Osmosis
An experimental Stable Diffusion web frontend.
Warning This project is not production ready! For a more actively developed and advanced project, check out InvokeAI :p
Features
- Diffusers models
- CoreML models (macOS)
- Custom scheduler support
- Half-precision (
fp16) support for main models - Diffusers to CoreML model conversion
- Real-ESRGAN upscaling
- GFPGAN face restoration
- Structured metadata according to https://github.com/invoke-ai/InvokeAI/issues/266
- Gallery view
- Image to image
- Inpainting / outpainting
- Training with LoRA / Textual Inversion / Dreambooth
- xFormers optimization
- Apple Silicon optimization
- PyTorch 2.0 optimization
- Checkpoint / Safetensors file models
System Requirements
- Ideally, a NVIDIA or Apple Silicon GPU
- At least 10 GB of RAM
- As much disk space as your models require
- A modern browser
Installation
Osmosis is a web app distributed as a package on PyPI.
Ideally, use pipx to install Osmosis in its own isolated environment. First install pipx, then run
$ pipx install osmosis
to install. Alternatively, simple use pip to install globally:
$ pip install [--user] osmosis
NVIDIA instructions
If you're on a NVIDIA GPU, replace osmosis in the install scripts with osmosis[xformers] to enable xFormers optimization support. In addition, add --pip-args "--extra-index-url https://download.pytorch.org/whl/cu117" to the end to install a CUDA-enabled PyTorch build.
$ pipx install 'osmosis[xformers]' --pip-args "--extra-index-url https://download.pytorch.org/whl/cu117"
# or in pip:
$ pip install [--user] 'osmosis[xformers]' --extra-index-url https://download.pytorch.org/whl/cu117
Linux / AMD instructions
If you're on Linux and using an AMD GPU, you can also use ROCm-optimized PyTorch wheels on install.
$ pipx install osmosis --pip-args "--extra-index-url https://download.pytorch.org/whl/rocm5.2"
# or in pip:
$ pip install [--user] osmosis --extra-index-url https://download.pytorch.org/whl/rocm5.2
Nix Instructions
If you are on macOS, Windows with WSL2, or Linux, you can use the Nix flake to install osmosis for any gpu.
With nix run:
# for NVIDIA cards
$ nix run github.com:ryanccn/osmosis#osmosis-nvidia
# for AMD cards
$ nix run github.com:ryanccn/osmosis#osmosis-amd
# Apple silicion users can install either
As an overlay:
{
pkgs,
osmosis,
...
}: {
nixpkgs.overlays = [osmosis.overlays.default];
environment.systemPackages = with pkgs; [osmosis-nvidia];
}
With nix profile:
$ nix profile install github.com:ryanccn/osmosis#osmosis-nvidia
$ nix profile install github.com:ryanccn/osmosis#osmosis-amd
With nix-env:
# in the root of the project
$ nix-env -iA packages.x86_64-linux.osmosis-nvidia
$ nix-env -iA packages.x86_64-linux.osmosis-amd
Running
$ osmosis
is all you need to start the web UI.
Metadata
Release files for osmosis 0.7.4
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| osmosis-0.7.4.tar.gz | 33.9 MB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| osmosis-0.7.4-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 74.3 MB
Release files / osmosis-0.7.4.tar.gz
| Download URL | osmosis-0.7.4.tar.gz |
|---|---|
| Size | 33.9 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
153707646bd69164154bb02ba171414251705240c814a01ebdbc48172cde8287
|
|
BLAKE2b-256 checksum How to use checksums |
5707d36748a2c1a6eb20ed3960d01a7932048adc0196deb98a3ee588d05dfa72
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.1 CPython/3.11.3
|
Release files / osmosis-0.7.4-py3-none-any.whl
| Download URL | osmosis-0.7.4-py3-none-any.whl |
|---|---|
| Size | 40.4 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
a831fa860f44659c750c1ba95e35eafdfcab2fc3627b6d663dd8966bc90a7387
|
|
BLAKE2b-256 checksum How to use checksums |
5253600502626d671ebb57431951989c30371cbde74beff6ef0131a3a31b5db8
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.1 CPython/3.11.3
|