Acly

> Seems like it happens when there's transparency for the background (the checkered gray background). When I painted it with white, it worked fine That sounds like you used transparent/black...

Can you try and [export a workflow](https://github.com/Acly/krita-ai-diffusion/wiki/Common-Issues#comfyui-workflow-export) which results in black image and run that in ComfyUI directly? I'd expect output to be black there too, but it might be...

This looks related: https://github.com/comfyanonymous/ComfyUI/pull/4548 If I read it correctly you can try providing `--force-upcast-attention` to ComfyUI startup If you are using manage server install: [How to pass arguments](https://github.com/Acly/krita-ai-diffusion/wiki/Common-Issues#how-do-i-pass-additional-arguments-to-comfyui)

The hype is real when 40sec for a fp8 distilled low-step model is considered "perfect" :D I added support for the checkpoint versions ([flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors), [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors)). * Make sure to update...

You can use [krita_ai_diffusion-1.22.0-pre1.zip](https://github.com/Acly/krita-ai-diffusion/releases/download/v0.1.0/krita_ai_diffusion-1.22.0-pre1.zip) It also has a built-in default style for Flux [schnell] Run local: make sure your ComfyUI is setup to run [this](https://comfyanonymous.github.io/ComfyUI_examples/flux/#simple-to-use-fp8-checkpoint-version) Cloud: use the "Online Service"...

Please use the ones from https://huggingface.co/Comfy-Org - they are also fp8. They're only bigger size because they include the vae + text encoder, which you'd need separately for the Kijai...

> Do we know what the RAM requirement for this model is? roughly 12 GB for fp8, 24 GB for fp16 that's for the diffusion model, you tend to need...

no, not really. you can already do what this video does in the plugin, but it only works ok for some simple scenarios.

Not sure, there were no changes in that regard. The presets located in `krita/ai_diffusion/presets` aren't touched during updates. Only the built-in ones in `krita/pykrita/ai_diffusion/presets` are replaced.

It works for me. Check client.log for something like this: ``` 2024-08-15 19:47:47,647 INFO Loaded 2 sampler presets from C:\Users\Acly\AppData\Roaming\krita\ai_diffusion\presets\samplers.json ```