I produced these visuals recently as a study on AI Text Prompts that output visuals showing various paper arts.

AI TEXT PROMPT EXPERIMENTS

All of the following images are generated on an old pc laptop with nvidia 1060 and a custom python script (stable diffusion 1.5). The following visuals are the raw output from the AI Generator and do not have the standard upscaling, inpainting, manual human artist edits that are typical in art production pipelines.

Referring to my previous experiments where I looked at the various Stable Diffusion Sampling Method options, I decided to use the following sampling methods for this set of experiments because they are probably best suited for generating environment visuals and paper arts. Sampling Method Details

	ai text prompt: feathery paper art, quilling, cut paper, colorful feathers Sampling Method: DDIM (to quickly design a text prompt)
	ai text prompt: feathery paper art, quilling, cut paper, colorful feathers, Super high details, 8k. hyper real, photorealistic, detailed –ar 16:9 –s 42000 –uplight, blur Sampling Method: DDIM (to quickly design a text prompt)
	ai text prompt: feathery paper art, quilling, cut paper, colorful feathers, Super high details, 8k. hyper real, photorealistic, detailed –ar 16:9 –s 42000 –uplight, blur Sampling Method: LMS Sampling Steps: 13 CFG Scale: 9
	ai text prompt: feathery paper art, quilling, cut paper, colorful feathers, Super high details, 8k. hyper real, photorealistic, detailed –ar 16:9 –s 42000 –uplight, blur Sampling Method: LMS Karras Sampling Steps: 13 CFG Scale: 9 ** Note: this ran fast with preferred visual output
	ai text prompt: feathery paper art, quilling, cut paper, colorful feathers, Super high details, 8k. hyper real, photorealistic, detailed –ar 16:9 –s 42000 –uplight, blur Sampling Method: PLMS Sampling Steps: 13 CFG Scale: 9

CONCLUSION

I found the LMS Karras was the fastest sampling method and the quality of the generated visual art is preferable. As always, DDIM sampling method ran fast and also produced visual outputs that I wanted my AI Text Prompt designs to generate. Using the AI Text Prompt tips that I’ve written about in previous AIGeneration.blog articles helped me to quickly design text prompts that are within the specific requirements for the experiment described in this article. Now I can refer to the results on this article for my future productions of any topic, style, color, etc.