Simple Docker Deployment for the popular Anime AI Illustrator
** WARNING: Using this content might be illegal in your country or location. **
Prerequisites for out-of-the-box execution (without rebuilding the modules and exchanging the transformer models)
- NVIDIA GPU with >35 GB Memory
- NVIDIA Driver == 515.xx (or greater)
- Get the model by executing:
curl -LO https://cloudflare-ipfs.com/ipfs/bafybeiccldswdd3wvg57jhclcq53lvsc6gizasiblwayvhlv6eq4wow7wu/animevae.pt
curl -L https://cloudflare-ipfs.com/ipfs/bafybeibhlcqjgvjtc3ck2bcazae5yizaetm7y7nm5rux674gsxgd4l5uce/animefull-latest.tar.zst | zstdcat | tar xf -
- Get the prebuild container
docker pull rimurosu/novai:withpt
If you do not with to use the pre-built OpenAI Transformer model (which is frequently updated, so might need a refresh, and I don't really have any bandwidth to do so), please use this instead:
docker pull rimurosu/novai:latest
This will trigger the download when you execute the container.
- Run the container with the following command:
docker run --gpus all -v $PWD/animevae.pt:/workspace/animevae.pt -v $PWD/animefull-latest:/workspace/animefull-latest -p 80:80 -p 8080:8080 rimurosu/novai:withpt
assuming that your downloaded models are located in the current directory with the name animefull-latest and animevae.pt.
-
Use local browser and navigate to
http://127.0.0.1:8080 -
In the website, use the backend address
127.0.0.1:80 -
Enjoy
If you want to generate a large amount of content please use the generate.py script.
By default it will generate 1000 images of the same kind.
This project is based on the frontend from novelai-simple by nidbCN. Container is based on NVIDIA NGC Public Container