My Private Local Waifu Diffusion colab

Feel free to use and edit to your liking!

▶️ ◀️

  • ~5 minutes startup time guaranteed.
  • Booru tags suggestions included. (configurable)
  • Supports hypernetworks (w/o fast-stable-diffusion)
  • 3 server tunnels alternative to Gradio Server (No more big images causing stuck, hopefully)
  • Output to google drive. Make your output permanent, no more anxiety from unknown colab's time limit. (imagine using a cell to zip lol)
  • Load models from your drive, direct links, drive folders, drive links, and public/private hugging face.
  • Batch count max. 200 (instead of 100)
  • Custom UI config which added useful settings to quick settings, add a lot of negative prompts, default prompt, and much bigger prompt textarea
  • UmiAI to help you generate character with random prompts
  • XFormers to help increasing performance by 1.2x
  • Easily upload your merged model to hugging face
  • Waifu Diffusion + VAE
  • Stable Diffusion V1.5 + VAE
  • Trinart + VAE
  • Stable Diffusion Inpainting
  • H Diffusion V17
  • SD v1.4
  • Anything v3 + VAE
  • Elysium Anime V2 + WD VAE
  • Gyokai


  • I will not include NovelAI to my colab due to legal reason.
  • I'm no longer using fast-stable-diffusion as base because it have a very "hacky" approach to increase performance, thus making me relying on their code, AND need him to fix everytime A1111 breaks something which can take days.


Dead-simple. One-cell. Stable Diffusion Automatic1111 Webui Colab.

Model DL Colab

Download models from torrents and drive in colab then distribute them to drive and hugging face



My HF Space for img to booru tags

WD + FD support + fast-stable-diffusion + localtunnel + output to drive colab:

WD local + fast-stable-diffusion colab:

WD + fast-stable-diffusion + drive as storage colab:


comparison 1boy, spiky hair, blue cape, blue eyes, light particles, dark background, hair over one eye. img2img Inpainted. WD pre1.3 epoch 8

Pub: 11 Oct 2022 11:27 UTC
Edit: 09 Dec 2022 09:52 UTC
Views: 6873