make the character walking from side more often.
or make the character tend to left or right side.
Description
walking from side
FAQ
Comments (10)
Thanks for everyone, achieved this donation~~
I can't seem to get this LoRA to work. I just end up with janky output or the character shrinking into the background but no walking.
I'm running:
- I2V (Input image is a cartoony character with a white background)
- Wan 2.1 I2V 1.4b GGUF 720p Q8 model
- Clip skip 2
- The prompts for the LoRA
- Output resolution of 480 wide x 720 high
Wondering if maybe it doesn't work with GGUF or should I be using 1.3b or something?
maybe you can try 14B version (not 1.3B version), and I think GGUF is not the problem. the others are fine I think.
@Aseer I'm already running on 14B. The exact model I'm using is: https://huggingface.co/city96/Wan2.1-I2V-14B-720P-gguf/blob/main/wan2.1-i2v-14b-720p-Q8_0.gguf
Also woops, meant "14b", not "1.4b" in my original comment
Is there any way you could provide the metadata for the sample videos? Saving and dragging them to comfyUI doesn't seem to load anything
@RobertBobertson https://blog.comfy.org/p/comfyui-wan21-flf2v-and-wan21-fun and https://blog.comfy.org/p/wan21-video-model-native-support I use the official workflow, maybe you can check the difference and try to correct and fit it to GGUF nodes.
Unable to get this working with any of the i2v models included in the GPU-poor WanGP v4.5 by DeepBeepMeep.
I get the error:
Error while loading Loras: Lora '{path}' contains non Lora keys '{trunc(invalid_keys,200)}'
same error with wan gp: Error while loading Loras: Lora '{path}' contains non Lora keys '{trunc(invalid_keys,200)}'
I'm a beginner, how do I use this? I use it like a regular LoRA and it only generates one image — how do I make several that form a sequence?
For the video generation, you should firstly use the defult workflow in ComfyUI which are official made. Then, the batchs of image of latent empty should be the frames that you want to generate.
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.