Go to [Hugging Face](https://huggingface.co/CompVis/stable-diffusion-v-1-4-original), and click "Access repository" to Download ```sd-v1-4.ckpt``` (~4 GB) to ```~/Downloads```.
You'll need to create an account but it's quick and free.
They provide a flexible, reliable way to build and deploy Stable Diffusion. We also use a Docker volume to store the largest model file and image outputs as a first step in decoupling storage and compute. Future enhancements will do this for other model files and assets. See [Processes](https://12factor.net/processes) under the Twelve-Factor App methodology for details on why running applications in such a stateless fashion is important.
This example uses a Mac M1/M2 (arm64) but you can specify the platform and architecture as parameters when building the image and running the container. You'll also need to specify the Stable Diffusion requirements file that matches your OS and architecture e.g. Linux on an arm64 chip if running a Linux container on Apple silicon.
The steps would be the same on an amd64 machine with NVIDIA GPUs as for an arm64 Mac; the platform is configurable. You [can't access the Mac M1/M2 GPU cores from Docker containers](https://github.com/pytorch/pytorch/issues/81224) and performance is reduced compared with running it directly on macOS but for development purposes it's fine. Once you're done with development tasks on your laptop you can build for the target platform and architecture and deploy to an environment with NVIDIA GPUs on-premises or in the cloud.
On the Docker Desktop app, go to Preferences, Resources, Advanced. Increase the CPUs and Memory to avoid this [Issue](https://github.com/lstein/stable-diffusion/issues/342). You may need to increase Swap and Disk image size too.
Copy the model file (we'll need it at run time) to the Docker volume using a lightweight Linux container. You just need to create the container with the mountpoint; no need to run it.
```condaarch``` will restrict the conda environment to the right architecture when installing packages. It can take on: ```linux-64```, ```osx-64```, ```osx-arm64```.
If you're on a **Linux container** the ```dream``` script is **automatically started** and the output dir set to the Docker volume you created earlier.
If you're **directly on macOS follow these startup instructions**.
With the Conda environment activated (```conda activate ldm```), run the interactive interface that combines the functionality of the original scripts ```txt2img``` and ```img2img```:
Use the more accurate but VRAM-intensive full precision math because half-precision requires autocast and won't work.
By default the images are saved in ```outputs/img-samples/```.
```strength``` is a value between 0.0 and 1.0 that controls the amount of noise that is added to the input image. Values that approach 1.0 allow for lots of variations but will also produce images that are not semantically consistent with the input. 0.0 preserves image exactly, 1.0 replaces it completely.
Make sure your input image size dimensions are multiples of 64 e.g. 512x512. Otherwise you'll get ```Error: product of dimension sizes > 2**31'```. If you still get the error [try a different size](https://support.apple.com/guide/preview/resize-rotate-or-flip-an-image-prvw2015/mac#:~:text=image's%20file%20size-,In%20the%20Preview%20app%20on%20your%20Mac%2C%20open%20the%20file,is%20shown%20at%20the%20bottom.) like 512x256.
