Hello, > I wonder if this regression comes from Ocrad itself or from the input > Qemu screeshot. I fixed it by removing the OCR delay altogether with: 822eacc6bb0878323e6687d4460a7c53066545e1. Thanks, Mathieu