1. Workflow Overview

This workflow, named "Wan2.1 Start/End Frame Video Stabilization (Industry Standard)", generates smooth transitional animations from start/end frames with:

• Frame interpolation between input images

• Video stabilization via Wan2.1 model

• 4x super-resolution (4xFaceUpSharpDAT)

• Frame rate boosting (RIFE VFI)

2. Core Models

3. Key Nodes

4. Workflow Structure

1. Input & Preprocessing

   • Input: Start frame (output (6).png) and end frame (output (7).png)

   • Processing:

     • Resize images (default: 480x768)

     • Extract features via WanVideoClipVisionEncode

2. Video Generation

   • Load Wan2.1 model and VAE

   • Generate intermediate frames (total: 81 frames ≈5s @16fps) using WanVideoSampler

3. Post-Processing

   • Upscaling: 4x super-resolution with 4xFaceUpSharpDAT

   • Interpolation: Boost to 32fps via RIFE VFI

   • Export: MP4 video (H.264, CRF=19)

5. Inputs & Outputs

• Input Parameters:

  • Required: Start/end frame images (PNG/JPG)

  • Optional: Total frames (default:81), resolution (recommended ≤768px), prompts (e.g., "cartoon girl turning")

• Output: Video file (AnimateDiff_*.mp4 in ComfyUI/output)

6. Notes

1. Dependencies:

   • Manual download required for Wan2.1 and RIFE models.

   • Super-resolution models via ComfyUI Manager.

2. Hardware:

   • Recommended: ≥12GB VRAM (NVIDIA GPU).

   • High frame counts (e.g., 81) may take 10-30 minutes.

3. Troubleshooting:

   • Reduce resolution/frame rate if errors occur.

   • Enable super-resolution or adjust prompts for facial artifacts.