TeleStyle V2

Summary

TeleStyle V2 is a content-preserving style transfer and image-editing model built with self-distillation and distribution-matching distillation. It extends TeleStyle V1 beyond the original realistic-content and stylized-reference setting to handle realistic and stylized references in both content and style roles.

Key Features

Supports content-preserving style transfer with flexible reference types.

Handles realistic-realistic, realistic-stylized, stylized-realistic, and stylized-stylized settings.

Uses self-distillation to construct training triplets from TeleStyle V1.

Uses Distribution Matching Distillation to reduce inference cost.

Preserves broader text-guided image editing ability from Qwen-Image-Edit.

Uses Qwen2.5-VL-7B to generate content and style prompts.

Supports optional numbers of reference images for style or content workflows.

Provides model, code, and Hugging Face demo links.

About TeleStyle V2

The model constructs self-distilled triplets with TeleStyle V1, then uses Distribution Matching Distillation to reduce inference cost while preserving Qwen-Image-Edit's general editing ability. It also uses Qwen2.5-VL-7B to generate content and style prompts to reduce content/style reference order confusion.

TeleStyle V2 is useful for image-editing applications where content and style references may be realistic, stylized, or mixed. The project links to model weights, code, and a Hugging Face demo for experimentation.

Get more likes & reach the top of search results by adding this button on your site!

Subscribe to the AI Search Newsletter

Get top updates in AI to your inbox every weekend. It's free!