MinDalle_StableDiff/README.md

1.6 KiB

MinDalle_StableDiff

Create pictures with Mindalle and stable diffusion. Python script with dependencies installation bat file.

How to set up dalle mini + stable diffusion

  1. Run "get_dependencies.bat";

    • if you see connections errors, just run the script again to be shure everything is installed;
  2. Run "login.bat" to login to your HuggingFace profile;

    2.1 Go to https://huggingface.co/CompVis/stable-diffusion-v1-4 and make sure to accept the terms.

    2.2 Go to Your Profile > Settings > Access Tokens;

    2.3 Create a write token/Copy an existing Token key and enter in the cmd window. - It can be done once and forever untill you delete token from C:\Users\%YOUR_USERNAME%\.huggingface

  3. You are redy to go.

How to run

  1. Clone this repo;
  2. Write your promt to "promt.txt";
  3. Run "run_all.bat";
    • First run will be slow due to downloading weights
  4. When amount is asked, enter any number you want;
    • Quantity of output pictures = amount * 60
  5. Just wait. It can be a long process;
  6. You can find output files in .\data\out\%Timestamp% folder. Jpeg contain commentaries in metadata with some settings used while generating.

How to update

Small update

  1. Get new scripts from this repo;
  2. Run "rem_dependencies.bat";
  3. Run "get_dependencies.bat".

Big update

  1. Delete old files;
    • Don't forget to save your fenerated pictures
  2. Clone this repo from scratch;
  3. Run "get_dependencies.bat".

Requirements

  • 10GB of free space.
  • Nvidia card with 4GB+ video memory.

Tests

  • RTX3070: 12 - 14 seconds per picture.
  • RTX2060_Laptop: 21 - 25 seconds per picture.