1. Workflow Overview

This workflow utilizes Aliyun's Wan2.1 model for Text-to-Video (T2V) generation. It integrates text encoding, video diffusion, and VAE decoding to produce dynamic video content. Key features:

• Supports Chinese prompts (e.g., "滑雪的男人" - "a man skiing")

• Configurable frame rate (default: 16fps) and resolution (480x768)

• Includes negative prompts for quality filtering

2. Core Models

3. Key Nodes

• LoadWanVideoT5TextEncoder
  Loads the Chinese text encoder (umt5-xxl-enc). Use bf16 precision to save VRAM.

• WanVideoTextEncode
  Processes positive/negative prompts. Example negative prompts filter low-quality content.

• WanVideoModelLoader
  Loads the main video model with options for fp32/fp16 and VRAM optimization.

• WanVideoSampler
  Core sampler parameters:

  • steps: 10 (lower for faster video generation)

  • cfg_scale: 6 (lower for creative freedom)

  • sampler: dpm++

• VHS_VideoCombine
  Combines frames into MP4 video with configurable:

  • Frame rate (16fps)

  • Output format (H.264, CRF=19)

  • Filename prefix (WanVideo2_1_T2V)

4. Workflow Structure

Group 1: Text Processing

• Input: Chinese prompt

• Output: Text embeddings

• Key nodes: LoadWanVideoT5TextEncoder → WanVideoTextEncode

Group 2: Video Generation

• Input: Text embeds + empty image embeds (480x768)

• Output: Latent video data

• Key nodes: WanVideoSampler

Group 3: Video Export

• Input: Decoded image sequence

• Output: MP4 file

• Key nodes: WanVideoDecode → VHS_VideoCombine

5. I/O Specifications

Input Parameters:

• Resolution: 480x768 (set in WanVideoEmptyEmbeds)

• Seed: Fixed/Random (example: 1057359483639287)

• Prompts: Natural Chinese language (avoid complex syntax)

Output:

• MP4 video (saved to ComfyUI output folder)

• Includes generation metadata

6. Notes

⚠️ VRAM Requirements

• Minimum 12GB (16GB recommended)

• Enable offload_device for optimization

⚠️ Model Installation

• Download Wan2.1 models manually from official sources

• Text encoder path: models/wan_t5/umt5-xxl-enc-bf16.safetensors

⚠️ Dependencies

• Requires ComfyUI-WanVideoWrapper & VideoHelperSuite

• Install via ComfyUI Manager