CyanCeeBee Style (Krea 2)
This style LoRA is based on the digital illustrations of CyanCeeBee. It gives images clean linework, smooth cel-style shading and a soft, warm palette of creams, peaches and dusty pinks. Characters come out stylized and slightly anime-influenced, with expressive faces, sleek hair and confident, playful poses. Most of the training set shows pin-up-style women on plain or pastel backgrounds, so the LoRA is strongest at character illustration. It can also handle portraits, city scenes and landscapes, giving them the same flat, polished finish.
Trigger word: CCBeeStyle
Usage
Put the trigger word first, then describe the character, outfit, pose and background.
Start at strength 0.8–1.0.
Plain or pastel backgrounds match the source material best.
Signs and other text in the image may pick up the trigger word, showing “CCBEE” on shop signs, for example. A small signature-like mark can also appear in a corner. Add “no text, no watermark, no signature” to reduce this.
The training set includes nudity, so outputs can lean suggestive even with neutral prompts.
Training details
Base model: Krea 2 Raw. Tested for inference on Krea 2 Turbo.
46 images, auto-captioned with Qwen3-VL.
1000 steps, rank 32, 768 px.









