Tem um workflow do ComfyUI? Gere uma documentação legível direto no navegador.Torne um workflow legível.Experimentar agora ↗
👑 Wan2.1-VACE-14B (LoRA Accelerated): 10x Speed with CausVid LoRA for 3-Step Video Creation
Perfil de execução
Descrição da fonte
👑 Wan2.1-VACE-14B (LoRA Accelerated): 10x Speed with CausVid LoRA for 3-Step Video Creation
🎬 Skyrocket Your Video Creation: Achieve ~10x Speed with Wan2.1 & the CausVid LoRA! 🎬
📌 Overview
The Wan2.1-VACE-14B video diffusion model, when supercharged by the CausVid LoRA , is designed for high-quality, highly efficient video generation. It particularly excels at 480p and 720p resolutions through a streamlined 3-step ComfyUI workflow. This guide will walk you through the setup process to unlock this accelerated video generation capability, including options for full precision and quantized models like the fast Q3KL GGUF.
🔑 Key Components
Diffusion Model (14B):
Comentário gerado por IA
Explicação gerada por IA com base nos detalhes da fonte e da configuração. As sugestões são claramente identificadas.
Wan2.1-VACE-14B(pro) is a ComfyUI workflow for generating 480p or 720p video clips from a text prompt or an image, using the Wan2.1-VACE-14B model with CausVid LoRA and producing a combined video output.
Once the models are loaded, enter a text prompt, sample with KSampler or an equivalent node using the selected model and LoRA, then send the result to Video Combine.
Use a text prompt for text-to-video or an image for image-to-video.
Video Combine is the named output stage for the generated video clip.
Place the diffusion model in ComfyUI/models/diffusion_models, the text encoder in ComfyUI/models/text_encoders, the LoRA in ComfyUI/models/loras, and the VAE in ComfyUI/models/vae.
Sugestão · não verificado
The setup names these model files: wan2.1_vace_14B_fp16.safetensors, wan2.1_vace_14B_Q4KM.safetensors, wan2.1_vace_14B_Q3kl.gguf, Wan21_CausVid_14B_T2V_lora_rank32.safetensors, wan_2.1_vae.safetensors, q3kl.gguf, umt5_xxl_fp16.safetensors, and umt5_xxl_fp8_e4m3fn_scaled.safetens
Named node packs include ComfyUI_Comfyroll_CustomNodes, comfyui_controlnet_aux, comfyui_essentials, comfyui_layerstyle, comfyui-custom-scripts, comfyui-easy-use, and comfyui-frame-interpolation.
Other named node packs include ComfyUI-GGUF, comfyui-kjnodes, comfyui-ollamagemini, comfyui-videohelpersuite, rgthree-comfy, and teacache.
Choose wan2.1_vace_14B_fp16.safetensors, wan2.1_vace_14B_Q4KM.safetensors, or wan2.1_vace_14B_Q3kl.gguf, pair it with Wan21_CausVid_14B_T2V_lora_rank32.safetensors, and apply the LoRA at a strength around 1.0.
Sugestão · não verificado
Use umt5_xxl_fp16.safetensors for compatibility with existing examples; umt5_xxl_fp8_e4m3fn_scaled.safetensors is the smaller fp8 option and can save VRAM.
Sugestão · não verificado
If you choose wan2.1_vace_14B_Q3kl.gguf, verify that a GGUF loader is available; the setup says this option requires one.
Sugestão · não verificado
Before relying on ComfyUI-GGUF, comfyui_controlnet_aux, q3kl.gguf, or teacache, verify the exact name and installation source for each.
Sugestão · não verificado
Isso precisa ser editado?
Faça login para enviar uma solicitação de edição.
Fontes
1 fonteTrechos de fonte
1 trechoSource context: 1714 downloads · Type Workflows · Base model Wan Video 14B i2v 720p
Full Precision: wan2.1_vace_14B_fp16.safetensors (Recommended for compatibility with LoRA examples)
Quantized (Civitai): wan2.1_vace_14B_Q4KM.safetensors
Quantized (GGUF - Civitai): wan2.1_vace_14B_Q3kl.gguf (Used in the 5-min example, requires GGUF loader)
This isn't the same as the GGUF format from Hugging Face (they missing !). I tested that one, and it didn't work for vid2vid tasks. So, I developed my own types specifically designed to work well with vid2vid. These are optimized and structured differently to ensure compatibility and better results , if u need another type do "a comment" after test hugging face one!
