[Feature Request]: Add [filewords] option to prompts.

[kyrieru]

Is there an existing issue for this?

• [X] I have searched the existing issues and checked the recent builds/commits

What would your feature do ?

There is currently no way to take the individual elements of a picture into account during img2img or batch processing, since it does not use caption files.

Feature request is to add the following options to img2img and batch processing:

• Add the option to use [filewords] in prompts, so that the placement of captioned files can be placed anywhere and with any weight.
• And/or automatically interrogate the image before processing, automatically adding it to the existing prompt.

Proposed workflow

1. Go to batch processing. Type a prompt as such, " a person sitting at a table, [filewords] "
2. When batch process runs, it adds the filewords of each image's caption file to the prompt of that image's processing.

Auto interrogate option would be:

1. Enable Deepboru or Clip auto-interrogate checkbox.
2. Type positive prompt as normal and hit generate.
3. image is interrogated before the generation begins, and the new prompt is added to the end of the regular one.

Additional information

I couldn't find this anywhere, but if it already exists, let me know.

[Theophite]

+1 to this one. This would be really great for reprocessing earlier results (or training examples) into the correct style when doing iterative model development -- extremely useful augmentation.

[hcaihao]

+1 to this one.

[tarunabhdutta]

Is this feature implemented yet? Are there any other ways to add corresponding caption text file to image?