SELF-HOSTED AI
Gemma Eve
Zero-cost AI voice agent powered by Gemma 12B + Chatterbox + FlashHead
Whisper STT
→
Gemma 12B
→
Chatterbox TTS
→
FlashHead Video
Eve Cyberdyne Systems
Gemma 12B Brain
Google's open-weight LLM running locally via Ollama. Full reasoning, zero API costs. Unlimited conversations with no per-token billing.
Whisper STT
OpenAI's Whisper Large V3 Turbo for speech recognition. Real-time transcription of your voice input, running entirely on-device.
Chatterbox TTS
Voice cloning that reproduces Eve's exact voice signature at 24kHz. Natural prosody, zero cloud dependency.
FlashHead Video
Lip-sync video generation from a single portrait image. Generates talking-head video matched to Eve's audio response.
Pipeline Comparison
Gemi Live 3.1
Gemma Eve
Brain
Gemini 3.1 Flash (Cloud)
Gemma 12B (Self-Hosted)
Voice
Native Gemini Voice
Chatterbox Clone
Latency
~100-200ms
~5-15s per turn
Cost
Per-token API billing
$0.00
Video
None
FlashHead Lip-Sync
Dependencies
Google Cloud, VoIP.ms
GPU Only (RTX 4090/5090)