41 lines
1.1 KiB
Markdown
41 lines
1.1 KiB
Markdown
|
+++
|
||
|
|
date = "2023-11-08"
|
||
|
|
tags = ["diffusion"]
|
||
|
|
title = "stable-diffusionのanimatediff-cliを試してみた"
|
||
|
|
slug = "stable-diffusion"
|
||
|
|
+++
|
||
|
|
|
||
|
|
https://github.com/neggles/animatediff-cli
|
||
|
|
|
||
|
|
```sh
|
||
|
|
$ git clone https://github.com/neggles/animatediff-cli
|
||
|
|
$ cd animatediff-cli
|
||
|
|
$ python3.10 -m venv .venv
|
||
|
|
$ source .venv/bin/activate
|
||
|
|
# install Torch. Use whatever your favourite torch version >= 2.0.0 is, but, good luck on non-nVidia...
|
||
|
|
$ pip3 install torch torchvision torchaudio
|
||
|
|
|
||
|
|
$ python
|
||
|
|
>>> import torch
|
||
|
|
>>> torch.backends.mps.is_available()
|
||
|
|
True
|
||
|
|
|
||
|
|
# install the rest of all the things (probably! I may have missed some deps.)
|
||
|
|
$ python -m pip install -e '.[dev]'
|
||
|
|
```
|
||
|
|
|
||
|
|
modelのloraは、基本的にアニメ絵がうまいやつで、このツールはloraの`motion-module`を使います。
|
||
|
|
|
||
|
|
motion-lora : https://drive.google.com/drive/folders/1EqLC65eR1-W-sGD0Im7fkED6c8GkiNFI
|
||
|
|
|
||
|
|
```sh
|
||
|
|
$ mv *.ckpt ./data/models/motion-module/
|
||
|
|
|
||
|
|
$ vim config/prompts/01-ToonYou.json
|
||
|
|
"motion_module": "models/motion-module/mm_sd_v15_v2.ckpt"
|
||
|
|
|
||
|
|
$ animatediff --help
|
||
|
|
```
|
||
|
|
|
||
|
|
model : [sd v1.5](https://huggingface.co/runwayml/stable-diffusion-v1-5)
|