Hey, I’m messing around with Diffusion Bee on my Mac, trying to train a model with 5 SleepyGimp images to capture that art style. My results are not good so far, so I need your wisdom to level up! Here’s what I’m wondering:
- Cropping: Should I crop each SleepyGimp image to focus on one character, or keep the full scene? How do you prep images for training to get the best results?
- Text Prompts: Do you write super detailed descriptions for each image (like every color, pose, and vibe), or go with short, generic tags to let the model do its thing?
- Training Tips: Any hacks for training Diffusion Bee to nail SleepyGimp’s intense, dynamic aesthetic? Like, how many images do you use, or do you tweak stuff like LoRA weights, epochs, or guidance scale?
If anyone’s got a pre-trained model that vibes with SleepyGimp’s style, would you be kind enough to share it or point me to one? I’d love to study it and learn! 🙏 Drop your best training tips, workflows, or favorite prompts below.
Background removal is better than cropping, though. Prepare separate portrait images for each char.
Alternatively you could put that time into learning how to draw using his style as a base.