Evados

/

DiffSynth-Studio-Lora-Wan2.1-ComfyUI

like 63

https://www.kaggle.com/code/rahul7star/wan2-1-lora-experiment/input

Do you have any plans for wan2.1 1.3b depth control lora?

Depth Control LoRA and Tile lines LoRA already exist for the 1.3B model.
These LoRAs are not made by me, but I can upload them here.
Normally, to use these LoRAs, you need to perform two generation passes with an advanced sampler.
You apply the Depth Control LoRA in the first pass, and you can either use a booster LoRA in the second pass, or just use the regular sampler on last one.

Otherwise, there's the VACE Wan 2.1 model, which I believe can also handle this, though I haven't tried that model yet.
Personally, I really like using I2V or T2V along with my personal LoRAs.

But if you were actually talking about the LoRAs and not the VACE model, I’ll upload them here so you can download them.
Just to be clear, again, they’re not mine.

You can find it here:
https://huggingface.co/Evados/DiffSynth-Studio-Lora-Wan2.1-ComfyUI/blob/main/wan2.1-1.3b-control-lora-depth-v0.1_comfy.safetensors
https://huggingface.co/Evados/DiffSynth-Studio-Lora-Wan2.1-ComfyUI/blob/main/wan2.1-1.3b-control-lora-tile-v0.2_comfy.safetensors
https://huggingface.co/Evados/DiffSynth-Studio-Lora-Wan2.1-ComfyUI/blob/main/wan2.1-1.3b-control-lora-tile-v1.0_comfy.safetensors

Thank you so much , will give it a try !!

For DG_Wan2_1_V1.3b_Fun_Inp to use as Lora , whats the base model ? I tried

lora "DG_Wan2_1_V1.3b_Fun_Inp/dg_wan2_1_fun_inp_Extra_new.safetensors" with base model "Wan-AI/Wan2.1-T2V-1.3B-Diffusers"

and got checkpoint error ( I am running via gradio )

your other Lora exp 'Wan2.1-1.3b-lora-speedcontrol-v1_new.safetensors" are working fine with base model "Wan-AI/Wan2.1-T2V-1.3B-Diffusers"

I know gradio is not best to run all these but given that your other models are working so I wan to keep going with same settings and try out new models :)

dg_wan2_1_fun_inp_Extra_new.safetensors is not a lora, it is a full model...