Best results are achieved from the front. Side and back views are currently less stable and may produce inconsistent details. Increasing the LoRA weight may help, but I generally don’t recommend it, as it can reduce overall image quality.
CuffstoCollar_concept_v1.safetensors