
Promptor - fully automated AI rendering
=======================================

Full version
============

The full version package contains promptor.exe, promptdb.txt
and this readme.

promptor.exe loads promptdb.txt, 
a text file which can be edited with plain ASCCI editors
like Notepad, Notepad++ or Depeche View Lite.

the file contains a syntax like

   $prompt: photo style, $subject

   $negativeprompt: nude, deformed, bad anatomy, bad proportion,
      disfigured, mutation, mutated, ugly, ...

   $subject: $woman; $man; $pair;
      $environment, afternoon light
   
   $woman: a 22 year old $eth girl,
      beautiful face, slim nose, beautiful body,
      $pose, (eye level, full shot)+.
      rough freckled skin, visible feet.
   
   $eth: $eth1 $eth2
   
   $eth1: swiss;german;norwegian
   
   $eth2: asian;nerdy;german
   
   $pose: standing;kneeling;squatting;sitting;
      lying;hero pose

This means: the symbol $prompt should produce a text starting 
with "photo style, " then add content produced by symbol $subject.

the symbol $subject should produce a text
- either from symbol $woman
- OR from symbol $man
- OR from symbol $pair
- OR from symbol $environment folled by ", afternoon light".

so ";" means to do a random selection from the text given before and after ";".

If you want to see what text is produced without starting an immediate render, type

   promptor -sim -maxpic=3

to create three examples. The negative prompt is always the same, static text.

If you type "promptor -h" you get a short help text with all available options.

While promptor runs, it will read promptdb.txt again for every new image
to be created. This means you can do live changes to the file which will
be applied immediately on the next image, without the need to restart
Promptor everytime.

Please note: although great care has been taken to avoid keywords
   within promptdb.txt that may trigger nsfw outputs, it cannot be 
   fully avoided that an unwanted nsfw image is rendered
   from time to time. therefore do not simply run a render of 
   100 images and present the output as is to your boss or family,
   but do a manual filtering first.

Full Version options overview
=============================

-db=x        load db file x instead of the default promptdb.txt
-maxpic=n    render up to n images. default is to render endless.
-sim         do not render, just simulate generated prompt output.
-todir=d     write images to folder d. default is 'rawpic'.
-saveprompt  also save a .txt file parallel to image with prompt text.
             note that prompt data may be included in the image file itself.
-postfix=x   add x to image filenames
-server=ip   render on a different PC
-verbose     tell current settings per image
-verbose=2   tell full request data sent
-quiet       don't tell prompt text
-quiet=2     don't tell render parms

-steps=n         default is 30
-cfgscale=n      default is 8
-width=n         default is 512
-height=n        default is 512
-restorefaces=n  default is 1 (=yes), else 0
-seed=n          default is -1


Promptor Test
=============

The test package should only make sure that:

-  the Stable Diffusion Web API is working correctly on your PC.
   Make sure that your webui-user.bat contains at least:
   
      set COMMANDLINE_ARGS=--api

   recommended is:
   
      set COMMANDLINE_ARGS=--api --xformers

-  your graphics card can produce a 512x512 image.

-  writing of an output image file works.

To do so, promptor-test.exe uses a static prompt text which never changes,
promptdb.txt is not included. It does an internal render of a 512x512 image,
but scales the written output image down to 256x256 (demo limitation).

The full version of Promptor writes PNG images with 512x512 pixel,
which can then be upscaled by a 3rd party tool like Upscayl to 2048x2048.

