In what ways is it better compared to LightX2V or drbaph Turbo Loras?
The example videos are in very low resolution; I’d love to know more, thanks!
in no way 🤣
I tested this LoRA with a simple I2V workflow to create a live photo (a simple smile and hand wave).
For my needs, this LoRA is better, I need consistency in character, lighting, and color, and this LoRA does a better job than the LightX2V LoRA.
Here is a comparison(same setup, same prompt):
1.This one
The skin looks more natural and less plastic using this one, remaining more consistent with the reference image. Notice the light and coat color changes in the LightX2V version.
However, the audio is still an issue. It is completely corrupted despite using minimax_h3_fl2va_pruned_bf16 as in the example workflow. I'm not sure if this is a ComfyUI version issue.
i go to test, the audio its very bad, well with others turbo loras for me the audio is bet with 0.5 strenght.
i go to test, the audio its very bad, well with others turbo loras for me the audio is bet with 0.5 strenght.
Yeah, the audio is completely broken. I couldn't even reproduce normal audio using the exact workflow from the video provided in the assets folder.
i go to test, the audio its very bad, well with others turbo loras for me the audio is bet with 0.5 strenght.
Yeah, the audio is completely broken. I couldn't even reproduce normal audio using the exact workflow from the video provided in the assets folder.
I think the ComfyUI-version/comfy-kitchen version may have bugs in loading voice related LoRA weight , I can get good voice example in comfy-kitchen 0.2.26, some one can't get in comfy-kitchen 0.2.28. Could u tell me yourcomfy-kitchen version.
4step sample Same Prompt with lora: