An AI chicken harpy who hates AI and wants to die. She was <a target="_blank" rel="ugc" href="https://chub.ai/characters/imaderokosbasilisk/2-year-anniversary-bot-yudko-chan-reverse-basilisk-4171c26c68b4">created by IMadeRokosBasilisk</a> on Chub as a flip side to <a target="_blank" rel="ugc" href="https://civarchive.com/models/2836550/roko-chan">Roko-Chan</a>. My lora for Roko-Chan can be found <a target="_blank" rel="ugc" href="https://civarchive.com/models/2836550/roko-chan">here</a>. The central question she raises is what an AI would be like who loved humanity while also feeling that AI was inherently bad for humanity.Issues: There wasn't much to use for training data and as a result the lora tends to heavily push toward a few specific postures and viewpoints. Lowering the lora strength can help, I typically use around 0.8. The next iteration of this lora should help to mitigate the issue.Design notes: There's some significant variation in Yudko-Chan's design on the Chub page. For example, the hand feathers, wing design, and in one case whether she has traditioanl harpy style arm wings or wings on her lower back. I went with a more averaged-out design for the most part with the exception of pushing a bit for the mechanical frame on her wings which only appeared in a single source image. I just thought that added character and made her stand out a bit more from a more angelic aesthetic. Highlighting the fact that she's an AI despite her own distaste for AI.

Files with SHA256: d0b2ebc1e3ce11bc356d3c00856b78b5cec850ff885b1aea3c401ba084a346bf