Kling 3 References, Better Uploads, Model Tools, and Export Fixes: What's New at Uwear

This week was about making Uwear Studio more reliable for real fashion production work: better AI fashion video control, fewer large-image upload failures, cleaner image downloads, improved model setup, and smoother agent workflows.

Kling 3 Pro Now Supports Reference Attachments for Fashion Video

Uwear video generation can now pass visual reference attachments to Kling 3 Pro. For fashion teams, this helps with a common AI video problem: the first frame shows the front of the garment, but the model also needs to understand the back, side, print, strap, pocket, or construction details.

Instead of forcing a strict last frame, you can now use the model image as the first frame and attach the back-view garment reference. That gives Kling 3 more product context while keeping the video motion more natural.

Large Clothing Uploads Are More Reliable

Fashion product images are often large: high-resolution ecommerce photos, heavy PNGs, detailed flat lays, and files exported from editing tools. Browser-side resizing can fail when memory is tight or when a resize library is blocked.

Uwear now handles more of that resize work on the backend. The result should be fewer "image too large" upload errors, fewer failed clothing item uploads, and a better path for teams uploading production-quality garment images.

Image Export and Download Fixes

We tightened several download details across image exports. PNG ZIP exports now use the expected format, single-image downloads behave more consistently, and export menus layer correctly over the gallery interface.

These are small fixes, but they matter when AI-generated product photos need to move into ecommerce pages, ad creative, social posts, creative review folders, or downstream editing tools.

Mouza Model Tools and Demo Avatars Are Easier to Use

Mouza can now help create reusable avatar/model candidates from a text description and save a selected generated image into the model library. It is a small but useful step toward faster AI model setup for consistent fashion photoshoots.

Uwear now has stronger demo avatar/model presets in Studio. Demo models appear more predictably in the model grid, can open into model details, and are easier to use as a starting point for teams testing AI model photoshoots.

This helps new users understand how consistent AI models work before they upload or generate their own reusable model assets.

Agent Chat, Mobile Layout, and Edit Context Polish

Mouza and the Studio interface also got a set of practical workflow fixes. Agent chat navigation is cleaner on mobile, visual panels have more reliable spacing, generation results refresh more simply, and edit generations inherit avatar context more consistently from the parent image.

In plain terms: when you edit an AI fashion image, generate another result, or move through the agent workflow on a smaller screen, Uwear should feel less brittle and easier to keep working in.

Media Thumbnails and Submission Status Are More Stable

We also fixed video thumbnail handling in media ingestion and improved how submitted generations move into the created state. These reliability improvements reduce the awkward cases where a generated video or image exists, but the gallery preview or generation status lags behind.