This model was trained on hot food images taken with V40, you might want to use it at 1 strength and use smoke/mist in negatives for reducing smoke intensity.
rfktrfodmstr-000004.safetensors