OpenAI integration
The assistant was shaped using JFK speeches, interviews, and secondary sources to keep responses historically grounded.
Back to projects
Multimodal AI Project
Integrating AI in Recreating Iconic Personalities
A proof-of-concept multimodal AI experience that explored how a historical figure could be represented through LLM behavior, voice synthesis, and animated visual presentation.
The project explores how language models, voice synthesis, and avatar presentation can be combined into a more convincing historical interaction experience.
Demo Preview
Open the interactive demo.
The preview links to footage showing the voice, avatar, and language-model pieces working together.
open demo video
offline
static
live
System Stack
Personality emulation across text, voice, and avatar layers.
The system combines OpenAI-driven personality emulation, ElevenLabs voice synthesis, and a digital portrait to create a more immersive historical interaction than a text-only chatbot.
Voice synthesis
ElevenLabs was used to generate speech output that matched the chosen historical figure's recognizable voice.
Avatar layer
A digital portrait was used to add visual presence, expression, and stronger multimodal immersion.
Challenges and impact
Believability mattered, but so did restraint.
- Historical accuracy had to be balanced against the tendency of generative models to hallucinate.
- Multimodal integration across LLMs, voice, and animation introduced meaningful engineering complexity.
- The educational use case was a major motivation: making history feel conversational and immediate.
Project Links
Demo materials
Demo footage and the original concept pitch for the project.