Krea2 GPTImage StyleLoRA
A Krea2 style LoRA trained on 300 images generated with GPT-Image-2.5.
The training images were created using prompts derived from my own past and recent works, which were then used to generate new images with GPT-Image-2.5.
The uploaded images and their corresponding prompts are provided as references to give an idea of the visual direction and characteristics represented in this LoRA.
The actual results may vary depending on the prompts and generation settings used.
Trigger Word
GPTImage
Optional. The LoRA works without it, but using it makes prompt adherence more stable.
Because the captions are based on my own prompts, this LoRA may respond best to prompts written in a similar style.
Recommended Strength
1.0
Generation Settings
Samplers used: euler
Scheduler: simple
Training
* Images: 300
* Network Rank: 64
* Training Steps: 3,000
Description
Initial release of Krea2 GPTImage StyleLoRA.
Trained on 300 images generated with GPT-Image-2.5, using prompts developed from my previous and recent works.
Recommended strength: 1.0.
Results may vary depending on prompts and generation settings.