Use -Realvis versions together with RealisticVision 5.1 for replicable results.

Suggested weights
Weight: 0.6 - 0.7

Bleeds quite a bit into the model.

Was trained on an internal model so you wont get the exact crisp results as in the promotion images.

Models like cyberrealistic, realisticVision and photoMovie tend to come close with the original LoRA. Otherwise use the -RealVis version.