Multimodal Generative AI Case Study: Generating Marketing Creatives for Bakery

This repository contains a Jupyter Notebook detailing a case study that leverages Multimodal Generative AI to produce professional-quality digital marketing materials for a small bakery, specializing in brownies and tres leches cakes.

The project demonstrates a cost-effective, AI-driven approach to content creation, addressing the common resource constraints faced by small businesses.

The Challenge

Browbake, a growing bakery, needs a strong online presence but lacks the time and budget for traditional professional marketing and design services. The goal is to use AI to streamline the creation of high-impact promotional content that showcases their rich brownies and creamy tres leches cakes.

Project Objectives

The primary goal is to use different Generative AI models to create a cohesive, ready-to-use marketing video package:

1. Promotional Poster: Generate a visually appealing promotional poster that highlights Browbake's unique offerings, exploring both a no-code and a low code-based solution.

2. Promotional Video & Tagline: Create an engaging short video showcasing the cakes and generate a catchy, brand-aligned tagline.

3. Voiceover Generation: Produce a fitting, natural-sounding voiceover for the video that enhances the brand tone.

4. Final Deliverable: Combine the generated visual (poster/video), tagline, and voiceover into a cohesive marketing video for social media campaigns.

Technologies & Key Libraries

This project uses a variety of Python libraries and generative models to handle different modalities (text, image, audio, and video):

Modality Task Libraries/Models

• Image Generation Promotional Poster; diffusers, Stable Diffusion XL (SDXL Base 1.0)
• Video & Text Video Generation, Tagline; google-genai (for Google's generative models like Veo/Gemini) = Audio/Speech Voiceover Generation; parler-tts, soundfile
• Video Editing Combining Assets; moviepy.editor (e.g., VideoFileClip, AudioFileClip)
• Utilities Display, Image Manipulation; torch, Pillow, mediapy

Getting Started

Prerequisites

To run the notebook locally, you will need:

1. A Google API key for access to the Google Generative AI models.

2. Python 3.x environment.

3. A strong GPU is recommended for faster execution of the large Stable Diffusion XL and video models.

Setup

1. Clone the repository:

   Bash

   git clone [Your Repository URL] cd [Your Repository Name]

2. Install the dependencies: The notebook uses a specific set of libraries, which can be installed using the command found within the first code cell:

   Bash

   !pip install torch torchvision diffusers soundfile parler-tts Pillow moviepy pydub google-genai mediapy -q Note: After installation, a kernel restart is recommended before proceeding with the rest of the notebook.

3. Run the Notebook: Open Multimodal_GenerativeAI_Browbake_Case_Study_Notebook.ipynb using Jupyter Notebook or Google Colab and execute the cells sequentially. You will need to set your API key in the appropriate cell to initialize the Google Generative AI client.