Back to Documentation

AI Agents

Use AI agents to create scripts and edit videos using natural language

AI Agents

TellFrame features two powerful AI agents that help you create and edit videos using natural language. The Script Agent helps you create video scripts from content, while the Editor Agent helps you modify and enhance existing videos.

Overview

TellFrame includes two specialized AI agents:
  1. Script Agent - Creates video scripts from content, ideas, or uploaded files
  2. Editor Agent - Edits and enhances existing videos using natural language
Both agents understand natural language and can execute complex tasks automatically, making video creation and editing faster and more intuitive.

Script Agent

The Script Agent helps you create video scripts through a conversational process. It can work with various content sources and generates structured scripts ready for video production.

What It Does

  • Creates video scripts from ideas, content, or uploaded files
  • Processes documents - PDFs, Word docs, spreadsheets, articles
  • Extracts content from URLs (Google Docs, Sheets, blog posts, social media)
  • Generates structured scripts with scenes, timing, and keywords
  • Creates data visualizations from spreadsheets (charts and KPI cards)
  • Optimizes for platforms - Tailors scripts for YouTube, TikTok, Instagram, etc.

How to Use

Starting a Script

  1. Open Script Tab
    • Go to the Script tab when creating a new video
    • The Script Agent chat interface appears
  2. Describe Your Idea
    • Type what you want to create
    • Examples:
      • "Create a video about cloud storage"
      • "Make a tutorial on baking sourdough bread"
      • "Create a history lesson about the Industrial Revolution"
  3. Or Upload Content
    • Click the upload button
    • Select a file (PDF, DOC, CSV, Excel, etc.)
    • Or paste a URL (Google Docs, blog post, etc.)
  4. Review Proposal
    • Agent proposes a script structure
    • Includes: title, duration, angle, script type
    • Review and request changes if needed
  5. Confirm Generation
    • Say "yes", "go ahead", or "generate it"
    • Agent creates the complete script
    • Script is ready for video creation

Content Sources

File Uploads

Supported Formats:
  • PDFs - Documents and slide decks
  • Word Documents - .docx files
  • Spreadsheets - CSV and Excel files
  • Text Files - Plain text content
How It Works:
  1. Upload your file
  2. Agent processes and extracts content
  3. Analyzes structure and key points
  4. Proposes script based on content
  5. Generates structured script with scenes

URL Import

Supported Sources:
  • Google Docs - Automatically exports and processes
  • Google Sheets - Converts to CSV and analyzes data
  • Blog Posts - Extracts article content
  • News Articles - Processes news stories
  • Social Media - Twitter, LinkedIn, Instagram posts
How It Works:
  1. Paste URL in chat
  2. Agent extracts content automatically
  3. Processes and analyzes content
  4. Proposes script structure
  5. Generates script when confirmed

Spreadsheet Data

When you upload CSV or Excel files:
  • Data Analysis - Analyzes structure and patterns
  • Chart Generation - Suggests appropriate chart types
  • KPI Creation - Identifies key metrics for KPI cards
  • Visualization - Creates frames with charts and KPIs
  • Narrative - Generates script that explains the data

Script Generation Process

Stage 1: Consultation

When you first describe your idea:
  1. Agent Acknowledges - Responds to your topic
  2. Strategic Insight - Shares why the topic is relevant
  3. Proposes Structure - Suggests:
    • Duration and word count
    • Narrative angle
    • Script structure
    • Key messaging points
  4. Shows Proposal - Displays in clean format:
    • Subject
    • Title
    • Duration
    • Angle
    • Script Type
  5. Waits for Confirmation - Doesn't generate yet

Stage 2: Refinement

If you request changes:
  • Discusses Modifications - Adjusts proposal based on feedback
  • Asks Clarifying Questions - About messaging, tone, structure
  • Presents Updated Proposal - Shows revised structure
  • Continues Discussion - Until you're ready
Not Confirmations:
  • Requests for changes ("make it shorter", "change platform")
  • Questions about proposal
  • Modifications or refinements

Stage 3: Generation

Only when you explicitly confirm:
  1. Confirms Action - "Perfect! Generating your script now..."
  2. Calls Tool - Executes script generation automatically
  3. Creates Structure - Generates:
    • Full script text
    • Scene breakdown (frames)
    • Timing for each scene
    • Keywords for media search
    • Word highlighting (optional)
    • Charts/KPIs (if data provided)
  4. Completes - "✅ Your script is ready!"

