AI Video Generation Guide

How It Works

Describe your scene, the camera movement, and the style, plus the platform you're using (each has slightly different prompt conventions). The guide returns a structured prompt and settings, plus guidance on stitching multiple clips together consistently.

What to Provide

Video Prompt

Replace [[tokens]] with your details. This is the finished deliverable.

Prompt Structure

[[Subject and action, specific: e.g. "a woman in a red coat walks through falling snow"]], [[camera movement, e.g. "slow tracking shot following from behind"]], [[setting/lighting, e.g. "dusk, streetlights just turning on"]], [[style, e.g. "cinematic, shallow depth of field, 35mm film look"]], [[duration if the tool supports specifying it]]

Multi-Clip Consistency (for a longer sequence)

To keep character/setting consistent across multiple generated clips:

1. Use the same detailed subject description verbatim in every clip's prompt.
2. Where the tool supports it, use the last frame of clip N as the starting reference for clip N+1: this prevents jarring jumps between clips.
3. Lock the style/lighting description across all clips in the sequence.

Aspect Ratio by Platform

Worked Examples

Example 1: Product Demo Clip

Prompt: "close-up of hands unboxing a sleek white product box on a marble surface, soft studio lighting, slow reveal, camera slowly pushing in, cinematic commercial style --ar 16:9"

Example 2: Character Intro for a Series (9:16, Multi-Clip)

Prompt (clip 1): "young woman with curly red hair, green jacket, standing on a city rooftop at sunset, wind moving her hair, camera static, cinematic golden hour lighting --ar 9:16"

Prompt (clip 2, using clip 1's last frame as reference): Same character description repeated verbatim, camera now slowly zooms out to reveal the city skyline, same lighting description.

Common Mistakes to Avoid

• Vague action descriptions that produce unpredictable, often warped motion
• Changing character/style descriptions between clips in a sequence, breaking visual consistency
• Wrong aspect ratio for the platform, requiring an awkward crop or pillarbox
• Overloading one prompt with too many simultaneous actions: AI video tools handle one clear action much better than several
• No plan for stitching clips, resulting in jarring jumps between scenes

Keep each clip's action simple and specific, and lock your character/style wording across every clip in a sequence.