UNIT 3

Generative Videos: AI Tools in Video Making

1. Definition: Generative videos use AI to automatically create video content by synthesizing

2. AI Tools: Include platforms for creating videos with pre-set templates, AI avatars, voiceovers,
and animations.

3. Automation: Reduce manual editing by using AI to handle transitions, scripting, and syncing.

Working of AI Video Makers

1. Input Processing: Users provide text, images, or voice scripts.

2. AI Synthesis: The tool generates video elements, such as virtual presenters, animations, or
voiceovers.

3. Customizable Templates: Allow users to select pre-designed formats and styles.

4. Export Options: Generated videos can be exported in various resolutions and formats.

Benefits of AI Video Makers

1. Time-Saving: Automates video creation, reducing production time.

2. Cost-Effective: Eliminates the need for expensive equipment or professional studios.

3. Scalability: Easily create multiple videos for different audiences.

4. Customization: Offers personalized content with dynamic elements like AI avatars and
voiceovers.

5. Accessibility: No prior technical skills required.

Popular AI Video Makers

1. Synthesia

2. Pictory

3. DeepBrain AI

4. Lumen5

5. Runway ML
Introduction to Synthesia

1. What is Synthesia?

o A leading AI video creation platform that uses AI avatars and voice synthesis to
produce professional videos.

2. Purpose: Ideal for creating instructional, marketing, or training videos.

Features of Synthesia

1. AI Avatars: Customizable, lifelike avatars that present video content.

2. Multi-Language Support: Create videos in over 120 languages.

3. Templates: Pre-designed templates for professional video formats.

4. Voice Synthesis: Realistic AI-generated voiceovers.

5. Ease of Use: Intuitive drag-and-drop interface.

Who Should Use Synthesia?

1. Businesses: For marketing and training materials.

2. Educators: To create engaging learning content.

3. Content Creators: For scalable video production.

4. HR Teams: For onboarding and internal communications.

Compatibility of Synthesia

1. Platforms: Web-based; works on any browser.

2. Integration: Compatible with tools like PowerPoint and LMS platforms.

3. File Types: Supports video export in MP4 and other popular formats.

Pros and Cons of Synthesia

Pros:

1. Easy-to-use interface.

2. Wide language support.

3. Cost-effective for businesses.

4. No need for professional video expertise.

1. Limited avatar customization.

2. Reliance on internet connectivity.

3. Voiceovers may lack emotional depth.

How to Use Synthesia

1. Sign Up: Create an account on the Synthesia website.

2. Choose a Template: Select a video format or template.

3. Customize Content: Input text, select avatars, and set voiceovers.

4. Preview: Review and make adjustments.

5. Export: Download the final video.

How to Make AI Videos in 10 Minutes

1. Login: Access your Synthesia account.

2. Select Template: Pick a template matching your needs.

3. Add Script: Input your script in the desired language.

4. Customize Avatars and Backgrounds: Choose avatars and set backgrounds.

5. Generate Video: Let the AI process the video.

6. Download: Save and use your video.

Practical Case Studies of Synthesia

1. Corporate Training: A company used Synthesia to create multilingual onboarding videos,

2. Education: An online educator created course videos in 5 languages, increasing global reach.

3. Marketing Campaigns: A startup used Synthesia for ad videos, cutting production time by
75%.
 UNIT-5

What is Generative AI?

1. Definition: Generative AI refers to artificial intelligence systems capable of generating new

2. Key Examples: Text generation (ChatGPT), image synthesis (DALL-E), and music creation.

3. Core Mechanism: It uses deep learning techniques like neural networks to model and
generate outputs.

AI vs ML vs DL vs Generative AI

1. Artificial Intelligence (AI)

o Broad field encompassing all technologies that simulate human intelligence.

o Includes reasoning, problem-solving, decision-making, and learning.

2. Machine Learning (ML)

o A subset of AI focused on training systems to learn patterns from data and make
predictions or decisions.

o Example: Recommendation systems, fraud detection.

3. Deep Learning (DL)

o A subset of ML using artificial neural networks to process vast amounts of data.

o Highly effective for image recognition, natural language processing, and speech
recognition.

4. Generative AI

o A specialized area within AI that creates new content (text, images, videos).

o Examples: ChatGPT, MidJourney, and Synthesia.

How OpenAI ChatGPT or LLaMA3 and LLM Models are Trained

1. Data Collection:

o Collect vast datasets from books, articles, codebases, and online resources.

o Include diverse languages and domains to ensure coverage.

2. Preprocessing:

o Tokenize text data into smaller chunks.

3. Model Training:

o Transformers: Use architectures like GPT or BERT with attention mechanisms.

o Optimization: Train on GPUs/TPUs using gradient descent.

o Fine-Tuning: Specialize the model on specific tasks or domains.

4. Evaluation and Testing:

o Validate the model's performance using benchmarks and user feedback.

5. Deployment:

o Serve models through APIs or integration with applications.

Evolution of LLM Models

1. Initial Models: Early models like word2vec focused on word embeddings.

2. BERT: Introduced bidirectional context understanding.

3. GPT: Focused on generative capabilities with autoregressive transformers.

4. LLama: Lightweight, fine-tunable models designed for efficiency and performance.

5. Next-Gen Models (Gemini, GPT-4, LLaMA3): Multimodal capabilities, better contextual

Getting Started with Google Gemini

1. Overview:

o Google Gemini is a next-generation AI model developed by Google, known for its

o Combines text, image, and video understanding in one model.

2. Features:

o Advanced natural language processing.

o Multimodal data handling (text, image, audio, video).

o Real-time adaptability and better reasoning capabilities.

3. Getting Started:

o Access through Google Cloud or supported platforms.

o Use APIs for integration into custom applications.

1. Define Objectives:

o Build a chatbot capable of answering user queries conversationally.

2. Set Up Google Gemini:

o Access Gemini Pro API via Google Cloud.

o Obtain API keys for authentication.

3. Prepare Data:

o Collect or create a dataset for Q&A pairs relevant to your domain.

o Format the data for model training or fine-tuning.

4. Model Configuration:

o Configure the Gemini Pro model for conversation-style output.

o Set parameters like context length and response style.

5. Development:

o Frontend: Build a user interface for the chatbot (e.g., using React or Flask).

o Backend: Integrate the Gemini API for query processing.

6. Test the Chatbot:

o Validate the chatbot's responses against test queries.

o Refine prompts or adjust model configurations for better accuracy.

7. Deploy:

o Host the chatbot on a cloud platform.

o Integrate it into web or mobile applications.

8. Monitor and Update:

o Collect feedback and improve the chatbot's performance iteratively.