Skip to main content

What is an AI Scenario?

The AI Scenario is the narration script that drives your video. When you generate a video, AI reads your entire document — text blocks, headings, code, tables, and screenshots — and creates a structured script that narrates each step of your tutorial.

How it works

The scenario generation uses Claude (via OpenRouter) to:
  1. Analyze document structure — identify the logical flow of steps, sections, and key information
  2. Interpret screenshots — understand what each screenshot shows and how it relates to the surrounding text
  3. Write narration — create natural-sounding narration text for each slide, matching the tutorial tone
  4. Set timing — assign duration and pacing for each slide based on content complexity

Scenario structure

The AI generates a JSON scenario with slides. Each slide contains:
FieldDescription
Narration textWhat the TTS voice will say for this slide
Screenshot referenceWhich screenshot to display
DurationHow long the slide is shown (driven by audio length)
TransitionHow the slide enters/exits

Tips for better scenarios

1

Write clear headings

Headings help AI understand the tutorial structure. Use them to mark distinct steps or sections.
2

Add descriptive text

The more context you provide around screenshots, the better the AI narration. Brief captions or explanations significantly improve output.
3

Use logical ordering

Arrange blocks in the order you want them narrated. AI follows the document flow top to bottom.
4

Keep screenshots focused

Crop screenshots to show only the relevant area. AI narration is more precise when screenshots are focused.
The AI uses up to 8,192 tokens for scenario generation. Very long documents with many screenshots are supported, but keeping documents focused produces better narration quality.

Language handling

The narration language is set when you start generation, not determined by the document text. The AI will:
  • Generate narration in the selected language regardless of the document language
  • Handle mixed-language documents (e.g., English UI with Polish explanations)
  • Use proper SSML markup for natural prosody in the target language
Polish text with special characters is handled automatically. The AI pipeline includes quote repair and Unicode sanitization to prevent JSON parsing errors.

Currently, the scenario is generated and rendered in a single pipeline. To change the narration, edit your document content and regenerate.
AI rewrites your content into spoken narration. It paraphrases, simplifies, and adds transitions to create a natural video flow. The meaning is preserved.
The scenario is generated using Claude via OpenRouter. The model selection is optimized for quality and cost.
If the narration is inaccurate, try adding more descriptive text to your document. The AI relies on context — more detail produces better results.