SUPIR
Research model that restores very damaged photos with rich detail.
FreeOpenRuns on your PC
- Best for
- Restoring badly damaged photos, locally
- On your PC
- Technical
- Works on
- Windows, Linux
- License
- Non-commercial
What’s good
- Rebuilds detail from very low quality photos
- Follows text hints about what the photo shows
Downsides
- Non-commercial license; business use needs written permission
- Needs a 12 GB NVIDIA card even in low-memory mode
- Python setup; slow, and SeedVR2 is now often better
At a glance.
- Best for
- Restoring badly damaged photos, locally
- Price
- Free, open source
- Version
- SUPIR v0Released 2024-01
- Works on
- Windows, Linux
- Open source
- Yes, Non-commercial
- Made by
- Fanghua Yu, Jinjin Gu and others
- Website
- github.com
Should you use it?
It is #9 of 10. Good for restoring badly damaged photos, locally; for most people, Topaz Gigapixel is the better start.
Can I run it myself?
Yes. Technical: command line or python.
Use Python (Gradio demo), or community ComfyUI nodes.
- You need
- NVIDIA card with 12 GB in low-memory mode, slower
- Setup
- TechnicalCommand line or Python
- Laptop, 8 GBToo small
- PC, 16 GB+Too small
- Mac, 16 GBToo small
- Mac, 32 GBToo small
- Mac, 64 GB+Too small
- PC, 8 GB GPUToo small
- PC, 12 GB GPURuns
- PC, 16 GB GPURuns
- PC, 24 GB GPURuns
- WorkstationRuns