RedShark News - Video technology news and analysis

Prompt attention: How to get the best results with generative AI prompts for video

Written by David Winter | Dec 11, 2024 3:00:00 PM

Generative AI video creation can seem simple on the surface, but there's a whole world of detail and complexity to it that can supercharge your results. 

Unlike traditional tools with their familiar buttons and sliders, AI video generators work through conversation—specifically, through carefully crafted text instructions called prompts.

But here's the challenge: although the conversational aspect feels human-like, these AI systems don't think like human directors, cinematographers, or editors.

Simple prompts can create great results for a brief clip or as an example of AI in action, but for any serious or longer-term video projects, there are more factors at play. They must be explicitly guided to translate your creative vision into compelling video.

And that's what we aim to help you achieve with this article. 

Understanding AI's Visual Language

When planning a shoot with a camera operator, you might say, "Let's make sure we get a dramatic shot of the product." 

Some elaboration might be needed around the specifics, but overall it gives a sense of what’s desired because humans can draw on shared cultural understanding and professional experience to interpret what "dramatic" means in this context.

Or, if your collaborator doesn't understand, they will ask, which currently AI does not do unless specifically prompted. 

AI doesn't have this intuitive understanding. Instead, it learns by analyzing patterns in millions of videos and their descriptions. It’s not operating intuitively but instead algorithmically.

To get the desired results, you need to speak its language by breaking down each element of your vision into specific, concrete details that map to its training data.

Another way to look at it is simply as a form of maths. "Dramatic", in this instance, is the answer to the equation. Our role as the prompter is not to provide the answer but to provide the question that leads to it. X + Y + Z = Dramatic. 

But how do we know the right question to ask? First, let's look at a simple example:

Human direction 

“Make it dramatic." 

AI-friendly prompt:

"Light the subject from a 45-degree angle with high contrast between shadows and highlights, using a slow upward camera tilt from a low angle."

The second version works better because it translates the abstract concept of "dramatic" into specific visual elements the AI can understand and reproduce. There is little or no subjectivity to it. Lighting (X) + contrast (Y) + Camera movement (Z) = Dramatic! 

Building Blocks of Effective Prompts

If prompts are conversational maths, then it makes order and structure essential. Let's break down the components:

1. Setting the Scene

This is your foundation. Just like an actual video shoot, you need to establish:

  • Where is this taking place?
  • What time of day is it?
  • What's the lighting situation?
  • What's in the background?

Examples 

Let's say you're trying to create a simple shot of a modern office.

Poor: 

"Office scene" 

 

Better:


"Modern open-plan office with floor-to-ceiling windows, early morning sunlight streaming in, desks and workstations visible but out of focus in the background"

Why it matters: 

The AI needs these details to create a coherent environment. Without them, it will make random choices that might not match your vision.

2. Defining Action and Motion

This is where you establish what's happening in your video. Be specific about:

  • What's moving?
  • How is it moving?
  • How fast is the movement?
  • What's the timing?

Examples: 

Poor: 

"A person walks through the office." 

Better: 

"Professional woman in her 30s walks confidently through the office at a steady pace, smartphone in hand, other employees visible but blurred in the background as she passes."

Why it matters: 

AI systems need explicit instructions about movement to create natural-looking animation and maintain consistency throughout the clip.

3. Technical Specifications

These details ensure your video meets your specific look, tone, and quality requirements:

  • Resolution and frame rate
  • Camera movement and angles
  • Depth of field
  • Color grading preferences

Example: 

Poor: 

"Make it high quality." 

Better:

 "Generate in 4K resolution (3840x2160) at 24fps, with shallow depth of field. Camera slowly tracks from left to right at shoulder height, maintaining medium shot framing. Use warm color grading with slightly lifted blacks for a cinematic look."

Why it matters: 

Without these specifications, the AI will make default choices that might not match your needs or industry standards.

Learning Through Iteration: Refining Your Prompts

Let's walk through the process of improving a prompt, understanding what each change adds to the final result.

Examples

Starting point:

"Create a tech product reveal for a smartphone."

Why this isn't effective:
  • No information about the product's appearance.
  • No guidance on lighting or atmosphere.
  • No technical specifications.
  • No style reference.

First improvement

"Generate a premium smartphone reveal in a minimalist studio setting. The device is matte black with chrome edges, first silhouetted by edge lighting."

What we added and why:
  • Product details (matte black, chrome edges) → Helps AI create consistent product appearance
  • Setting (minimalist studio) → Establishes a clean, professional environment
  • Lighting concept (edge lighting, silhouette) → Creates dramatic reveal effect
  • Premium positioning → Influences overall quality and style

Further refinement: 

"Create a 6-second premium smartphone reveal. Start with dramatic edge lighting silhouetting a matte black device with polished chrome edges. The camera moves clockwise while ascending slightly (15 degrees). Match modern tech aesthetic with navy and electric blue accent colors. Generate in 4K at 24fps with smooth camera movement."

NOTE: Website compression will adjust the example's technical specs.