Performance LoRA (Essential for Speed):
Wan21_CausVid_14B_T2V_lora_rank32.safetensors
VAE:
wan_2.1_vae.safetensors
Text Encoder: Choose one:
umt5_xxl_fp16.safetensors (Recommended to match Kijai's wrapper compatibility for the LoRA workflow)
umt5_xxl_fp8_e4m3fn_scaled.safetensors (Smaller, fp8 version)
📁 File Organization
Place the downloaded files in the following structure within your ComfyUI directory:
ComfyUI/ ├── models/ │ ├── diffusion_models/ │ │ └── wan2.1_vace_14B_fp16.safetensors # Or Q4KM.safetensors, or Q3kl.gguf │ ├── text_encoders/ │ │ └── umt5_xxl_fp16.safetensors # Or the fp8 version │ ├── loras/ │ │ └── Wan21_CausVid_14B_T2V_lora_rank32.safetensors │ └── vae/ │ └── wan_2.1_vae.safetensors
🎨 Model Showcase: Rapid 720p Cinematic Shots
This setup, featuring Wan2.1-VACE-14B and the CausVid LoRA, excels at producing 720p (and 480p) video clips with remarkable speed, even faster with quantized GGUF models. It's ideal for quick iterations, creative experimentation, and efficient content creation, all streamlined by a 3-step workflow.
💡 Usage Tips
Model & LoRA Configuration: For maximum speed and quality, ensure you are using the appropriate 14B model (e.g., wan2.1_vace_14B_fp16.safetensors or wan2.1_vace_14B_Q3kl.gguf ) paired with the Wan21_CausVid_14B_T2V_lora_rank32.safetensors LoRA. The LoRA should be applied with a strength typically around 1.0.
Text Encoder: The umt5_xxl_fp16.safetensors text encoder is recommended for best compatibility with existing examples and Kijai's original demonstrations. The fp8 version can save VRAM.
Resolution: This setup is optimized for 480p and 720p video generation.
Performance Gains:
Without LoRA (fp16): An 81-frame 720p video might take ~40 minutes on an RTX 4090.
With CausVid LoRA (fp16): The same video can be generated in ~4 minutes on an RTX 4090.
With CausVid LoRA & Q3KL GGUF: Potentially even faster, around 5 minutes or less for similar output on capable hardware with a GGUF loader.
Workflow Simplicity: The primary advantage, beyond speed, is the reduction to a 3-step generation process once models are loaded. This typically involves: 1. Prompting (Text Input), 2. KSampler (or equivalent node with LoRA and chosen model), 3. Video Combine (Output).
🌟 Credits & Acknowledgements
Original Wan 2.1 models repackaged for ComfyUI by Comfy-Org : Wan 2.1 ComfyUI Repackaged on Hugging Face . The performance-boosting CausVid LoRA ( Wan21_CausVid_14B_T2V_lora_rank32.safetensors ) was extracted and shared by Kijai . Original announcement and details: Kijai's Reddit Post . Quantized GGUF and Safetensors versions available on Civitai, enabling broader accessibility and speed. Gratitude to the developers of the underlying CausVid technique (presumably available under an MIT License or similar open terms).
👨💻 Developer Information
This guide was created by Abdallah Al-Swaiti:
Hugging Face
GitHub
ComfyUI-OllamaGemini
For additional tools and updates, check out my other repositories.
✨ Create Dreamy Videos with WAN 2.1 VACE and Pastel Dream! ✨
Estimativa de requisito VRAM
Estimativa indisponível
Os tamanhos de arquivos do modelo ainda não estão completos o suficiente para calcular este requisito.
Requisitos
Requisitos 17q3kl.gguf
Não resolvidoCheckpoint · Unknown
Text encoder · Hugging Face · Comfy-Org/Wan_2.1_ComfyUI_repackaged · Comfy-Org/Wan_2.1_ComfyUI_repackaged · Hugging Face
umt5_xxl_fp8_e4m3fn_scaled.safetensors
Arquivo não verificadoText encoder · SAFETENSORS · Unknown
VAE · Hugging Face · Comfy-Org/Wan_2.1_ComfyUI_repackaged · Comfy-Org/Wan_2.1_ComfyUI_repackaged · Hugging Face
Pacote de nós · Registry
comfyui_controlnet_aux
PossívelPacote de nós · Registry
Pacote de nós · Registry
Pacote de nós · Registry
Pacote de nós · Registry
Pacote de nós · Registry
Pacote de nós · Registry
ComfyUI-GGUF
PossívelPacote de nós · Registry
Pacote de nós · Registry
Pacote de nós · Registry
Pacote de nós · Registry
Pacote de nós · Registry
teacache
Não resolvidoPacote de nós · Registry
