Presento: AI Presentation and Podcast Maker

Overview

This application leverages the power of Vertex AI Agents API and the Imagen text-to-image model to generate presentations and podcasts based on a user-provided topic.

Sample PDF	Sample Podcast	Try Locally
Try in Notebook

Features

• Vertex Agents for Content Creation:
  • Story Generation: Creates a comprehensive story related to the user's topic.
  • Slide Generation: Structures the story into a basic slide deck format.
  • Slide Refinement: Iteratively improves the slide deck's clarity, conciseness, and engagement.
  • Image Description Generation: Creates detailed prompts for image generation tailored to each slide's content.
  • JSON Conversion: Transforms the refined slide deck text into a structured JSON format for easier processing.
  • Podcast Generation: Generates a podcast conversation between a host and a guest expert based on the slide content.
• Imagen for Visuals: Generates relevant and engaging images for each slide based on the AI-generated descriptions.
• PDF Generation: Compiles the slides, including titles, descriptions, key takeaways, and images, into a downloadable PDF presentation.
• Podcast Synthesis: Synthesizes the generated podcast conversation using Google Cloud Text-to-Speech, with different voices for the host and guest.
• Gradio Interface: Provides a user-friendly interface for topic input, refinement level selection, and presentation and podcast download.

Architecture

graph LR
    A[User Input] --> B(Story Generation)
    B --> C(Slide Generation & Refinement)
    C --> D(JSON Conversion)
    D --> E(Image Description & Generation)
    E --> F(Podcast Generation)
    F --> G(PDF and Podcast Generation)

Loading

1. User Input: The user provides a presentation topic through the Gradio interface.
2. Story Generation: A Gemini agent generates a detailed story relevant to the topic.
3. Slide Generation & Refinement: Another agent converts the story into a slide deck, which is then refined iteratively by a refinement agent.
4. JSON Conversion: The refined slide deck is converted to JSON format.
5. Image Description & Generation: An agent creates image descriptions for each slide. Imagen then uses these descriptions to generate relevant images.
6. Podcast Generation: An agent generates a podcast conversation based on the slide deck content.
7. PDF and Podcast Generation: The application creates a PDF presentation from the slides and images and synthesizes the podcast audio using Google Cloud Text-to-Speech.

Getting Started

1. Set up
   • Set the PROJECT_ID and LOCATION variables in the code.
   • Enable the Vertex AI API, Text-to-Speech API.
2. Install Dependencies:

   pip install -r requirements.txt

3. Run the Application:

   gradio main.py 

4. Deploy to Cloud Run

gcloud run deploy presento  --project <PROJECT_ID>  --port 8080 --region us-central1 --min-instances 1 

Please note some of the services here are either private preview,public preview and GA. Please refer documentation for more details.