I'm new to this whole image generation thing, any tips? I'm using ComfyUI with a custom workflow. I mess a bit with prompts.
A few things I noticed about the workflow:
There are no negative prompts. I recommend adding common negative tags such as: low quality, worst quality, normal quality, text, JPEG artifacts, bad anatomy, old image, early image, watermark, copyright notice, artist name, and signature. This usually improves image quality and reduces unwanted artifacts. The OpenRouter LLM node has been removed because I don't have access to it or the information required to configure it. In the first KSampler, the denoise value is set to 0.35. For the initial image generation, it should normally be 1.0. A value of 0.35 is intended for image refinement or img2img, not for creating the first image. The second KSampler (presumably the "Hi-res fix" stage) also appears to be configured incorrectly. Instead of refining the image generated by the first sampler, it seems to be generating a completely new image. I'd recommend checking the node connections and making sure it's receiving the output image from the first KSampler rather than starting from noise.
Try to keep this thread to fan images please.
However, hard to really find a game like Starkmaker really it has it's own niche with the multiple different directions it takes.
