I've recently gotten into drawing with AI as a hobby
I installed Stable Diffusion locally and have been generating various images. At first, I didn't know much about prompts, so I only got weird pictures, but now I'm starting to get the hang of it. I especially find it fun to customize LoRA models to train specific characters. The downside is that without a high-end GPU, the generation speed is slow. For anyone else into AI image generation, could you share some good prompt sites or tips?
3 answers
I've also gotten really into AI art lately! Stable Diffusion is so much fun. For prompts, checking out examples posted by others on sites like Civitai or PromptHero is really helpful. Especially adding 'bad anatomy, extra fingers' to the negative prompt really boosts the quality. For GPU, cloud services or even the free version of Google Colab work fine. It's amazing that you're even doing custom LoRA!
I've had a similar experience! I started with a 1060 6GB, but it was too slow, so I ended up buying a used 3090. It's definitely much smoother now. As for prompt tips, adding lighting and composition details to basic tags like 'masterpiece, best quality' really makes the results pop. By the way, what character LoRA did you make? I'm practicing with Genshin Impact characters, but the skin tone often breaks.
Nice to meet you as a fellow AI art enthusiast! I mainly use NovelAI, but I'm envious of your local setup. I'd also recommend Lexica.art as a prompt site. I totally relate to the slow generation speed—lowering the resolution, using 'highres-fix', or reducing the batch size to 1 can help a bit. Have you tried extensions like ControlNet? They're really useful for setting poses or compositions.