ComfyUI_Krea2_ModelMerger_Converter_int8_int4

ModelMerge Krea 2 + Quant (INT8/INT4)

This node lets you merge two full Krea 2 bf16 models and up to three LoRAs. It relies solely on ComfyUI's internal kitchen. Merging and quantization — int8, int4, or none, in which case it stays in bf16 — happen on the fly: you can plug it straight into a KSampler to check the result right away (quality, inference speed), and optionally save the quantized model to disk in the chosen format.