HiAPI
21:9 高调光晕宽幅:左侧留白,右侧悬浮半透明玻璃渲染面板与光带轨道,冷蓝、珊瑚粉与暖金渐变,模型名以玻璃蚀刻发光立体字呈现。

Meta

Muse Image 1.0

muse-image-1.0/image-to-image

Meta's reasoning-based image generation and editing model that first breaks down complex prompts before rendering, can fuse multiple reference images, and leverages web retrieval to ensure factual accuracy

Try muse-image-1.0/image-to-image

Input

Output

image
Fill in the parameters and run — results show up here.

Pricing

muse-image-1.0/image-to-image per task

Unit Price $0.015

Overview

muse-image-1.0/image-to-image is a reference-image-driven image generation and editing model released by Meta, accessible through the unified HiAPI.ai interface. It takes one or more reference images as input, works with text prompts to perform image generation, element replacement, and style rewriting, and breaks down multi-part complex instructions with layout planning before rendering. It is suitable for developers who need to use existing assets as base images and pursue prompt adherence and batch output efficiency, and can be used in workflows such as asset variants, product scene images, and personalized compositing.

Key Highlights

  • Reason-then-render generation workflow — before rendering, it breaks down multi-part prompts and plans the layout, reducing the chance that complex instructions are dropped in a single forward generation.
  • Generation and editing happen within the same model — reference image fusion and instruction-based rewriting require no extra editing toolchain, reducing the number of calls.
  • Multiple reference images can be submitted at once — accepts up to 10 reference images, and you can specify the subject, style, or composition role for each image in the prompt.
  • Output specifications controlled directly by parameters — you can specify the aspect ratio and the jpeg, png, or webp output format, making it easy to connect to downstream asset pipelines.

Capabilities

Reference-image-driven image editing: takes reference images plus text prompts as input, modifies or preserves the content in the images according to the prompt, suitable for targeted adjustments on existing assets rather than generating from scratch.

Multi-image fusion: a single request can carry 1 to 10 reference images, with the subject, style, or composition role labeled for each image in the prompt, for compositing tasks.

Decomposed execution of complex prompts: plans multi-part descriptions before rendering, making prompts that contain multiple subjects and layout requirements easier to fully implement.

Output specification control: supports nine aspect ratios from 1:1 to 21:9, and lets you choose jpeg, png, or webp format to output directly according to delivery or storage needs.

Use Cases

  • Advertising and media teams: batch-generate variants from existing assets — upload brand base images and rewrite the visuals and ratios according to channel prompts, producing multiple sets of ad creatives for A/B testing and multi-channel distribution.
  • E-commerce operations: contextual transformation of product images — use real product photos as references, prompt to replace backgrounds and scenes, and output catalog images or scene images for product detail pages.
  • Personalized compositing: blend user photos into new scenes — use user-uploaded photos as reference images to composite themed scenes, producing customized images for campaign pages or social sharing.

Limitations and Selection Advice

Public materials do not disclose technical details such as model architecture and output resolution, so generation results need to be evaluated based on actual performance. Multi-step reasoning introduces additional latency and is not suitable for scenarios extremely sensitive to response latency; output is limited to images and does not involve video or audio. If the task only requires text-to-image generation from scratch without reference images, the text-to-image line in the same family can be used; if video generation or lower latency is required, other models should be considered.

Frequently asked questions

View all FAQs →
  • How do I choose generation or editing?

    Choose generation to start from a written brief, or editing with references to build on an existing subject or layout.

  • How many images can I get per task?

    Each task returns one image with no count selector. Submit a new task with an adjusted prompt for another composition.

  • Can I select 1K, 2K or 4K?

    This entry has no resolution tiers. Choose framing; the model determines dimensions, typically around 2.5MP. Inspect the downloaded file.

  • Which file formats can I choose?

    Choose JPEG, PNG or WebP; the default is WebP. Select a format for your web or design workflow.

Related Models

View all models →

Same Type