itch.io

I've come up with a really interesting idea.

Do you have any interesting ideas for creating content?

I'd suggest this approach: feel free to experiment with the settings and customize them to suit your needs.

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.

The images have all the generation metadata embedded. Just load one into a metadata viewer (or back into Stability Matrix/ComfyUI), and you'll be able to see the exact prompt, model, sampler, steps, CFG, seed, LoRAs, and other settings used.

Well, I see that your negative promt is the same as your positive one—why is that? The CFG is also pretty high at 10, and there are 40 steps too—which, in my opinion, is way too many

Hey! I’m using ComfyUI for generating these images. The main model I use is Anima, with a custom LoRA trained for it called “Starmaker Story” to get this specific style and consistency. It gives me more control over the characters, outfits, and overall look compared to using a regular web generator.

Anna, looking a little sad, is waiting for Update 1.9

thx Do you use any LoRa?

Какая модель используется для генерации?

I agree with each of the 7 points.