AvatarAI
An interactive API project
Description:
AvatarAI is a prototype project I created during an internship, serving as a prototype for potential future integration into other software to facilitate a more interactive Q&A interaction between users and the AI agent. By integrating the APIs from ChatGPT and GooeyAI, I created an animated PNG avatar that provides a dynamic and immersive user experience. Rather than offering standard AI responses, this avatar speaks in a realistic manner, incorporating distinctive personality traits, styles, and even popular slang.
Role:
Internship Project
Date:
May ~ Jun, 2023
Skills:
API Integration (ChatGPT, Gooey AI)
Avatar Animation
Conversational AI Development
User Experience (UX) Design
Testing & Iteration
Note: This video contains audio.
Inspiration
I've always been fascinated by the potential of AI and the myriad ways it can be utilized to improve and enrich our lives. This project was particularly inspired by the idea of creating an AI interaction that goes beyond the conventional, one that could mimic human conversation with uncanny realism. The challenge of creating an AI with distinctive, human-like personality traits was an exciting proposition that drove this project forward.
Chatgpt ai chat
D-ID ai chat
Development process
Explore the capabilities of the Gooey API
The development process of the entire project started by exploring the capabilities of the Gooey API and sketching out the initial UI layout on Glitch. The goal was to create an engaging interactive platform that allows users to easily interact with AI.
Testing Gooey AI's animation capabilities and initial UI placement
Exploration with ChatGPT
Next, the focus shifted towards experimenting with the GPT-3 model's output to incorporate certain personality traits and styles. The aim was to enhance the human-like nature of AI responses, with adjustments for sarcasm, usage of slang, and even mimicking personalities like Anthony Jeselnik. This step involved understanding and exploiting the GPT-3's versatility.
Testing Prompt
"Respond to the question like Anthony Jeselnik with VERY dark humor, EXTRA FUNNY, EXTRA sarcastic, stand up comedy style, using POPULAR slang words."
enter
"How are you doing today?"
Result
GPT-3 Response Example 1 : Oh, I'm fine, my friend. I mean, who doesn't like being surrounded by constant reminders of our own mortality? Life is nothing but an endless parade of pain and disappointment. But hey, I guess I can't complain, right?
11:45 PM
GPT-3 Response Example 2 : Since you want to know so much, how many times have you cried today? Oh, and don’t forget, how many existential crises have you had since breakfast? Let’s keep it lighthearted, okay?
11:45 PM
Integration and Final Coding
The final stage revolved around integrating the APIs from ChatGPT and GooeyAI, synchronizing them to create an animated PNG avatar. Overcoming various tech nical challenges during this stage not only solidified my problem-solving abilities but also expanded my technical skills.
Additionally, I implemented a "CHANGE" button to allow users to switch the "personality" or "individual" they are interacting with.
Reflection
Learning Experience
This project was an enriching journey that significantly deepened my understanding of AI and machine learning principles. I had the opportunity to delve into effective API integration, honing my skills in prompt engineering. I further explored the capabilities of language models like ChatGPT and improved my full-stack development capabilities by working with JavaScript and Node.js.
Project Insights
Creation of AI Speech Avatar Prototype: The effective combination of various APIs led to the development of the company's first-ever AI-generated speech avatar prototype. This opens new avenues for user interaction and engagement and diversifies the company's AI product line.
Enhanced Conversational Personality: The adjustments I made in prompt engineering refined the AI's conversational style to exhibit more customized characteristics, such as sarcasm, deviating from the typical bot-like responses. This unique approach has the potential to elevate user satisfaction and engagement over time.
Scalability and Iteration: Through my hands-on work on both the front-end and the back-end, I've laid the groundwork that makes the system more scalable and makes it easier to make changes in small steps. Because of what I did, the system is now better set up for long-term adaptability and stability. This makes it easier to add new features and make changes in the future.