- Home
- Image Upscaling Software
- DiffBIR
On the recordRead 12 Oct 202619 facts
DiffBIR
DiffBIR · China · founded 2023
Plans and prices
DiffBIR
Free
Open-source research code and pretrained weights; GPU memory can limit inference
Read 12 Oct 2026 · github.com
On the record
- Additional components
- The project says it is based on ControlNet and BasicSR and uses Stable Diffusion v2.1 for training stage two.github.com · read 12 Oct 2026
- Apple Silicon
- The project announced Apple Silicon support and refers users to instructions for CPU, CUDA, and MPS devices.github.com · read 12 Oct 2026
- Batch processing
- Yesgithub.com · read 22 Sep 2026
- Deployment
- The instructions clone the repository, create a Python 3.10 Conda environment, install requirements, and run inference locally.github.com · read 12 Oct 2026
- License
- The project is released under the Apache 2.0 license.github.com · read 12 Oct 2026
- Maximum scale
- 8xgithub.com · read 22 Sep 2026
- Model downloads
- Pretrained weights are linked from Hugging Face, Baidu Netdisk, and OpenXLab, and the README says inference weights download automatically.github.com · read 12 Oct 2026
- Model integrations
- Inference uses BSRNet for super-resolution, SwinIR-Face for face restoration, and SCUNet-PSNR for denoising as stage-one models.github.com · read 12 Oct 2026
- Online demo
- The README announces an online demo on OpenXLab for the general and face models.github.com · read 12 Oct 2026
- Purpose
- DiffBIR is a blind image restoration pipeline that uses a generative diffusion prior.github.com · read 12 Oct 2026
- Quality trade-off
- The README says using the condition as the sampling start point can improve stability but may decrease sample quality.github.com · read 12 Oct 2026
- Restoration tasks
- The repository provides inference examples for blind image super-resolution, aligned and unaligned face restoration, and blind image denoising.github.com · read 12 Oct 2026
- Support
- The README directs questions to [email protected].github.com · read 12 Oct 2026
- Tiled sampling
- Tiled sampling supports super-resolution at large scale factors on low-VRAM graphics cards.github.com · read 12 Oct 2026
- Training limitation
- The README states that the stage-two training script does not support a validation set.github.com · read 12 Oct 2026
- Upscaling method
- aigithub.com · read 22 Sep 2026
- Versions
- The README lists DiffBIR v2 and v2.1, and says v2.1 includes a model trained on Unsplash captions generated with LLaVA, more samplers, and improved tiled-sampling support.github.com · read 12 Oct 2026
