for started i found this:

https://huggingface.co/akameswa/tiny-sd-onnx-fp16

this is tiny sd, it is one of the smallest models which is from this i believe:
https://huggingface.co/segmind/tiny-sd
correct me but i think these models are listed as 'distilled', they have their own pipeline which will be great someday we have them here too:
https://huggingface.co/docs/diffusers/main/en/using-diffusers/distilled_sd

as the official page says:
"distilled models are upto 80% faster than the Base SD1.5 Models"

anime based LCM model which needs to be converted to onnx first:
https://huggingface.co/furusu/LCM-Acertainty

tutorial on how to convert sd models to ONNX fp16 for cpu and gpu by @dakenf:
https://github.com/saddam213/OnnxStack/issues/9#issuecomment-1803320472

int8 quantized sd models which are small but quality is lower sometimes:
https://huggingface.co/ClashSAN/miniSD-quantized-onnx/tree/main

https://huggingface.co/ClashSAN/miniSD-anything-v3-swapped-onnx-quantized/tree/main
[use int8 folder in it as it is smaller]

the script that is being used [needs to be updated as i remember needs older onnx versions, versions that are lower 16-15 i think:
https://github.com/LowinLi/stable-diffusion-streamlit/blob/main/src/stable-diffusion-streamlit/pages/model/quantization.py

please feel free to share if u ever found something interesting here

Real-Time LCM!

https://github.com/radames/Real-Time-Latent-Consistency-Model

this is something amazing, didn't try it, but with the looks of it u can use ur webcam and turn urself to anything in real time! will be cool someday we have it here too

fp16 model lf LCM-Dreamshaper v7:
https://huggingface.co/aislamov/lcm-dreamshaper-v7-onnx/tree/main

some more models [sd 1.5], they are directml fp16:

https://huggingface.co/sharpbai/stable-diffusion-v1-5-onnx-directml-fp16
https://huggingface.co/sharpbai/chilloutmix-onnx-directml-fp16

u can find fp32 and more versions on sharpbai's page

chilloutmix is excactly what i was looking for! it even can generate low resolution images 256x256 320x320 384x384 in very high quality

a tip to generate realistic images with lcm dreamshaper:
give it a location sometimes, and use = real, detailed human image, sharp focus
example:
a handsome man with brown hair walking, in a beautiful alley in japan, real, detailed human image, sharp focus

New Model template to Import into OnnxStack

Deliberate-v3
deliberate-v3.json

Download file, Open OnnxStack, go to Model section, click Import, select the json file

my new favorite way to make new images is one of those 2x2frames with sd, have it 512x512 in lcm, activate automation to generate batch of images and:
quadtych photo of a woman talking to another man, in japan, detailed human image, sharp focus, real

now what it does is gives u an image with multiple frames in it and if u get lucky those images will be frames of same people with different motions!
u can easily crop the frames out of the image, put them together and make a gif or put it in flowframes [another ai app to generate more frames and make it more smooth]. voila, u have an animated image.

a bit off topic:
@saddam213 after ur onnxstack and maybe ur other projects, do u have any plan for bringing sound and music to c# onnx? i tried AudioCraft and it's pretty amazing! a c# onnx version will be great:
https://github.com/facebookresearch/audiocraft
https://github.com/GrandaddyShmax/audiocraft_plus

some olive optimized fp16 models [Realistic Vision, Deliberate, Photon and Dreamshaper]:
https://huggingface.co/softwareweaver/realisticvision/tree/main
https://huggingface.co/softwareweaver/deliberate/tree/main
https://huggingface.co/softwareweaver/photon/tree/main
https://huggingface.co/softwareweaver/dreamshaper/tree/main

LCM-Dreamshaper v7 Olive optimized:
https://huggingface.co/softwareweaver/LCM_Dreamshaper_v7_Olive_Onnx
more here:
https://huggingface.co/softwareweaver

HI, recently reached this forum... will follow it...as you are sharing resources, will like to share my own resources. a UI for onnx to low mem gpu or cpu, giving control on where the models are loaded during runtime , and also including a hi-res pipeline approach....(also, LCM support is on development and almost ready to be implemented...)
https://github.com/NeusZimmer/ONNX-ModularUI-StableDiffusion
PD: i moved from .Net to python to work with stable diffusion models good to see someone working on .Net approach, will try to contribute

Ideas0:

• Ability to use Refiners (two or more checkpoints to generate an image). More than (1) refiner would be very cool.
• Ability to convert .pth, .safetensors, and .ckpt, to .onnx from tools build into or side-by-side with the app (it will improve-usability)
• ability to more easily swap out models in the codebase. Perhaps this is already implemented and I just have not explored this yet. I will read the codebase on this now