|
| 1 | +# Stable Diffusion Unity Tools |
| 2 | + |
| 3 | +## Summary |
| 4 | + |
| 5 | +Editor utility to generate assets in Unity Editor via self-hosted & managed Stable Diffusion installations |
| 6 | + |
| 7 | +## Installation |
| 8 | + |
| 9 | +1) Setup Stable Diffusion on your local machine - https://github.com/CompVis/stable-diffusion (skip if you will use managed services) |
| 10 | +2) Add package in Packages/manifest.json: |
| 11 | + ``` |
| 12 | + "com.konh.stable-diffusion-unity-tools": "https://github.com/KonH/StableDiffusionUnityTools.git?path=Packages/com.konh.stable-diffusion-unity-tools" |
| 13 | + ``` |
| 14 | +3) Click **SD Tools/Setup**, it creates settings file<br/> |
| 15 | + |
| 16 | +4) Open settings at **Edit/Project Settings/Stable Diffusion** and configure: |
| 17 | + - **Command** - command to execute stable diffusion: |
| 18 | + - Should contain required placeholders: `$PROMPT` for generation prompt, `$OUT_DIR` for destination directory, `$WIDTH` and `$HEIGHT` for image dimensions (these arguments assigned at generation process) |
| 19 | + - Could be multiline for better usability, line breaks are ignored |
| 20 | + - Windows WSL example (in that case paths should be related to Linux file system): `wsl |
| 21 | + --cd "linux-mnt-path-to-stable-diffusion" |
| 22 | + python3 scripts/txt2img.py |
| 23 | + --prompt "$PROMPT" --plms --n_samples 1 --n_iter 1 --outdir "outputs/txt2img-custom/$GUID"` |
| 24 | + - If you want to use managed service - write custom script which communicate with it and store results on file system in txt2img.py compatible format |
| 25 | + - **ResultPath** - full path to directory with generated files on local file system (in case of WSL - Windows host machine) |
| 26 | + |
| 27 | +5) Additionally, you could add **Assets/StableDiffusionUnityTools** to .gitignore (it's related to current machine, unfortunately we can't use UserSettings for that) |
| 28 | + |
| 29 | +## Asset file generation |
| 30 | + |
| 31 | +Now you can use **SD Tools/Generate Image** to generate images, it provide you ability to tweak settings for next generation: |
| 32 | + - Required settings: **Prompt**, **Directory**, **Width**, **Height** |
| 33 | + - Optional settings: **Name** (if not assigned, GUID is used) |
| 34 | + |
| 35 | + |
| 36 | +When generation finished, you will see result in given directory |
| 37 | + |
0 commit comments