I took out many things most would not understand and confuse everyone... This can still do a lot. If enough people like this flow i can fix up the full ones that fixes many of the issues with this flow but it would...
Perfil de execução
Descrição da fonte
simplified wan 2.1 vace
I took out many things most would not understand and confuse everyone... This can still do a lot. If enough people like this flow i can fix up the full ones that fixes many of the issues with this flow but it would take time and i dont know if its worth it with the pace of LTX taking over. LTX cant do some things like skipping, or flips but its really good at everything else.
This is not fast, its a slow, and difficult process to make this work for the things i show... I have done this a LONG time and know how to manipliate the flow to get results, I spent 8 hours to 4 days for each video... Multiple takes, multiple batches... On a 3060. This is why i never understood the hate i have recieved :P. I worked harder to make the ai copy than the people making the originals usually. I now use a 5060 and 2.1 vace still takes ages.
This flow was designed to be used with my face fix inpaint flow that fixes faces.
Fontes
1 fonteTrechos de fonte
1 trechoSource context: 651 downloads · Type Workflows · Base model Wan Video 14B t2v
Faces in this output can be .... not great. Vace face inpaint is actually VERY good.
Going further than 4 batches you will notice a significant drift in contrast. This is worse with a static camera, anything static on the screen will drift and not look good after 6 batches. Moving cameras work well because it wont over contrast the areas remaining the same. PIDnet copies background movement.
That being said nothing copies like vace... yet.
Remaining hardware or download-size estimates are not available yet.
Requisitos
depth_anything_v2_vitl.pth
Não resolvidoCheckpoint · Unknown
dw-ll_ucoco_384_bs5.torchscript.pt
EncontradoCheckpoint · 129 MB · PT · Unknown
lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank256_bf16.safetensors
LoRA · Unknown
lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank32_bf16.safetensors
Não resolvidoLoRA · Unknown
Object Detector · Hugging Face
umt5_xxl_fp8_e4m3fn_scaled.safetensors
Arquivo não verificadoText encoder · Unknown
VAE · Hugging Face · Comfy-Org/Wan_2.1_ComfyUI_repackaged
wan2.1_vace_1.3B_fp16.safetensors
EncontradoCheckpoint · 135 B · Hugging Face · memescreamer/Wan_2.1_ComfyUI_repackaged
Checkpoint · 17.4 GB · GGUF · Unknown
Checkpoint · 208 MB · PT · Unknown
Pacote de nós · Registry
ComfyUI_Comfyroll_CustomNodes
EncontradoPacote de nós · Registry
comfyui_controlnet_aux
PossívelPacote de nós · Registry
Pacote de nós · Registry
comfyui-easy-use
EncontradoPacote de nós · Registry
comfyui-gguf
PossívelPacote de nós · Registry
comfyui-impact-pack
EncontradoPacote de nós · Registry
comfyui-impact-subpack
EncontradoPacote de nós · Registry
Pacote de nós · Registry
comfyui-videohelpersuite
EncontradoPacote de nós · Registry
masquerade
Não resolvidoPacote de nós · Registry
masquerade-nodes-comfyui
EncontradoPacote de nós · Registry
pr-was-node-suite-comfyui-47064894
Não resolvidoPacote de nós · Registry
rgthree-comfy
EncontradoPacote de nós · Registry