Krea releases Krea-2-Raw base model for fine-tuning

Krea-2-Raw is a raw checkpoint not recommended for inference, designed as a base for finetuning or post-training. It supports Diffusers and can be used with custom LoRAs trained on midtrain.