[Technical Q&A + 1000Buzz Reward]
Hello everyone, I’ve run into an issue while training and testing my own LORA, and I’d love your insight:
When I generate images locally using my familiar workflow, parameters, and base model, the linework is smooth and the colors are clean and natural—like in the example below:



However!
After uploading my LoRA to Civitai, I’ve noticed (and other users have reported) that when generating with the “Civitai public online generator,” the results often show noisy, blurry, oversharpened, jagged, or generally rough linework, as shown below:



My Questions
It’s the exact same LoRA—why are the results so different?
Where does the problem lie? Is it the generation parameters, base model, sampler, or is it something about the Civitai platform itself?
Is there any way to make the LoRA produce consistent results on both local and remote (Civitai) environments? How should I train or tune my LoRA to be more universal?
Reward
I’m offering a 1000Buzz reward for anyone who can provide the correct answer and demonstrate ComfyUI-quality results using this LoRA at weight 1.0 (or at least provide a viable solution—if absolutely impossible, let’s discuss).
The reward is limited to the first two people with successful solutions!
——————————————————————————————
Don't mind the composition and content, just look at the style details (parameters are exactly the same, there is a workflow in the image that you can download and try).

