As of now this 8-Steps LightX2V LoRA produces better output
This LoRA trained for a very specific style: close-up portraits featuring platinum blonde / pale silver hair, extremely pale skin with visible freckles and natural texture, and full, plump lips.
The training dataset consisted exclusively of tight headshots and facial close-ups , no full-body, half-body, or wider compositions were included.
performs best in close-up portrait scenarios ( "close-up face", "portrait", "headshot").
For anything beyond tight crops (upper body, full body), you should lower the LoRA strength (0.4–0.5 range recommended) to avoid distortions or over-application of the trained features.
Sampler: euler
Scheduler: sgm_uniform
Steps: 10
CFG: 1