Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Kamal-Nayan-Kumar/AI-Video-Gen/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

The ContentGenerator class is responsible for creating the initial presentation structure. It uses Google’s Gemini AI to generate a complete slide-by-slide breakdown including titles, content text, and metadata about visual needs (images/animations).

​
Class Definition

from generators.content_generator import ContentGenerator

content_gen = ContentGenerator()

​
Constructor

def __init__(self)
Initializes the content generator with Google Gemini AI configuration. Configuration:
  • Model: Uses Config.GEMINI_MODEL (configured in environment)
  • Response Format: JSON (application/json MIME type)
  • API Key: Configured via Config.GEMINI_API_KEY

​
Methods

​
generate_content

Generates complete presentation structure with slide content and visual requirements. 
def generate_content(topic: str, num_slides: int = 5) -> Dict
​
topic
string
required
The presentation topic or subject matter
​
num_slides
int
default:"5"
Number of slides to generate
​
return
Dict
Presentation content structure with slides data
Returns structure: 
{
  "topic": "Newton's Laws of Motion",
  "total_slides": 5,
  "slides": [
    {
      "slide_number": 1,
      "title": "Introduction to Newton's Laws",
      "content_text": "Overview of the three fundamental laws",
      "needs_image": false,
      "image_keyword": "",
      "needs_animation": false,
      "animation_description": "",
      "duration": 6.0
    }
  ]
}
​
topic
string
The presentation topic
​
total_slides
int
Total number of slides generated
​
slides
array
Array of slide objects with content and metadata

​
Data Models

​
SlideContent

Pydantic model representing individual slide content. 
class SlideContent(BaseModel):
    slide_number: int
    title: str
    content_text: str
    needs_image: bool
    image_keyword: str = ""
    needs_animation: bool
    animation_description: str = ""
    duration: float
Validation Rules:
  • Mutual Exclusivity: A slide cannot have both needs_animation=True and needs_image=True
  • Auto-correction: If both flags are true, the generator prioritizes animation and clears image fields

​
PresentationContent

Pydantic model for complete presentation structure. 
class PresentationContent(BaseModel):
    topic: str
    total_slides: int
    slides: List[SlideContent]

​
Usage Example

From backend/app.py:242-243: 
# Step 1: Generate PPT content structure
update_progress(generation_id, 10, "generating_content", 
                "📝 Generating presentation content...")

content_gen = ContentGenerator()
content_data = content_gen.generate_content(topic, request.num_slides)
Output statistics: 
total = len(content_data['slides'])
text_only = sum(1 for s in content_data['slides'] 
               if not s.get('needs_image') and not s.get('needs_animation'))
with_image = sum(1 for s in content_data['slides'] if s.get('needs_image'))
with_animation = sum(1 for s in content_data['slides'] if s.get('needs_animation'))

print(f"Slide breakdown: Text={text_only} Image={with_image} Animation={with_animation}")

​
Content Generation Rules

The generator follows specific rules enforced via the AI prompt:
  1. Text-First Design: 70-80% of slides should be text-only
  2. Image Usage: Sparingly for people, places, objects, and static diagrams
  3. Animation Usage: Very rarely (1-2 max) for concepts requiring motion visualization
  4. Mutual Exclusivity: Slides can have either image OR animation, never both
  5. Duration Estimation: Based on content complexity (typically 4-10 seconds per slide)

​
File Persistence

Generated content is automatically saved to: 
Config.SLIDES_DIR / "{topic_sanitized}_content.json"
Filename Sanitization:
  • Max 30 characters from topic
  • Replaces spaces with underscores
  • Removes special characters (:, /, etc.)

​
Error Handling

The generator includes comprehensive error handling: 
try:
    response = self.model.generate_content(prompt)
    text = response.text.strip()
    
    # Clean markdown formatting
    text = text.replace('json', '').replace('`', '').strip()
    
    # Parse and validate JSON
    content_data = json.loads(text)
    
    # Add missing fields if needed
    if 'topic' not in content_data:
        content_data['topic'] = topic
    if 'total_slides' not in content_data:
        content_data['total_slides'] = len(content_data['slides'])
    
    # Fix mutual exclusivity violations
    for slide in content_data['slides']:
        if slide.get('needs_animation') and slide.get('needs_image'):
            if slide.get('animation_description'):
                slide['needs_image'] = False
                slide['image_keyword'] = ""
            else:
                slide['needs_animation'] = False
                slide['animation_description'] = ""
    
except Exception as e:
    print(f"Content generation error: {e}")
    traceback.print_exc()
    raise

Build docs developers (and LLMs) love

Get started for freeTalk to us