chfm
/

WanVideo_comfy / LongVie2 /README.md
chfm's picture
Duplicate from Kijai/WanVideo_comfy
3d39a5e
|
Raw
History Blame Contribute Delete
601 Bytes
LongVie2:
Source project: https://huggingface.co/Vchitect/LongVie2
`LongVie2_attn_layers_bf16.safetensors` replaces base I2V model's attention layers, can be used as extra_model, this ensures the full weights of the original.
`longvie2_attn_layers_lora_rank_64_bf16.safetensors` This is a LoRA extracted from the attention layer diff, slightly weaker effect than full weights but functional.
Use either of those, not both.
`LongVie2_dual_controller_controlnet_bf16.safetensor` is the controlnet model weights, loaded as extra_model in WanVideoWrapper and used with `Add Dual Control Embeds` -node