to-Image support?

The model requires a text prompt and supports optional controls for aspect ratio, number of images, and output format. You can use aspect ratio presets such as square, widescreen, portrait, and tall formats depending on the final layout.

What output formats can I generate?

Meta Muse Image Text-to-Image can return JPEG, PNG, or WebP images. WebP is the default for compact image delivery, while PNG is useful for lossless workflows and JPEG is useful for broad compatibility.

Meta Muse Image Text to Image image preview

Example inputhover

prompt: "cinematic wildflower meadow during late golden hour, tall soft grasses and delicate burgundy, peach and cream flowers moving gently in the breeze, an old wooden chair placed alone among the flowers with an open sketchbook resting on it, distant rolling hills fading into warm atmospheric haze, dreamy but highly realistic photography, rich autumn-inspired colors, soft natural sunlight, subtle film grain, poetic editorial aesthetic, sophisticated composition"
num_images: 1
aspect_ratio: "16:9"
output_format: "png"

Meta Muse Image Text to Image

Image·muse-image·by Meta

Meta Muse Image Text-to-Image creates detailed images from prompts with strong instruction following, accurate text rendering, and flexible aspect ratios.

Try it now →

API reference

Runtime (p50): 1m
Estimated price: $0.01 / image

Call the API

prediction.sh

curl -X POST \
  -H "Authorization: Bearer $EACHLABS_API_KEY" \
  -H "Content-Type: application/json" \
  --data '{
    "model": "meta-muse-image-text-to-image",
    "version": "0.0.1",
    "input": {
        "prompt": "cinematic wildflower meadow during late golden hour, tall soft grasses and delicate burgundy, peach and cream flowers moving gently in the breeze, an old wooden chair placed alone among the flowers with an open sketchbook resting on it, distant rolling hills fading into warm atmospheric haze, dreamy but highly realistic photography, rich autumn-inspired colors, soft natural sunlight, subtle film grain, poetic editorial aesthetic, sophisticated composition",
        "num_images": 1,
        "aspect_ratio": "16:9",
        "output_format": "png"
    },
    "webhook_url": ""
}' \
  https://api.eachlabs.ai/v1/prediction/

Documentation8 sections

Overview
Meta Muse Image Text to Image Overview

Meta Muse Image Text to Image is Meta’s in-house generative image model that creates and edits rich visuals directly from natural language prompts. Built by Meta Superintelligence Labs as part of the broader Muse family, it focuses on instruction-following fidelity, accurate text rendering inside images, and multi-step “agentic” reasoning over complex requests. Within a few weeks of launch, Muse Image ranked among the top models on community text-to-image leaderboards, making it a strong choice for high-quality visuals where prompt precision and layout control matter. On each::labs, the Meta Muse Image Text to Image endpoint provides convenient access to this capability, so developers and creators can generate on-brand ads, posters, and concept art from concise prompts without managing Meta’s consumer interfaces.
Capabilities
Capabilities
- High-fidelity text-to-image generation from natural language prompts, suitable for ads, posters, thumbnails, and social content.
- Multi-image composition, allowing Muse Image to blend multiple reference photos into a coherent scene while preserving key subjects and brand elements.
- Fine-grained image editing, including region-constrained edits guided by sketches or markup, so users can adjust specific areas without rebuilding an entire image.
- Accurate in-image text rendering, leveraging code-based rendering and reasoning for readable headlines, labels, QR codes, and charts embedded in the output.
- Agentic self-refinement, where the model drafts, evaluates, and improves images based on instructions for style, branding, or constraints.
- Contextual alignment with social formats, having been deployed across Meta AI, Instagram Stories, WhatsApp, and Ads Manager, with outputs tuned for feed, story, and ad placements.
- Ad creative optimization support, powering Advantage+ image variations that keep core brand assets while generating multiple layout and style alternatives.
- Competitive text-to-image quality, consistently ranking near the top of public text-to-image leaderboards by user votes.
Use cases
Use Cases for Meta Muse Image Text to Image

The Meta Muse Image Text to Image model is particularly valuable for marketing and design workflows that demand both speed and control. For performance marketers, its ad variation capability lets you generate multiple on-brand creatives from a single brief, preserving logos and product placement while exploring new backgrounds and hooks. Example prompt: “Create five variations of this product photo for a feed ad, keep the logo and product fixed, change background color and headline text to test different messages.” For creators and social managers, its Instagram-native aesthetic and flexible aspect ratios help produce story and reel cover art from a short description. Prompt: “Design a vertical story cover with a cinematic photo of a city at night, bold title ‘Midnight Launch’ at the top.” For developers using each::labs, the Meta Muse Image Text to Image API can power template-driven generation for landing pages, thumbnails, and campaign assets directly from structured prompts.
Tips & tricks
Tips and Tricks

