All projects05 / 43
AI & Data
Minimoes
Conversational AI avatar platform
At a glance
- Discipline
- AI & Data
- Key features
- 6 delivered
- Screens
- 4 in gallery
Next.jsReactTypeScriptMongoDBWebSocketsSSE
Visit siteMinimoes
Overview
A real-time 3D digital-twin platform with streamed AI conversations, synchronised speech and lip-synced facial animation.
I took a partially working avatar platform to a stable production system. Responses stream in over Server-Sent Events, while audio playback, text rendering and avatar animation are coordinated independently through a dedicated speech state machine.
Lip sync is driven by real audio timestamps rather than estimates, and an embeddable widget brings the avatar to any website with text and voice queries. Mock analytics were replaced with real MongoDB-backed metrics and CSV export.
Key features
- 01Streaming AI conversations over SSE
- 02Audio-synchronised text rendering with requestAnimationFrame
- 03Blendshape lip sync aligned to real playback timing
- 04Embeddable avatar widget for external websites
- 05Multi-step Create Twin flow: voice, avatar, knowledge, agent
- 06Real analytics with latency metrics and feedback tracking
Gallery
Inside the product
4 screens
Up next06 / 43
Legend
AI football performance & coaching platform
