测试了一下使用视频训练hunyuan video的lora,使用了一个14秒的视频,根据分镜切成了六段,分辨率854 * 480。更大的话4090也会爆显存。提示词使用的是单词的提示词,没使用自然语言,所以使用该lora的时候,还是建议使用单词,强度推荐0.9-1,低了会偏向动漫风格
Tested the use of video training hunyuan video lora, used a 14-second video, according to the split screen cut into six segments, resolution 854 * 480. larger if 4090 will also burst the video memory. The cue word used is the word cue word, did not use natural language, so the use of the lora, it is still recommended to use the word, the intensity of the recommended 0.9-1, low will be biased towards the anime style!
Description
Details
Downloads
158
Platform
CivitAI
Platform Status
Deleted
Created
1/18/2025
Updated
7/7/2025
Deleted
5/23/2025
Trigger Words:
1girl, solo, red eyes
Files
lbk_comfy.safetensors
Mirrors
Huggingface (1 mirrors)
CivitAI (1 mirrors)