AUTOMATIC1111 / stable-diffusion-webui-promptgen

stable-diffusion-webui-promptgen
MIT License
484 stars 93 forks source link

Prompt generator

An extension for webui that lets you generate prompts.

Getting more models online

You can add models from huggingface to the selection of models in setting. The setting field is Hugginface model names for promptgen, separated by comma, and its default value is just:

AUTOMATIC/promptgen-lexart, AUTOMATIC/promptgen-majinai-safe, AUTOMATIC/promptgen-majinai-unsafe

Those are GPT2 finetunes I did on various datasets:

To add more models, add more items to the setting, for example like this:

AUTOMATIC/promptgen-lexart, AUTOMATIC/promptgen-majinai-safe, AUTOMATIC/promptgen-majinai-unsafe, succinctly/text2image-prompt-generator, microsoft/Promptist

Reloading UI is required to apply this setting.

Getting more models offline

Put your models into the models directort inside the extension. For example, my model can be made fully offline by placing fines from https://huggingface.co/AUTOMATIC/promptgen-lexart/tree/main into those directories:


 πŸ“ webui root directory
 ┗━━ πŸ“ extensions
     ┗━━ πŸ“ stable-diffusion-webui-promptgen
         ┗━━ πŸ“ models                            
             ┗━━ πŸ“ promptgen-lexart              <----- any name can be used
                 ┣━━ πŸ“„ config.json               <----- each model has its own set of required files;
                 ┣━━ πŸ“„ merges.txt                       mine requires all those
                 ┣━━ πŸ“„ pytorch_model.bin
                 ┣━━ πŸ“„ tokenizer_config.json
                 ┗━━ πŸ“„ vocab.json

Reloading UI is required to see new models.