Cartoon Diary

Turn your daily life into a fun cartoon

Comment

Inspiration

"What if my daily life could be a comic strip?" The project began with a simple desire to transform mundane daily routines into vibrant, memorable illustrations. We wanted to create a space where the most impactful moment of the day is captured in a single cartoon frame. Furthermore, we aimed to move beyond simple keyword searches, envisioning a diary that understands the context of our memories through natural language queries.

What it does

Cartoon Diary allows users to set up a personal avatar and document their daily experiences. The AI then selects the most significant event from the entry and illustrates it in a unique cartoon style. Users can manage their memories effortlessly using Semantic Search. Instead of remembering dates, you can simply ask, "When did I go out for fruit?" and the system will retrieve the relevant diary entries based on the meaning of your question.

How we built it

We built a modern, full-stack architecture designed for seamless user experience. The Frontend was developed using React, ensuring a fully responsive UI that automatically optimizes for both PC and mobile environments. The Backend was powered by Python, where we implemented the LangChain framework to orchestrate complex AI workflows. For the core AI capabilities, we integrated various Amazon Nova models and hosted the entire infrastructure on AWS for high scalability and reliability.

Challenges we ran into

Real-time Communication: Implementing Server-Sent Events (SSE) proved difficult; we faced several issues receiving server events correctly, requiring multiple architectural refactors.

AWS Optimization: Selecting the most cost-effective and efficient AWS services for our specific workflow was a constant balancing act.

Orchestration & QA: Using LangGraph for AI orchestration introduced complexities in Quality Assurance, leading to extensive testing phases to ensure stable logic flows.

Prompt Engineering with Nova: Since the Nova image generation model had specific constraints regarding system prompts, we underwent rigorous prompt engineering iterations to ensure the output aligned with the user’s intent.

Accomplishments that we're proud of

Consistent Character Rendering: We successfully tuned the prompts so that the generated characters strictly follow the user's initial configuration.

Intuitive Semantic Search: By implementing an Embedding Model, we achieved a seamless search experience that feels like talking to a friend rather than searching a database.

What we learned

Advanced Prompt Engineering: We gained deep insights into crafting precise and clear prompts to get the best performance out of generative models.

AWS Ecosystem: This project was a great opportunity to explore various AWS services and identify the best tools for rapid app deployment.

AI Orchestration: We significantly leveled up our knowledge in AI Orchestration (LangChain/LangGraph) and understanding the unique characteristics of the Nova model family.

What's next for Cartoon Diary

Multi-panel Narratives: Moving beyond a single frame to support multi-scene comic strips for more detailed storytelling.

Visual Consistency: Further refining the model to ensure even higher consistency in characters and artistic style across multiple days.

Supporting Cast: Adding a "Character Library" feature where users can save and reuse recurring characters (friends, family, or pets) in their diary entries.

Built With

Share this project:

Updates