This repository hosts adapter checkpoints for TADiSR: Text-Aware Real-World Image Super-Resolution via Diffusion Model with Joint Segmentation Decoders. Each checkpoint contains LoRA parameters for the diffusion backbone, VAE skip adapters, and the jointly trained text-segmentation decoder. It is not a standalone base model.

The CogView4 release is recommended for Chinese text. See the code repository for installation, checkpoint validation, and inference.

The adapter checkpoints are released for research use. They remain subject to the terms of their respective base models, including the commercial-use terms of Kolors. Do not treat this model card as a replacement for the base model licenses.