Knight Molly
使用したAI
Stable Diffusion XL
この初稿はすでにほぼ完璧でした。ChatGPTで生成されたもので、理論的にはCopilotと同様にDALL-Eで描かれているはずですが、おそらく理解後に自動生成されたプロンプトが優れていたのかもしれません。無料の使用時間が終わらない限り、ChatGPTで画像を生成する方が好きです。元の生成画像が1024x1024だったため、背景をBrushnetで置き換え、さらに顔部分のみをinpaintで置き換えました。というのも「イメージ to イメージ」ではこのような動作は生成できなかったからです。その後、一連の修正やKritaでの加工を行い、最終的に生成された手が理想的ではなかったため、初稿の手を重ねて使用しました。ほとんど持っているスキルを全て駆使しました。
The initial draft of this was already nearly flawless. It was generated by ChatGPT, and theoretically, like Copilot, it should be drawn using DALL-E. However, perhaps it was because the automatically generated prompts after comprehension were better. Unless the free time runs out, I prefer using ChatGPT to generate images. Since the original generation was 1024x1024, I used Brushnet to replace the background, and then I only used inpaint to swap the face because "image to image" couldn't produce this kind of action. After that, it was a series of redrawing and retouching with Krita. Finally, as the hand generated wasn't ideal, I layered the hand from the initial draft on top. I pretty much used all the skills I knew.
See more R18+ pictures of my work:
https://www.patreon.com/daniuchen
The initial draft of this was already nearly flawless. It was generated by ChatGPT, and theoretically, like Copilot, it should be drawn using DALL-E. However, perhaps it was because the automatically generated prompts after comprehension were better. Unless the free time runs out, I prefer using ChatGPT to generate images. Since the original generation was 1024x1024, I used Brushnet to replace the background, and then I only used inpaint to swap the face because "image to image" couldn't produce this kind of action. After that, it was a series of redrawing and retouching with Krita. Finally, as the hand generated wasn't ideal, I layered the hand from the initial draft on top. I pretty much used all the skills I knew.
See more R18+ pictures of my work:
https://www.patreon.com/daniuchen
呪文
入力なし
