What's new at FASHN AI

The Reframe tool has received a major upgrade.

Full 4K Resolution Support

Reframe now preserves high resolution outputs, allowing reframing up to 4K resolution. Reframing at 2K or 4K costs 2 credits (instead of 1).

Smarter Reframing Decisions

Reframe is now context aware and can intelligently decide whether an image should be cropped or expanded using generative fill, based on the content and the requested aspect ratio.

For example, when converting a wide 16:9 image into a 1:1 format for social media, Reframe can focus on the main subject and create a tighter composition, instead of expanding the image and introducing unnecessary background.

We've introduced a major update to the Gallery to make it clearer, more powerful, and easier to understand how each result was created.

Cleaner Gallery Overview

Gallery items now use a simplified hover UI that clearly shows which tool generated the result, while removing unnecessary or hidden actions from the preview state.

Full-Featured Sidebar on Selection

When you click on an image or video, a new sidebar opens with all available actions and context in one place. From here you can:

• Zoom in and pan across the image
• Export results for web use
• Send outputs directly to other tools (Edit, Model Swap, Product to Model, Try On)
• Create videos from images
• Upscale images
• View all inputs used to generate the result, including image references

What's Coming Next

This is the first step in a broader Gallery overhaul. Upcoming updates will further unify Models and Faces into the same view and introduce stronger tools for tagging, organizing, and managing your Gallery results.

This update includes several improvements and fixes across the platform:

• Added support for HEIC and HEIF image formats
• The API Requests tab now displays all input images for every endpoint
• Fixed video generation issues for images narrower than a 9:16 aspect ratio
• Resolved cases where Creative 4K generations could get stuck for extended periods

You can now import images directly from your device into the Gallery.

This allows you to run features such as image upscaling or image to video on your own assets, even if they were not originally generated with FASHN. In future updates, imported images will also be easier to use across tools like Try On and Face Reference.

A new Enhance option is now available in the Gallery alongside the existing 2× and 4× upscaling modes.

Enhance improves image quality by adding detail, refining textures, and sharpening edges without changing the image resolution. This makes it especially useful for images that are already at 4K, where further upscaling is not possible but additional clarity and fidelity are still desired.

Use Enhance when your image has sufficient resolution but needs higher visual quality rather than a larger output size.

The “+ Add Model” option has been removed from Product to Model. This use case is now fully and more effectively handled by the Try On tool.

Product to Model is optimized for generating high quality images when starting from a product photo and creating an AI generated model to showcase it.

Try On is designed for placing products onto a specific existing model photo, whether the product is standalone or already worn by another person.

To reduce confusion and guide users toward the best results, this option has been removed.

If you want to showcase products on specific model images, Try On is the recommended tool.

Background changes in Edit are now handled automatically. The manual Background Change toggle has been removed, and the optimized background change model is now triggered whenever a background change is detected in your edit prompt.

Tip: Background Change also works with Image Context, allowing you to specify a target background using a reference image for more precise results.

New Input Parameters: image_context and mask

The Edit API endpoint has been expanded with additional capabilities. Alongside the newly introduced Image Context support, the endpoint now also accepts a mask parameter, enabling the same Brush Edit functionality available in the app.

The mask can be combined with Image Context to perform precise, localized edits guided by a reference image. This allows targeted changes such as placing logos, modifying buckles, or applying embroidery with greater accuracy.

New Runtime Error Type: UnavailableError

We've also improved error handling across the API and introduced a new UnavailableError. This error explicitly indicates temporary service unavailability or overload, making it easier to implement appropriate handling strategies such as retries with backoff or fallback workflows.

For more details see our API documentation:

• Edit API reference: https://docs.fashn.ai/api-reference/edit
• Error handling: https://docs.fashn.ai/api-overview/error-handling

Face Reference, our core feature for maintaining model consistency across Product to Model, Model Swap, and Model Creation, is now available to all plans.

Previously limited to Agency users, Face Reference can now be used by everyone through a curated set of pre-created realistic AI faces designed by the FASHN team. The initial release includes 4 faces (2 male, 2 female), with more planned in future updates.

This allows all users to experience model consistency workflows and understand how Face Reference works, without needing to upgrade to Agency just to try this feature.

Agency users retain full access to advanced capabilities, including creating custom AI faces and uploading their own images to apply unique identities to photoshoots.

The Edit tool and API endpoint now support Image Context, an optional reference image that provides visual guidance when text instructions alone are not sufficient.

Image Context can be used to:

• Apply a specific background from a reference image

• Mimic complex poses or viewpoints

• Transfer patterns, textures, or color characteristics

For best results, Balanced 4K and Creative 4K are recommended when using Image Context, as they handle visual reference guidance more reliably. Precise 1K remains effective for common tasks such as background or pose changes, but may be less reliable for detailed texture or material transfers.

This addition reduces the need for complex prompting and enables faster, more predictable editing workflows across both the app and API.