Meta Muse Image Text to Image responds best to prompts written like mini creative briefs rather than keyword lists. Clearly describe subject, style, layout, and any required text in the image, and keep constraints explicit. For example: “a square Instagram-style poster, headline at the top, product photo centered, call-to-action button at the bottom.” When using the Meta Muse Image Text to Image API through each::labs, structure prompts so that brand elements and must-keep regions are described first, then variations and style cues. If you are editing or composing from references, mention which parts should stay fixed and which can change. Example prompts:
- “Create a vertical story ad for a summer sale, bright pastel colors, centered sneaker photo, bold readable text ‘SUMMER DROP’ at the top, small body text at the bottom.”
- “Generate a clean product photo on a white background, subtle shadow under the object, include the brand name ‘Aurora Labs’ in a simple sans-serif font under the product.”
- “Blend two reference photos into a single lifestyle scene, keep the original logo intact, match lighting and color grading for a modern Instagram aesthetic.”
Technical spec
Technical Specifications
- Provider: Meta (Meta Superintelligence Labs), Muse family image-generation model.
- Primary modality: Text-to-image, with support for image editing and multi-image composition (text + image in, image out).
- Instruction-following: Uses an “agentic” pipeline that can invoke search and code-like rendering to improve factual grounding, charts, QR codes, and in-image text.
- Aspect ratios: Flexible; optimized for modern social formats and ad placements (e.g., stories, feed, square) drawn from Instagram and Meta surface requirements.
- Resolution: Meta has not publicly fixed a single native resolution; outputs adapt to target placement, with high-quality consumer-facing image size suitable for ads and social creatives.
- Latency: Competitive “frontier-class” text-to-image; arena rankings and platform usage suggest interactive-time generations suitable for chat and creative workflows.
- Access model: Consumer-facing Meta AI surfaces only; no official public Meta Muse Image Text to Image API from Meta as of August 2026.
Things to be aware of
Things to Be Aware Of

Meta Muse Image Text to Image is tightly coupled to Meta’s product surfaces, and Meta has not published a stable first-party API or model card for independent deployment. This means behavior can evolve as Meta tunes the system for safety, branding, and performance, which downstream integrators must track. Availability is also region- and account-dependent in Meta’s consumer apps, and some early features like reference via @-mentions have already changed post-launch. While Muse Image performs strongly on text-to-image leaderboards, users still report occasional artifacts, especially in very small, dense text or extreme edge cases such as heavy compositing with many constraints.
Key considerations
Key Considerations

Meta Muse Image Text to Image is designed first as a consumer feature inside Meta AI, so most official UX assumes chat-like prompts and social surfaces rather than raw API calls. Meta emphasizes instruction-following and brand-safe ad creatives, making this model especially suitable for marketers and designers who need coherent layouts and readable text. However, Meta has not released open weights or a first-party developer API, and access remains tied to Meta’s ecosystem and policies. On each::labs, the Meta Muse Image Text to Image API abstraction is intended for experimental and prototyping use; production deployments should factor in that upstream changes may affect behavior and style defaults over time.
Limitations
Limitations

Meta Muse Image Text to Image currently lacks open weights and an official public Meta Muse Image Text to Image API, limiting direct developer control and fine-tuning options. Access is bound to Meta AI experiences, so usage policies, safety filters, and content restrictions apply and may block certain prompts or styles. The model can struggle with highly technical diagrams, extremely small text, or complex scenes with many overlapping constraints, and there is no publicly documented fixed resolution or deterministic configuration for reproducible research-grade outputs.

Related models

4 models

Alibaba Qwen Image 3.0 Pro Text-to-Image AI model preview

Alibaba Qwen Image 3.0 Pro Text-to-ImageAlibaba

Recraft v4.1 | Text to Vector AI model preview

Recraft v4.1 | Text to Vectorrecraft

Luma | Uni-1 | Text to Image AI model preview

Luma | Uni-1 | Text to ImageLuma

Reve 2.1 Text to ImageReve

* FAQ

About Meta Muse Image Text to Image

01 / 03

What is Meta Muse Image Text-to-Image?

Meta Muse Image Text-to-Image is an image generation model that creates images from text prompts. It is useful for realistic scenes, stylized visuals, editorial concepts, product mockups, and designs that need accurate visual details or readable text.

Meta Muse Image Text to Image