FASHN Tryon v1.6

Positioning Overview

fashn-tryon-v1.6 is a virtual try-on image model for apparel e-commerce and visual content production: it takes a person photo and a garment product image as input, and outputs a composite image of the model wearing that garment. Compared with the previous generation, native output resolution is increased to 864x1296, with clearer reproduction of fabric textures, patterns, text, and other details, and more natural lighting. It is suitable for developers and e-commerce teams that need to produce on-model effect images in bulk for product detail pages, marketing materials, or design validation.

Core Highlights

• Native 864x1296 resolution output — about a 50% increase over the previous generation, making fabric textures, patterns, and text details clearer and usable at detail-page sizes.
• Compatible with both model images and flat-lay garment materials — specify model, flat-lay, or auto via garment_photo_type for automatic recognition, adapting to different shooting conditions.
• Three quality modes to choose from as needed — performance, balanced, and quality switch between generation speed and rendering precision, making tiered calls convenient in production environments.
• Fixed seed for reproducible results — passing the same seed with the same input yields consistent output, facilitating result reproduction and troubleshooting in batch production.

Capability Description

Try-on composition of person and garment images: submit two URLs for the person image and garment image to get an on-model effect image, output as png or jpeg, which can go directly into product image or material workflows.

Adaptation to garment categories and shooting formats: category supports tops, bottoms, one-pieces, and auto for automatic recognition; garment_photo_type distinguishes on-model garment images from flat-lay images, reducing preprocessing work.

Output quality control: mode selects speed priority or precision priority; moderation_level can block explicit content according to business requirements, and the conservative level also blocks lingerie and swimwear garments.

Reproducible batch generation: with a fixed seed, the same input can reproduce results, making it convenient to locate issues and rerun failed samples in production pipelines.

Applicable Scenarios

• Batch production of apparel product on-model images for e-commerce teams — batch-generate on-model effect images from product flat-lay images plus model images, for store detail-page main images and style display.
• Visual validation during apparel design sampling — quickly generate images using design drafts or sample garment photos together with model photos, to confirm fit and pattern presentation before investing in actual photography.
• Rapid image generation for multiple styles of marketing materials — generate images separately for the same model paired with multiple styles, for producing multi-variant materials such as campaign pages and social media placements.
• Converting flat product images into on-model display — convert existing flat-lay or invisible-mannequin shooting materials into on-model wearing effects, revitalizing the existing product image library.

Limitations and Selection Advice

This model targets full-body apparel display scenarios at a fixed resolution, and the output size cannot be adjusted arbitrarily. In quality mode, rendering precision is higher but takes longer; batch tasks need to allocate tiers between speed and effect. Generating multiple results per request is a common practice to improve usability, but it correspondingly increases billing. The conservative moderation level blocks lingerie and swimwear garments; when such items are involved, a looser level should be selected. If the business values long-running production stability more than precision, it is necessary to evaluate whether to use a more mature previous-generation version.