InspirationThe idea for TalkSnap AI came from a desire to make AI interactions more human, more expressive, and more personal. Most AI assistants today are still just “voices in a box” — no face, no emotion, no real presence.
We asked ourselves:
“What if an AI could talk to you the way a friend would — with a face, expressions, personality, and emotion?”
This sparked the vision for TalkSnap AI, an app that transforms simple photos into a fully animated 3D avatar that talks, listens, smiles, reacts, and evolves with you.
What it doesTalkSnap AI transforms your ordinary photos into a fully animated 3D AI avatar that you can talk to in real time — just like speaking with a virtual human.
Once your avatar is generated, TalkSnap AI becomes your interactive AI companion, capable of:
🎭 Creating a 3D Avatar From Your Photos
Upload 1–4 photos
The app reconstructs your 3D face, texture, and expressions
A complete 3D avatar (GLB) is generated automatically
Blendshapes allow realistic facial expressions & lip movement
How we built it3D Avatar Generation Pipeline
We built a custom photo → avatar pipeline:
User uploads 1–4 images
DECA reconstructs a 3D head model
Facial expressions and ARKit-style blendshapes are generated
The avatar is exported as GLB
Rendered in the app using BabylonJS / Three.js
This gives each user a unique, expressive avatar.
Challenges we ran into
Accomplishments that we're proud of
What we learnedAI & ML Engineering
3D face reconstruction
Blendshape systems
Viseme phoneme mapping
LLM behavior design
🎨 3D Animation
GLB optimization
WebGL blendshape animation
Syncing animation with audio timing
📱 Mobile Engineering
Handling WebView + native audio together
Managing background threads for STT/TTS
Keeping performance smooth on mid-range devices
🌍 Multilingual AI
Whisper + multilingual TTS
LLM translation + localization
What's next for Untitled
Built With
- apikey
- openai
- supacloud