Script Structure

Generated scripts include:
Full Script:
  • Complete narration text
  • Platform-optimized
  • Engaging and conversational
Frames (Scenes):
  • Title - Scene name (e.g., "Hook", "Context", "Solution")
  • Duration - Time in seconds
  • Script - Narration text for this scene
  • Keywords - 2-4 keywords for media search
  • Word Colors - Optional highlighting
  • Word Sizes - Optional size variations
  • Charts/KPIs - If data visualization needed
Plan:
  • Overview of script structure
  • Section descriptions
  • Timing breakdown

Data Visualization

When spreadsheets are uploaded:
Automatic Analysis:
  • Identifies data structure
  • Finds interesting insights
  • Suggests visualizations
  • Creates narrative around data
Chart Generation:
  • Bar charts for comparisons
  • Line charts for trends
  • Pie charts for proportions
  • Area charts for cumulative data
  • And more chart types
KPI Cards:
  • Key metrics extraction
  • Trend identification
  • Format selection (currency, percentage, etc.)
  • Auto-positioning in layout
Smart Filtering:
  • Different data per scene
  • Filters based on narrative
  • Matches data to script content
  • Avoids showing all data in every frame

Language Support

  • Automatic Detection - Detects your language
  • Consistent Language - Responds in same language
  • Multi-language Scripts - Generates scripts in your language
  • Platform Optimization - Adapts to language-specific platforms

Best Practices

For Best Results:
  • Be Specific - Describe what you want clearly
  • Provide Context - Include target audience, platform, purpose
  • Review Proposals - Check structure before confirming
  • Request Changes - Don't hesitate to refine
  • Use Uploads - Upload files for better context
For File Uploads:
  • Clean Files - Use well-formatted documents
  • Clear Structure - Organized content works best
  • Complete Content - Include all relevant information
  • Proper Format - Use supported file types
For URLs:
  • Public Access - Ensure content is accessible
  • Full URLs - Use complete URLs
  • Google Docs/Sheets - Share settings allow access
  • Quality Content - Better source = better script

Editor Agent

The Editor Agent helps you edit and enhance existing videos using natural language. It understands your video's current state and can make changes automatically.

What It Does

  • Modifies Scenes - Add, remove, or update scenes
  • Changes Text - Edit text content, styling, positioning
  • Manages Media - Add, replace, or remove backgrounds
  • Adjusts Settings - Change format, effects, audio
  • Adds Elements - Create background elements, charts, KPIs
  • Updates Styling - Fonts, colors, animations, layouts
  • Searches Media - Finds stock or user library media
  • Generates Content - Creates AI images, videos, components

How to Use

Accessing the Editor Agent

  1. Open Your Video
    • Open a video in the editor
    • Click the AI Agent button or panel
    • Editor Agent chat appears
  2. Describe What You Want
    • Type your request in natural language
    • Examples:
      • "Make the text bigger and change it to blue"
      • "Add background music to this scene"
      • "Change the transitions to fade"
      • "Add a logo to the top right corner"
  3. Agent Executes
    • Understands your request
    • Executes appropriate tools
    • Makes changes automatically
    • Confirms what was done

Understanding Video State

The Editor Agent has access to:
Current Scenes:
  • All text blocks/scenes
  • Their content and IDs
  • Positions and timing
  • Types and properties
Video Settings:
  • Format (portrait, landscape, square)
  • Duration and frame count
  • Audio tracks and volumes
  • Voice settings
  • Video effects
Elements:
  • Background elements
  • B-roll elements
  • Logo
  • AI-generated components
Context Awareness:
  • Knows what's in your video
  • Understands scene positions
  • Tracks element IDs
  • Monitors current state

