Introduction

Here you will learn about AI tools!

Upscale#

Upscaling is required for most photos you find online. From art to cards it all needs to be a high resolution by the time you print. You will need an upscaling software and a safetensors file.

Upscale software#

• chaiNNer Github is a stand-alone upscaling GUI. You can use templates Github to get started.
• ComfyUI Github is a modular diffusion model GUI, api and backend with a graph/nodes interface. You can upscale and remove text with this.

Safetensors file#

OpenModelDB Website has many custom upscaling models. An interactive visual comparison of upscaling models Website is available.

• Real-ESGRAN-x4plus Hugging Face
• 4x-UltraSharpV2 Github by Kim2091 Ko-fi

Remove text#

If you need to translate a card to your native language, you have to remove the original language text from the card. Use a FLUX.1-Kontext-Dev template Website to modify the original card photo, then upscale. You can integrate upscaling directly into the pipeline of text removal using ComfyUI.

Safetensors file#

• FLUX.1-Kontext-Dev Hugging Face - 12GB file size

  Warning
  Requires 16-20GB of VRAM

  • Latest Image-to-Image diffusion model from Black Forest Labs Hugging Face
  • If you want to test this before downloading you can try the live demo Hugging Face
  • Model source available on Github
• Nunchaku-FLUX.1-Kontext-Dev Hugging Face - 7GB file size

  Warning
  Requires 8-12GB of VRAM

  • 4-bit version of FLUX.1-Kontext-Dev from MIT hanlab Website
  • Different safetensors files for users with non-Blackwell GPUs (<=40-series) Hugging Face and for users with Blackwell GPUs (50-series) Hugging Face
  • See ComfyUI-nunchaku Github for ComfyUI setup
  • Model source available on Github

Prompt example#

Remove the “sample” text in the center. Remove the text at the top. Remove all card text at the bottom. Keep the top left numbers. Remove all card text at the bottom.