A Unity package to run pretrained diffusion models with Unity Sentis
OpenUPM · Documentation (coming soon) · Feedback/Questions
This is essentially a port of Hugging Face’s diffusers library.
It is still very early though, so as of today only a limited number of pipelines and schedulers are supported (see below).
via OpenUPM
In Edit -> Project Settings -> Package Manager, add a new scoped registry:
Name: Doji
URL: https://package.openupm.com
Scope(s): com.doji
In the Package Manager install com.doji.diffusers either by name or select it in the list under Package Manager -> My Registries
Some things that might be worked on next are:
- add/fix classifier-free guidance
- support more models
- SD 1.5
- SD 2.1
- SDXL
- SD-Turbo
- SDXL-Turbo
- LoRA models
- more scheduler implementations
- PNDM
- DDIM
- EulerDiscrete
- EulerAncestralDiscrete
- LCM
- DDPM
- KDPM2Discrete
- KDPM2AncestralDiscrete
- more pipeline/model types
- txt2img
- img2img,
- inpaint,
- upscale,
- depth2img
- marigold
- async pipeline methods (sliced inference)
- support multiple images per prompt
- write some documentation