Common Tasks

Text Editing

Examples:
  • "Make the title text larger"
  • "Change all text to white"
  • "Make the text in scene 2 blue"
  • "Bold the important words"
What It Does:
  • Updates text content
  • Changes font sizes
  • Modifies colors
  • Adjusts styling
  • Updates word highlighting

Scene Management

Examples:
  • "Add a new scene with text 'Welcome'"
  • "Delete the second scene"
  • "Duplicate the first scene"
  • "Move scene 3 to position 1"
What It Does:
  • Creates new scenes
  • Removes scenes
  • Reorders scenes
  • Duplicates scenes
  • Updates scene properties

Media Management

Examples:
  • "Add a background image to scene 1"
  • "Replace the background with a video"
  • "Add stock video of mountains"
  • "Use my uploaded cat image"
What It Does:
  • Searches stock media
  • Searches your library
  • Adds media to scenes
  • Replaces existing media
  • Applies to multiple scenes

Styling Changes

Examples:
  • "Change all transitions to fade"
  • "Add slide animation to text"
  • "Use modern subtitle style"
  • "Apply card layout to all scenes"
What It Does:
  • Updates transitions
  • Changes animations
  • Modifies subtitle styles
  • Updates layouts
  • Applies effects

Element Creation

Examples:
  • "Add a countdown timer"
  • "Create a progress bar"
  • "Add floating particles"
  • "Generate animated gradient background"
What It Does:
  • Creates background elements
  • Generates AI components
  • Adds dynamic elements
  • Positions elements automatically

Advanced Features

Voice Input

How to Use:
  1. Click microphone button
  2. Speak your request
  3. Audio is transcribed
  4. Text appears in input
  5. Send to execute
Supported:
  • Natural speech
  • Multiple languages
  • Automatic transcription
  • Real-time processing

Reference Images

How to Use:
  1. Click camera button
  2. Upload images (up to 5)
  3. Or select from library
  4. Ask agent to use them
  5. Agent generates based on images
Use Cases:
  • "Generate an image like this"
  • "Create a video with this style"
  • "Use these as reference for generation"

Multi-Scene Operations

Examples:
  • "Make all text blue"
  • "Add fade transition to all scenes"
  • "Change font size in scenes 1, 2, and 3"
What It Does:
  • Applies changes to multiple scenes
  • Uses efficient batch operations
  • Maintains consistency
  • Updates all at once

Smart Features

Auto-Execution

When intent is clear:
  • Executes Immediately - No confirmation needed
  • Batch Operations - Handles multiple changes
  • Efficient Processing - Optimizes tool calls
  • Smart Chaining - Links related operations
Example: "Add stock image of mountains to scene 1, 2, and 3" → Searches media → Shows results → Adds to all three scenes

Context Understanding

Scene References:
  • "Scene 1" → First scene
  • "Scene 2" → Second scene
  • "Last scene" → Final scene
  • "Scene with 'Hello'" → Finds by content
Element References:
  • "The gradient background" → Finds AI element
  • "The countdown timer" → Finds by type
  • "Background element 1" → Finds by ID

Clarification

When ambiguous:
  • Asks Questions - Clarifies intent
  • Shows Options - Presents choices
  • Waits for Input - Doesn't guess
  • Confirms Understanding - Verifies before acting

Examples

Example 1: Text Styling

Request: "Make the title bigger and change it to blue"
Agent Actions:
  1. Identifies title scene
  2. Increases font size
  3. Changes text color to blue
  4. Confirms changes
Result: Title is larger and blue

Example 2: Media Addition

Request: "Add a stock video of mountains to scene 2"
Agent Actions:
  1. Searches stock media for "mountains"
  2. Shows preview results
  3. Waits for selection (or auto-selects if clear)
  4. Adds video to scene 2
  5. Confirms addition
Result: Mountain video added to scene 2

Example 3: Multi-Scene Update

Request: "Change all transitions to fade and make text white"
Agent Actions:
  1. Updates transitions for all scenes
  2. Changes text color to white for all scenes
  3. Confirms both changes
Result: All scenes have fade transitions and white text

