Skip to content
/ sd_batch_runner Public

This is an application for batch execution of AUTOMATIC1111 UI launched on a local PC via api. (This is NOT sdwebui extension.)

0 stars 0 forks Branches Tags Activity
Star
Notifications You must be signed in to change notification settings

s9roll7/sd_batch_runner

BranchesTags

Repository files navigation

SD Batch Runner

This is an application for batch execution of Stable Diffusion Web UI(A1111) launched on a local PC via api.
(This is NOT sdwebui extension.)

It is easy to do the following.

  • Image generation by simply selecting lora without manually setting trigger word or lbw
  • Batch generation of images with only the character or style changed with specific settings
  • Easy to try multiple lora combinations. (ex. Generate a combination of a specific style and all character loras)
  • Complex image generation procedures, such as those involving controlnets or consecutive txt2img,img2img, can be saved and reused at a later date.
    (ex. After generating txt2img once, use it as an input image and execute img2img 3 times with slightly different conditions, and execute this procedure 20 times, changing the character and style each time. )

Installation(for windows)

Python 3.10 and git client must be installed

git clone https://github.com/s9roll7/sd_batch_runner.git
cd sd_batch_runner
py -3.10 -m venv venv
venv\Scripts\activate.bat
pip install -r requirements.txt

Preparation on the sdwebui side

Installation

This is a client application using the sdwebui api, so it is necessary to launch sdwebui beforehand.

First, of course, is to run webui with --api commandline argument
example in your "webui-user.bat": set COMMANDLINE_ARGS=--api

Install the following extensions.

Face-1:1,0,0,0,0,0,1,1,1,0,0,0
Face-2:1,0,1,0,1,0,0.8,1,1,0.6,0,1
Face-3:1,0,0,0,1,0.3,1,1,1,0.8,0,0.4
Noface:1,1,1,1,0,1,0,0,0,1,1,1
Wear:1,1,1,0,0,0,1,1,1,0,0,0
Pose:1,0,0,0,0,1,1,1,0,0,0,0
ArtStyle:1,0,0,0,0,0,0,0,0,1,1,1
Char:1,1,1,0,0,0,1,1,1,1,1,1
BG:1,1,1,1,0,1,0,0,0,1,1,1
Soft:1,0,0,0,0,0,1,1,1,1,1,1
Due to the overwhelming complaints about GroundingDINO installation and the lack of substitution of similar high-performance text-to-bounding-box library, 
I decide to modify the source code of GroundingDINO and push to this repository. 
Starting from v1.5.0, you can choose to use local GroundingDINO by checking Use local groundingdino to bypass C++ problem on Settings/Segment Anything. 
This change should solve all problems about ninja, pycocotools, _C and any other problems related to C++/CUDA compilation.

Organizing Lora Files

This application assumes that lora files are organized in subfolders by type.
The following is an example. Some files may be difficult to classify, but there is no need to be so strict.
(If there are too many files to organize, you can put only the files you want to use in a subfolder.)

YOUR_SD_PATH/models/Lora/Pony/character
YOUR_SD_PATH/models/Lora/Pony/style
YOUR_SD_PATH/models/Lora/Pony/pose
YOUR_SD_PATH/models/Lora/Pony/item
YOUR_SD_PATH/models/Lora/Pony/etc
YOUR_SD_PATH/models/Lora/SDXL/character
YOUR_SD_PATH/models/Lora/SDXL/style
YOUR_SD_PATH/models/Lora/SDXL/pose
YOUR_SD_PATH/models/Lora/SDXL/item
YOUR_SD_PATH/models/Lora/SDXL/etc

Restart sdwebui when you have finished organizing the files.
Then, use the Stable-Diffusion-Webui-Civitai-Helper you just installed to collect the lora preview image and trigger word and other information.
(Civitai Helper Tab -> "Scan Models for Civitai" -> press "Scan" button)

How To Use

run sdwebui

  • GUI
    run launch_GUI.bat
    (Perform Initial Setting when starting up for the first time.)
    Go [Generate -> Create New Sequence -> common -> prompt]
    Select some lora files. By default, one character lora is selected at random.
    Go [Generate -> Create New Sequence -> common -> prompt -> preset_tags]
    Select quality tag.
    Go [Generate -> Create New Sequence -> common -> prompt -> header(or footer)]
    Enter prompt.
    Go [bottom left of app screen -> batch_count]
    For now, set about 4.
    Go [bottom left of app screen -> Generate]

  • CUI
    run launch_cmd.bat
python main.py

Initial Settings

There are a few items to be set up at first startup.

  • Lora Directory Setting
    Set the path of the directory containing lora. Multiple environments can be set up, but only one is needed at first.
  • Config -> main
    Select a stable diffusion checkpoint file.
    Select the environment name set in [Lora Directory Setting].
  • Generate -> Lora Update
    Run every time you add or remove a lora file.

Advanced Settings

  • Controlnet Alias Setting
    Several settings are included as samples. Set the model if you want to use them.
  • Config -> lora_generate_tag
    Feature to automatically add train tags extracted from safetensors to the prompt.
    If you find the automatically added tags intrusive, please adjust them.
  • Config -> lora_block_weight
    If you do not need lbw, disable it.
    It is safer to use it when combining multiple lora, as it has the effect of making the image less likely to break up.
  • Config -> adetailer
    If you want to generate even higher quality images, set [person_yolov8n] to 1 and [face_yolov8n] to 2.
  • Config -> segment_anything
    Set sam model.
    Set dino model.
    ( Retrieving the dino model name is not supported by the api, so you need to enter the model name by text.
    See sdwebui txt2img -> Generation -> Segment Anything -> Enable GroundingDINO -> GroundingDINO Model )

TODO

Create some sample sequence json files

Related resources

About

This is an application for batch execution of AUTOMATIC1111 UI launched on a local PC via api. (This is NOT sdwebui extension.)

Topics

stable-diffusion automatic1111 stable-diffusion-api

Resources

Readme
Activity

Stars

0 stars

Watchers

1 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published

Languages