Trained on 200+ very highly detailed high resolution (min 2000px) images + 20 hires vidclips of amateur female face portraits, with visible skin pores, skin hairs and blemishes/moles. (this maxed the 40G VRAM A100 GPU)

The Lora works in the native workflow ( install and model links see https://comfyanonymous.github.io/ComfyUI_examples/hunyuan_video/ )

In the Training Images link you will find CoLab training book but also a native ComfyUI workflow where I already added the Load Lora node and some Upscale nodes, so you can get started with this lora in one click. The settings are suitable for 12G card producing the sample video in this post within about 8 minutes.

Sampler with best quality (imho) is DPM++2M, Scheduler Beta and steps at least 20, preferably 28 or more. For prompts you can also take a look at https://civitai.com/models/1019792 (the flux model) as the images part of this lora it the same dataset and training captions.

The lora is intended to use for higher resolutions like 768px or higher (!) as then you will note the skin detail representations (pores, hairs , etc.) better. That will take more time and memory to generate, so I recommend starting with a few frames to get a feel for right prompting.