Example 4: Element Creation

Request: "Create an animated gradient background"
Agent Actions:
  1. Generates animated gradient component
  2. Adds as background element
  3. Positions automatically
  4. Confirms creation
Result: Animated gradient background added

Key Differences

Script Agent vs Editor Agent

Script Agent:
  • Purpose: Create new video scripts
  • Input: Ideas, content, files, URLs
  • Output: Structured script with scenes
  • Process: Consultation → Refinement → Generation
  • Focus: Content creation and structure
Editor Agent:
  • Purpose: Edit existing videos
  • Input: Natural language requests
  • Output: Video modifications
  • Process: Request → Execute → Confirm
  • Focus: Video editing and enhancement

When to Use Each

Use Script Agent When:
  • Creating a new video from scratch
  • You have content to convert
  • You need a structured script
  • Working with data/spreadsheets
  • Starting from an idea
Use Editor Agent When:
  • You have an existing video
  • You want to make changes
  • You need to add elements
  • You want to modify styling
  • You're refining a video

Best Practices

For Script Agent

  1. Be Clear - Describe what you want specifically
  2. Provide Context - Include audience, platform, purpose
  3. Review Proposals - Check structure before confirming
  4. Use Uploads - Upload files for better results
  5. Request Refinements - Don't hesitate to ask for changes
  6. Confirm Explicitly - Say "yes" or "generate" when ready

For Editor Agent

  1. Be Specific - "Scene 2" is better than "that scene"
  2. One Task at a Time - Break complex requests into steps
  3. Use Natural Language - Write as you would speak
  4. Review Changes - Check that agent understood correctly
  5. Reference Elements - Use clear descriptions
  6. Test Incrementally - Make changes step by step

Tips for Success

Script Agent Tips

  • Start with Ideas - Describe your concept clearly
  • Upload Content - Files provide better context
  • Review Proposals - Check structure matches your vision
  • Request Changes - Refine until perfect
  • Confirm When Ready - Explicitly say "generate"

Editor Agent Tips

  • Reference by Position - "Scene 1", "Scene 2" works well
  • Be Specific - "Make text blue" not "change color"
  • Use Examples - "Like scene 1" helps clarify
  • Check Results - Verify changes match intent
  • Combine Requests - Can handle multiple changes

Troubleshooting

Script Agent Issues

Script not generating?
  • Ensure you explicitly confirmed ("yes", "generate")
  • Check that proposal was accepted
  • Try requesting generation again
Content not processing?
  • Verify file format is supported
  • Check file isn't corrupted
  • Ensure URL is accessible
  • Try a different file/URL
Script structure wrong?
  • Request refinements before confirming
  • Be specific about changes needed
  • Review proposal carefully

Editor Agent Issues

Changes not applying?
  • Check that request was clear
  • Verify scene/element exists
  • Try being more specific
  • Check for error messages
Agent not understanding?
  • Use clearer language
  • Reference scenes by number
  • Break into smaller requests
  • Check current video state
Tool execution failing?
  • Verify video state is correct
  • Check element IDs exist
  • Try simpler request first
  • Review error messages

Advanced Usage

Script Agent Advanced

Data-Driven Scripts:
  • Upload spreadsheets with data
  • Agent analyzes and creates visualizations
  • Generates narrative around data
  • Creates charts and KPIs automatically
Multi-Language:
  • Write in your language
  • Agent responds in same language
  • Script generated in your language
  • Maintains language consistency
Platform Optimization:
  • Specify target platform
  • Agent optimizes script structure
  • Adjusts duration and pacing
  • Uses platform-appropriate style

Editor Agent Advanced

Complex Operations:
  • "Make scenes 1-3 have blue text and fade transitions"
  • "Add logo to all scenes in top right"
  • "Create animated background and apply to scene 2"
Element Management:
  • "Update the gradient background to be more vibrant"
  • "Move the countdown timer to bottom center"
  • "Change KPI card colors to match brand"
Batch Updates:
  • "Apply card layout to all scenes"
  • "Change all text to white"
  • "Add fade transition between all scenes"
Both AI agents make video creation and editing faster and more intuitive. Use the Script Agent to create content, and the Editor Agent to perfect it!