Takekitdown_Ai
patreon
Notice: 少女の身体実験 updated
🕑 Added 2025-04-13 10:11:14 +0000 UTCRewards(6P+6V/219P+83V) updated in Discord.
This update, similar to the previous one, involved selecting several images I was satisfied with from my debugging and experimentation process, refining them, and generating videos. The delay in this update was due not only to attending a funeral but also to the increased workload of producing six 5K-resolution images and generating additional video content.
Before using ComfyUI, I relied on SD WebUI to generate images purely through text prompts based on themes. Due to WebUI’s functional and performance limitations, as well as the unpredictability of its outputs, I had to focus on a "single image + variations" approach to balance quality and quantity for each update. Many promising experimental results had to be abandoned, even if I found them satisfying. This led to an astonishing number of generated images (currently estimated at 15,000), but my updates were infrequent and sparse. While I could have followed the common practice of flooding platforms like Pixiv or subscription sites with dozens or hundreds of images every day—given that 15,000 images could sustain such a pace for a year—I chose not to. In this era of rapid, ever-evolving AI technology, I prefer to maintain my own style, slow down, and create works that leave a personal mark.
After switching to ComfyUI, as I grew more familiar with it, optimized my workflows, and incorporated new technical models, I achieved a balance between quality and increased output of high-resolution 5K images. Combined with video content, this boosted my posting frequency on Pixiv and social media, leading to greater exposure. Barring unforeseen circumstances, I plan to continue prioritizing this approach for future updates.
Of course, this doesn’t mean I’ll abandon the "single image + variations" method entirely. I will continue exploring high-controllability workflows to improve consistency and efficiency in creating variations. Additionally, I still have one or two older unreleased works, which I’ll publish once my ComfyUI setup is further refined.
