Beta

CONTINUUM Guide

Generate broadcast-ready video prompts with AI. This guide walks you through creating your first prompt.

What This Does

Continuum transforms your shot descriptions into detailed, platform-optimized prompts for AI video generators like Google Veo 3, Kling, and Sora. It handles the technical translation so you can focus on the creative.

You describe the shot. Continuum returns a production-ready prompt with camera direction, lighting cues, and platform-specific formatting. It also analyzes your shot and recommends which AI video platform will render it best—with confidence scores and explanations.

AI-generated Porsche 911 driving through tunnel

This frame was generated using a Continuum prompt → Veo 3

View full gallery →

Get Started

  1. Go to continuum.video
  2. Sign in with Gmail or create an account with email authentication.
  3. You'll receive beta tokens to generate prompts. Generation costs 10 tokens; refinements cost 2.
Sign Up →

Set Up Your Brand

Brands keep your work organized and teach the AI your context. Each brand is isolated—your Porsche prompts never mix with your Tesla prompts.

Brands screen with numbered callouts
  1. Click "Brands" in the top navigation.
  2. Click "+ CREATE NEW BRAND" to add a new brand.
  3. Add a description — this context informs how prompts are written. Be specific: "High-end kitchen appliances, premium aesthetic, stainless steel focus" beats "appliances."

You'll select your brand on the Generate screen when creating prompts. Use "None" for non-brand-specific work.

Generate Your First Prompt

Generate screen with numbered callouts
  1. Select Brand — Choose from your saved brands or "None."
  2. Describe your shot — Be specific about action, camera, and style. Example: Red 911 headed South on PCH during Golden Hour. Camera is tracking next to car during entire clip. Vehicle moves screen right to screen left. Driver in car wearing sunglasses. Aspect ratio 16:9.
  3. Review Platform Recommendations — Continuum analyzes your shot and ranks which AI video platforms will handle it best. You'll see confidence percentages and explanations (e.g., "Best for motion-heavy automotive shots").
  4. Choose Target Platform — Accept the top recommendation or override based on your preference or subscription access.
  5. Set Clip Duration — 7 seconds is standard for commercial shots.
  6. Click "Generate Prompt" — Takes ~10 seconds.
Token costs: Generating a new prompt uses 10 tokens. Refining an existing prompt uses 2 tokens. Your balance shows in the top-left.

Use Your Generated Prompt

Generated prompt output with Copy button

After generation, you'll see a detailed prompt optimized for your target platform:

  1. Copy your prompt — Click "Copy" to grab the full generated prompt, then paste it into your video AI platform (Veo 3, Kling, Sora, etc.) and render your video.

Not quite right? Update your shot description with what you want changed, then click "Refine Prompt" to iterate.

Agent learning: Each prompt you generate teaches Continuum about your brand. The "Agent learning activated" message means your next prompt for that brand will be informed by this one.

Tips for Better Results

How to use Refine Prompt

Refine lets you iterate without starting over. First, edit your shot description with what you want changed—more motion, different angle, new lighting. Then click "Refine Prompt." The system compares your updated description against the last generated prompt and adjusts accordingly. Important: if you click Refine without changing your description first, you'll get the same output and still be charged.

Trust the platform recommendations

The confidence scores are based on shot type analysis. If Veo 3 shows 87% for automotive and Sora shows 65%, that gap matters. Use the top pick unless you have a specific reason not to.

Be specific about vehicles

If there's a car, request a driver. AI video often renders empty cars that look wrong. Same for boats, planes—anything that should be operated.

Include camera direction

Terms like "tracking shot," "low angle," "dolly in," "aerial pullback" translate well. Describe placement and movement.

Specify time of day

"Golden hour," "overcast midday," "dusk with city lights" dramatically affect the output quality and mood.

Add style references

Film noir, claymation, anime, documentary, 70mm cinematic, VHS aesthetic—style keywords help the AI commit to a look.

Current Limitations

Continuum generates prompts—the final video quality depends on the AI video platform you use. Here's what to expect:

Fingers and hands

AI video still struggles with accurate hand rendering. Count fingers in your output. Re-render if needed.

Text and typography

On-screen text, signs, and logos often render incorrectly or illegibly. Plan to add text in post.

Physics realism

Complex motion, reflections, and object interactions may look unnatural. Simpler compositions render better.

Consistency across renders

Same prompt, different results. You may need multiple renders to get the shot you want. This is normal.

Send Feedback

This is a beta. We want to know what's working, what's broken, and what's missing.

Report issues or suggestions

Email: feedback@continuum.video