Guide: Generating Veo 3 & Sora Prompts

Video generation models like Google's Veo 3, OpenAI's Sora, and Runway Gen-3 Alpha require a different prompting structure than image generators. They rely heavily on sequential action, camera movement, and temporal consistency.

The Anatomy of a Video Prompt

To get the best results out of Veo 3 or Sora, your prompt should follow this hierarchy:

1. Cinematography & Format: Is it a 35mm film shot, a drone flyover, or a macro shot?
2. Subject & Action: Who or what is the focus, and exactly what are they doing over time?
3. Camera Movement: How does the camera move? (e.g., "Slow pan from left to right," "Push in steadily on the subject's face").
4. Lighting & Atmosphere: Golden hour, cinematic lighting, volumetric fog.

How ProPeek Automates This

Writing these prompts from scratch is difficult and time-consuming. ProPeek automates the process using a technique called Reverse Engineering.

Step-by-Step Workflow:

• Find a Reference: Find a stock video that has the exact camera movement and lighting you want.
• Upload to ProPeek: Drop the video into the ProPeek workspace.
• Select 'Video to Prompt': ProPeek will extract keyframes and send them to the Vision AI.
• Get Your Prompt: The AI will return a highly structured prompt formatted specifically for models like Veo 3. It will describe the temporal changes and camera moves it observed in the reference video.

Using Inspire Mode for Unique Generations

If you want to use the camera movement from a reference video but need a completely different subject, use ProPeek's Inspire Mode. It will keep the cinematographic instructions intact but generate a new subject and setting, giving you a completely unique prompt ready for Veo 3.