Badaramoni has released a new adapter for the Krea-2-Turbo model that combines public realism packs of different formats into a single file. The release addresses the limitation where standard merger tools fail to combine classic LoRA and LoKR adapters because they do not share the same low-rank format.
The fusion algorithm operates in weight-delta space by reconstructing each adapter's true delta, normalizing conventions, blending them, and re-factoring the result into a rank-128 LoRA using truncated SVD. This process produces a single portable .safetensors file that avoids the need for multi-adapter stacks at inference time.
The model is available under the Krea 2 Community License and requires access to the base krea/Krea-2-Turbo model to run.