Skip to main content
Wan-Alpha is a specialized text-to-video model that generates high-quality videos with alpha channel transparency. Built on the Wan2.1-14B-T2V base model, it creates videos with transparent backgrounds and semi-transparent objects, perfect for compositing workflows. The model excels at generating transparent backgrounds, semi-transparent objects (bubbles, glass, water), glowing effects, and fine details with proper alpha channels (hair, smoke, particles).

​
Video tutorial

​
Resources

Make sure your ComfyUI is updated.Workflows in this guide can be found in the Workflow Templates. If you can’t find them in the template, your ComfyUI may be outdated.If nodes are missing when loading a workflow, possible reasons:
  1. You are not using the latest ComfyUI version (Nightly version)
  2. Some nodes failed to import at startup

​
Wan-Alpha Text-to-Video Workflow (14B)

​
1. Download the Workflow

Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find “Wan 2.1 Alpha T2V 14B” in the Template Library under Workflow → Browse Templates → Video. Wan-Alpha T2V workflow

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “Wan 2.1 Alpha T2V 14B” in Template Library

​
2. Install Models

All models mentioned can be found at Wan_2.1_ComfyUI_repackaged. Diffusion Models : Choose one version:

wan2.1_t2v_14B_fp8_scaled.safetensors

FP8 scaled diffusion model. Place in ComfyUI/models/diffusion_models/

wan2.1_t2v_14B_bf16.safetensors

BF16 diffusion model. Place in ComfyUI/models/diffusion_models/
Text Encoder

umt5_xxl_fp8_e4m3fn_scaled.safetensors

FP8 text encoder. Place in ComfyUI/models/text_encoders/
VAE Models

wan_alpha_2.1_vae_rgb_channel.safetensors

RGB channel VAE. Place in ComfyUI/models/vae/

wan_alpha_2.1_vae_alpha_channel.safetensors

Alpha channel VAE. Place in ComfyUI/models/vae/
LoRA Models

wan_alpha_2.1_rgba_lora.safetensors

Alpha generation LoRA. Place in ComfyUI/models/loras/

lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors

Lightning LoRA for fast inference. Place in ComfyUI/models/loras/
File storage locations:

​
3. Complete the Workflow Step by Step

  1. Ensure the correct diffusion model is loaded in the Load Diffusion Model node
  2. Ensure the Load CLIP node has loaded umt5_xxl_fp8_e4m3fn_scaled.safetensors
  3. Ensure the Load VAE node for RGB has loaded wan_alpha_2.1_vae_rgb_channel.safetensors
  4. Ensure the second Load VAE node for Alpha has loaded wan_alpha_2.1_vae_alpha_channel.safetensors
  5. Ensure the LoRA Loader node has loaded wan_alpha_2.1_rgba_lora.safetensors
  6. (Optional) Load lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors in the Lightning LoRA node for faster generation
  7. Set your positive prompt in the CLIP Text Encode (Positive Prompt) node
  8. (Optional) Modify video dimensions in the EmptyHunyuanLatentVideo node
  9. Click the Run button or use Ctrl(Cmd) + Enter to execute generation