Using the same basic trick as my previous version ( https://openart.ai/workflows/molkemon/thicc-lines-comic-style/yKLIVdo3XvuzMxwh47NJ ), by blending controlnet preprocessor lineart with the original video and only then sending it to the AI.

However, this version works almost completely differnt than the original version, as in it doesn't use controlnet besides the preprocessing at all, instead using ipadapter, low denoising strength and upscaling.

Benefits of this method are that it can run on almost any hardware, you just have to reduce upscale tilesize and likewise make the results better by increasing both the base upscale to x4 or x8 and increasing the scaling factor higher at cost of time or hardware requirement.

Despite the low denoising strength, the effect will always be strong due to the basic lineart trick. However, higher denoise obviously yields more interesting, but less coherent results.

Video Demo:

https://reddit.com/r/StableDiffusion/comments/19euzhx/gta_6_trailer_in_gta_artwork_style_with_workflow/