Lora Trained using Fizgig for Minimax H3-Image (not to be confused with the video model converted to generate Images). No trigger word necessary but you can add "kafkaeffect4" if it's not coming through. Strength depends on how many LoRas you're stacking and what type, as well as how many steps, but usually 1.2-1.6 is a good spot if you're running the 6-step workflow and no other LoRas. Trained on 100 high quality hand-made renders that I made myself over the past few years.

