Changelog

You can now use a Background Reference image in Product to Model to anchor your generations to a specific location or setting.

The provided image is used as the backdrop, ensuring consistent environments across campaigns. If a person appears in the reference image, they will be ignored — only the background is used.

Generated models now interact naturally with the scene, adapting to lighting, perspective, and physical elements like ledges or furniture to produce realistic, integrated compositions.

This update brings several major improvements across image and video generation tools:

4× Image Upscale

Image Upscale now supports 4× enlargement, delivering true 4K-level detail with reliable results.

While 2× remains recommended for the most natural and stable outcomes, 4× offers an excellent option when maximum resolution is desired.

Face Reference in Product to Model

You can now apply Face Reference directly in Product to Model, enabling complete identity consistency from product photo to final model showcase.

This expands full identity control across Product to Model, Model Swap, Model Creation, and Edit — ensuring the same face can appear seamlessly across your entire workflow.

Prompt Support in Video Creation

You can now add short text prompts to guide motion or mood in video generation.

For best realism and aesthetics, we still recommend using no prompt, but this option gives added flexibility when specific direction is required.

This update introduces several improvements across the app, focused on identity control, creative flexibility, and a cleaner interface.

Face Swap in Edit

You can now swap or restore faces directly within Edit. This helps recover likeness after pose or lighting adjustments, or replace identities while maintaining consistent skin tone and realism.

Currently available for Agency plan users, with broader access planned soon.

Improved Model Creation Pipeline

The Model Create tool is significantly upgraded with a new internal generation system, delivering better aesthetics, prompt accuracy, and reference alignment by default.

Face Reference in Model Creation

The Model Create tool now also supports Face Reference, enabling you to generate new images of an existing model with full control over identity, pose, and composition. You can combine prompt + image reference + face reference for complete creative precision.

Refreshed UI Navigation

We’ve refined the action navbars across the app with new popovers and a cleaner, minimal design. These updates prepare the interface for upcoming features and improve usability.

Transition from Consistent Models

The Consistent Models feature is being phased out in favor of Face Reference, which offers superior consistency and flexibility without fine-tuning. You can still find Consistent Models under the 🧬 icon for now.

Video creation at 480p has been significantly upgraded. It now renders faster, in around 30 seconds, while delivering higher fidelity, smoother motion, and fewer distortions in faces or garments.

These improvements bring more natural, stable visuals at lower resolutions and are now live for both Basic plan users in the app and API users with the parameter resolution: 480p.

Model Swap now more reliably changes identities, even when the prompt closely resembles the current subject (for example, “a Brazilian woman”).

This update ensures swaps occur as intended, even with similar or missing prompts, while improving attention to extremities such as hands and feet. These areas now adjust more accurately in age or ethnicity swaps, reflecting realistic variations in skin tone and texture across the entire body.

The Image Upscaler now uses a streamlined and more reliable 2× enhancement option.

This update replaces the previous “2K” and “4K” modes with a smarter scaling system that:

• Delivers sharper, more natural texture improvements.

• Reduces distortion and artifacting in detailed product areas.

• Provides more consistent results across diverse product types and source qualities.

The new approach ensures that upscaling focuses on clarity, realism, and product detail preservation rather than arbitrary resolution labels, resulting in cleaner, more dependable upscaled images across the board.

The 720p option in Image to Video now delivers significantly enhanced visuals and motion quality.

This update brings the 720p output much closer to 1080p results, with:

• More natural model and camera movement for smoother, more lifelike motion.

• Improved facial identity preservation, ensuring models remain consistent across frames.

• Better product detail fidelity, maintaining sharper textures and realistic lighting.

• Refined aesthetic quality, matching the visual polish of our higher resolutions.

These improvements apply across both the app and API versions of Image to Video, providing a better balance between visual quality and generation cost.

Model Swap now includes Match options for finer control over how reference faces influence the final image.

Match Reference (default)

Adopts the head angle, gaze, and expression from the reference image for maximum resemblance. Best when you want the new model to fully follow the look and direction of your face reference.

Match Current Model

Keeps the head angle, gaze, and expression from the original photo while still applying the reference identity. Ideal for preserving the current pose while changing only the face.

Try it now → https://app.fashn.ai/studio?model=model-swap

Generate longer videos from your images, with options for 5-second or 10-second durations.

A new endpoint that turns a single image into a short video with motion tailored for fashion products. Motion is planned automatically based on product type (for example, jewelry, dress, bag) with tasteful camera work and model movement. For close-up product shots, it emphasizes professional camera motion rather than model movement. No prompt is required; optional guidance is supported but should be minimal.

• Duration: 5s or 10s

• Resolutions: 480p, 720p, 1080p

• Inputs: image (required), prompt (optional and minimal)

• Output: one MP4 per request

• Credits: 480p=1, 720p=3, 1080p=6; 10s applies a 2× multiplier

API Reference → https://docs.fashn.ai/api-reference/image-to-video