 

New elements and their purpose:
  • Duration (6 seconds) → Ensures proper timing
  • Specific camera movement → Creates dynamic but controlled motion
  • Color scheme → Establishes brand-appropriate atmosphere
  • Technical specs → Ensures broadcast-quality output

 

Final version: 

"Generate a cinematic product reveal for a premium smartphone. Begin with a silhouetted profile of a matte black device with polished chrome edges. Key light at 45 degrees creates dramatic edge definition. Camera moves clockwise at 20 degrees per second while ascending at 15 degrees. Background features subtle blue and purple gradient with floating particles. Style references: Apple and Samsung promotional material. Technical specs: 4K resolution, 24fps, cinematic motion blur, 2.39:1 aspect ratio."

NOTE: Website compression will adjust the example's technical specs.

 

Final improvements and their impact:

  • Precise lighting angles → Ensures consistent, professional lighting
  • Specific movement rates → Creates smooth, controlled motion
  • Style references → Helps AI understand the desired production value
  • Aspect ratio → Matches cinematic format
  • Environmental details → Adds depth and interest


Platform-Specific Considerations

Different AI video generators have distinct strengths and requirements and are evolving at an astonishing rate. That's why we've tried to keep this article focussed on universally helpful prompt advice.

But we do have some advice on adapting your prompts for maximum effectiveness for some of the major players in the space... at least as of the publishing date. If you're reading this in 6 months, it may well be a whole new prompting world out there! 

Adobe Firefly

Optimized for:

Commercial and brand content 

Key strengths:

  • Strong integration with Creative Cloud
  • Built-in commercial licensing
  • Professional-grade output

Prompt tips:
  • Use detailed lighting descriptions
  • Specify brand-aligned color schemes
  • Include technical export specifications
  • Begin with shot-type descriptions (close-up, wide shot, etc.)
  • End with a specific aesthetic direction

 

Runway Gen-2

Optimized for:

Creative and artistic applications 

Key strengths:
  • Advanced motion control
  • Style transfer capabilities
  • Artistic effects

Prompt tips:
  • Specify frame interpolation settings
  • Include detailed motion parameters
  • Reference artistic styles clearly
  • Use temporal coherence instructions for smooth motion
  • Combine concrete and abstract elements in prompts

 

Pika Labs

Optimized for:

Character animation

Key strengths:
  • Natural movement
  • Character consistency
  • Narrative sequences
 
Prompt tips:
  • Provide detailed character descriptions
  • Specify exact movement patterns
  • Include clear environmental context
  • Address proportion and scale explicitly
  • Maintain consistent character traits across scenes

 

Runway Gen-1

Optimized for:

Environmental and nature scenes

Key strengths:
  • Photorealistic environments
  • Natural phenomena
  • Atmospheric effects

Prompt tips:
  • Detail geographic locations
  • Specify lighting conditions
  • Include atmospheric elements
  • Reference real-world color palettes
  • Focus on environmental dynamics

Common Pitfalls and How to Avoid Them

1. Overcomplicating Prompts

Problem:

Adding too many elements makes it hard for the AI to prioritize what's important. 

Solution:

Break complex scenes into sequential prompts, focusing on one central aspect at a time.

Example

Complex prompt:

"Create a scene with a person walking through a busy city while it's raining at night with neon lights reflecting in puddles and cars passing by with their headlights creating lens flares while the camera does a complex movement."

Break down into:
  1. First prompt: Establish the environment (rainy night city scene with neon lights)
  2. Second prompt: Add the main character and their movement
  3. Third prompt: Add secondary elements (passing cars, reflections)
  4. Final prompt: Specify camera movement

2. Vague Instructions

Problem:

Abstract or subjective terms lead to unpredictable results. 

Solution:

Translate creative concepts into specific, measurable elements.

Example

Vague prompt:

"Make it moody."

Clearer:

"Use low-key lighting with a 4:1 contrast ratio, deep shadows, and a cool color temperature of around 5600K. Add subtle volumetric fog in the background."

3. Inconsistent Characters

Problem:

Character appearance varying between generations. 

Solution:

Create a character template with specific, reusable descriptions.

Example

Character base:
  1. Age: 30-35
  2. Height: 5'8" (1.73m)
  3. Build: Athletic, medium frame
  4. Hair: Shoulder-length brown hair with subtle waves
  5. Clothing: Charcoal grey tailored suit, white shirt, no tie
  6. Movement: Confident stride, measured pace
  7.  Expression: Neutral, professional demeanor

Looking Forward

Crafting effective prompts for generative AI video creation isn't just about typing out a wish list—it's a deliberate and structured process that mirrors the precision of professional filmmaking.

By translating abstract ideas into concrete, detailed instructions, you're not just asking the AI to create a video; you're guiding it, frame by frame, to execute a specific vision or concept. Your ability to direct the AI will ultimately determine the quality and impact of the final output.

Generative AI is not a replacement for human creativity—nor should we ever aim for it to be—but hopefully, this guide helps you tap into its potential as a collaborator. 

tl;dr

  •  Effective communication with AI video generators involves translating abstract concepts into specific visual details, ensuring that prompts are clear and actionable.
  • Structuring prompts around key components—setting, action, and technical specifications—establishes a solid foundation for generating desired video content.
  • Provide detailed contextual information, such as location, time of day, and background elements, to help the AI create a coherent and aligned environment with your vision.
  • Continuously refining prompts based on the AI's output helps improve the precision and quality of the generated videos, ensuring they meet your creative expectations.