Z-Image / Krea 2:

Trained on over 200 high-definition real anime screencaps across a wide range of characters, creatures, poses, detailed backgrounds, and even NSFW content. No Pony/Illustrious junk in the dataset!

Z-Image: It's far from perfect, but it definitely improves anime-style without complex prompting.

Krea 2: The difference is subtle since Krea 2 handles anime well by default.

Illustrious:

Tired of that generic "AI look" in your anime generations? So was I. Most "anime styles" LoRAs I tested left me with warped hands, blurry details, or muddy anatomy: hardly the crisp, expressive style found in actual anime scenes.

True Screencap was created to address this issue. Trained on over 150 high-definition anime screencaps across a wide range of characters, creatures, poses, detailed backgrounds, and even NSFW content, this LoRA delivers clean, sharp, and natural-looking images, all while preserving accurate anatomy, nice eyes, overall clarity, and flexibility across checkpoints and/or when using other LoRAs.

Every image in the dataset was carefully tagged and selected to avoid the common issues plaguing other screencap models. The result? A LoRA that enhances your anime generations with the feel of real anime frames 🔥