Consistency in Video Generative Models: from Clip to Wild (CVM @ AAAI '26)

Submission Instructions

Common Video Requirements (All Categories)

• Resolution: 1920x1080 or higher recommended

• Format: MP4

• Frame Rate: 24fps or 30fps recommended

• Duration:

  • Single shot: 3-5 seconds recommended

  • Full scene: 15-25 seconds recommended

Naming Conventions

• Use lowercase letters and numbers only

• No spaces or special characters allowed

• Must strictly follow the specified format for each category

Category-Specific Requirements

1. Intra-clip World Knowledge Category

• Generate videos based on prompts from JSON file

• Naming format: {ID}.mp4

2. Inter-Shot Category

Json Structure

• Scene Basic Information:

  • scene_id: Unique identifier (e.g., scene001)

  • scene: Scene name

  • style: Scene style

• Storyboard: Contains 5 consecutive shots

  • Each shot has an ID (1-5) and description

Submission Options

A. Full Scene Submission

• Combine all 5 shots into one complete video

• Naming format: scene{id}_full.mp4

• Example: scene001_full.mp4

B. Individual Shot Submission

• Submit each shot as separate videos

• Naming format: scene{id}_shot{number}.mp4

• Example:

  • scene001_shot1.mp4

  • scene001_shot2.mp4

  • scene001_shot3.mp4

  • scene001_shot4.mp4

  • scene001_shot5.mp4

3. Inter-clip Camera Category

• Generate videos based on prompts from JSON file

• Naming format: {ID}.mp4