InkPunk LORA was designed for those who crave the gritty, edgy look of inkpunk art, but also want the intelligence and coherence that SDXL LORA offers. It's the best of both worlds, blending unique artistic style with cutting-edge tech, to give you something genuinely new and engaging.
Details: Currently in alpha state, this is the Lora version of the SDXL InkPunk Style Model. Pretty good all around with text2img and img2img, but has difficulty with some animals. Setting the Lora weight to 0.5 is a good starting point, and adjusting the CFG can also help produce better results. With Img2img, you can also change the denoising strength as well, but it's good to start with 0.5 as well. Raising the denoising number can allow more of the style to shine through in situations where it's being shy.
Prompting: Nearly all images in the gallery are going to be using prompts which start with "inkpunk style illustration" and end with the lora itself.
Description
Trained on SoCalGuitarist's NightVision XL model, additional images added to the training set (total of around 110 images for the training), and trained with filewords.
Pros: a bit more coherency across a wider variety of subjects/topics. Likes the 0.8 range usually
Cons: tends to get weaker with longer prompts. Combat this by increasing the weight over 1.0 and describe the overall image as an "illustration".
Other notes: Responds pretty well to basic hair color and style suggestions. Avoid using "photo" or "photograph" if your prompt is longer than about 10 (not including the LoRA name). Also does pretty well with telling it you want color or monochrome etc. (see trigger words for list of potential options that can help guide results)
FAQ
Comments (2)
Details
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.