Changelog

We’ve released a patch to address a few issues reported by our users:

• Fixed a skin tone mismatch that could occur when switching from a long sleeve top to a short sleeve top, where the wrist area didn’t seamlessly blend with the newly generated skin.

• Resolved background discontinuities that were sometimes visible with certain background types.

Building on our May 3, 2025 update, we’ve further refined our algorithms to improve identity preservation. This update brings better retention of individual features such as skin tone, freckles, tattoos, and other distinguishing details — as well as improved handling of original, unswapped garments.

We’ve added two new parameters to give you more control over image delivery:

• output_format: Choose between 'png' (highest quality) and 'jpeg' (faster, optimized for real-time use cases like virtual fitting rooms). Default: 'png'.

• return_base64: When set to true, the API returns a base64-encoded image string instead of a CDN URL — ideal for fully private, ephemeral use. Default: 'false'.

Check out the updated API docs for usage details.

We have significantly enhanced the 'Quality' mode, focusing on preserving skin details and original garment integrity. This update delivers noticeable improvements in the following scenarios:

1. Preservation of Unique Skin Characteristics:

   • Improved retention of original model features such as tattoos, freckles, wrinkles, and other distinctive skin attributes.

2. Enhanced Garment Reconstruction:

   • Superior reconstruction when significantly altering garments. For example, changing a long top into a crop top now results in better imagining of underlying garments, ensuring realistic and consistent results.

3. Improved Bottom Garment Retention:

   • Enhanced handling of minimal bottom garments (e.g., underwear or panties) to prevent unintended conversion into shorts. Minimal garments are now preserved accurately during top garment changes.

4. Reduced Distortion in Non-swapped Regions:

   • Major improvement in maintaining original garment details when changing bottom garments. The top garments now experience significantly reduced distortion, preserving the original appearance and details

You can now select multiple images in your gallery to delete them all at once. Deleting generated videos is also supported.

1MP outputs and flexible input aspect ratios are now supported in both the API and Web App

You can now select a model directly from your saved models gallery when using the Try-On Studio. A “My Models” button will appear under the “Select Model” section if you have previously saved models.

FASHN Try-On now supports higher output resolutions and more flexible input image dimensions, available today in the nightly endpoint and 🧪 Experimental web app mode.

- Outputs are now larger (up to 1MP).

- No more strict 2:3 aspect ratio constraint.

- No changes needed to your API calls.

- No impact on runtime performance.

Planned rollout to all users: Wednesday, April 30, 2025.

👉 Read the full announcement

The Models Studio and AI model generation functionality in the Try-On Studio have been significantly improved, resulting in more realistic-looking models, higher overall quality, and better prompt adherence.

Web App:

The new FASHN v1.5 model is now the default model. The "Experimental" toggle is removed.

API:

The new FASHN v1.5 model is now the default model in the standard https://api.fashn.ai/v1/ endpoint. The /nightly endpoint now simply reroutes to the main endpoint. See updated docs.

We're excited to announce that our new v1.5 model is now available for testing!

Building on our recent Segmentation-Free Mode preview, this version represents a significant leap forward, bringing numerous improvements and simplifications.

What's New in v1.5:

• Smarter fits: Enhanced capabilities for more challenging fits, preserving body shape, skin texture, tattoos, and identity with much higher resemblance.

• Simplified Controls: We've removed the need for parameters like Restore Background, Restore Clothes, Long Top, Adjust Hands, and Cover Feet. Our new model handles these complexities seamlessly.

• Performance Boost: Experience double the speed! Approximate runtimes:

  • Quality: ~12 seconds

  • Balanced: ~6 seconds

  • Performance: ~4 seconds

• Segmentation-Free by Default: The new "Segmentation-Free" parameter is enabled by default, streamlining the garment-fitting process.

Turn it off by sending segmentation_free: false only if you encounter issues when removing original garments.

*API Users: This model is still in testing phase and is subject to changes, so we don’t recommend using it in production just yet.

How to Access:

• Web App Users:

  Activate the v1.5 model by selecting the "Experimental" toggle.

• API Users: Access via the /nightly endpoint. Learn more about /nightly →

Please note: You may experience occasional longer wait times as we fine-tune our hardware based on usage patterns.

API and Models Studio users can use their saved models by sending saved:<model_name> in the model_image field instead of a URL or base64. API parameters documentation →

Users can now save and reuse AI-generated models in Models Studio for try-ons.

Following our recent upgrade for automatic recognition of model vs. flat-lay garment photos, we've now introduced Automatic Garment Category Classification.

No more manually specifying garment categories (tops, bottoms, or one-pieces). You can now simply send category: auto

Here's how it works:

Flat-Lay or Ghost Mannequin Photos: It's simple—we instantly recognize the garment category.

On-Model Photos:

Full-Body Shots: We'll default to swapping the entire outfit. If you only want the top or bottom swapped, you'll still need to specify manually (sorry, we're good, but not psychic yet).

Focused Shots (upper or lower body): We will identify if the focus is on the top or bottom—even if both are visible—saving you the extra step.

Users can set automatic credit top-ups via the API billing page.