Stable Diffusion Crash Course - Setup, Basics, Advanced Prompting, ControlNet (Automatic1111 webui)
610:00 Installation
0:54 Setup
1:47 Install Extensions
3:29 Recommended Settings
6:59 Exploring the Interface
8:07 Sampling Steps
8:36 Batching
9:27 CFG Scale
10:46 Upscaling
15:17 Wildcards
16:31 Dynamic Prompts
17:52 Attention Grabber
18:58 Styles - Models / Textual Inversion Embeddings / Loras
20:20 Civitai - Finding Models, Embedings and Loras
21:11 Installing Models, Embeddings and Loras
22:12 Custom Model
23:20 Upscaling
24:38 LDSR - Latent Diffusion Super Resolution
25:40 Generating at High Resolutions
26:30 Unlocking Resolutions Sliders
27:10 GPU Memory Management
28:00 Textual Inversion Embeddings
29:18 Lora
30:03 Illuminati Diffusion - Stable Diffusion 2.1
32:07 Exploring Embedding with Dynamic Prompts
33:14 ControlNet
35:19 Combining Multiple ControlNets
36:47 ControlNet with Stable Diffusion 2.1
37:29 Exporting Normal Map from Blender
38:08 Using the Exported Normal Map
39:11 Exporting Depth Map from Blender
39:50 Using the Exported Depth Map
Automatic1111 webui:
- https://github.com/AUTOMATIC1111/stable-diffusion-webui
Arguments for webui-user.bat:
--xformers --medvram
Suggested file naming scheme for settings:
[seed]-[cfg]-[steps]
Download stable diffusion models, embedings and loras:
- https://civitai.com/
Control net models (smaller file size versions compared to official):
- https://civitai.com/models/9868/controlnet-pre-trained-difference-models
Control net models for Illuminati Diffusion:
- https://huggingface.co/thibaud/controlnet-sd21/tree/main
Database of upscalers:
- https://upscale.wiki/wiki/Model_Database
Interactive comparison of upscalers:
- https://imgsli.com/MTMxODgy/0/30
My favourite 3 upscalers:
- 4x-UltraSharp
- 4x_foolhardy_Remacri
- 4x_NMKD-Superscale-SP_178000_G
https://drive.google.com/file/d/1aeyvzn2ZqDSD3prL8l0vJrh-LpoSxqXj/view?usp=sharing
Loras in the video:
- epiNoiseoffset_v2
- dpep2 ep2
https://drive.google.com/drive/folders/14bvhx8OvbdQaNncs7TEBxDGYa57D9GOm?usp=share_link
Textual Inversion Embeddings seen in the video:
- verybadimagenegative_v1.3
- bad_prompt_version2
- bad-hands-5
- bad-picture-chill-75v
- easynegative
- nartfixer
- nfixer
- ng_deepnegative_v1_75t
- nrealfixer
- Unspeakable-Horrors-Composition-4v
https://drive.google.com/drive/folders/1XeUK3YtsVmBuK7OsTFgG1tjxRai2f7JD?usp=sharing
Models used in the video:
- meinamix_meinaV8
- illuminatiDiffusionV1_v11
- deliberate_v2
- darelitesFantasyMix_
- chilloutmix_NiPrunedFp32Fix
- meinapastel_V3
ControlNet segmentation color spreadsheet:
- https://docs.google.com/spreadsheets/d/1se8YEtb2detS7OuPE86fXGyD269pMycAWe2mtKUj2W8/edit#gid=0
System informer (for nicer GPU memory graphs):
- https://systeminformer.sourceforge.io/