Tegain Voice Agent — AI Playground

📝 Agent Instructions

System prompt sent to the agent

🤖 Agent Type

All-in-one model — handles speech natively with no separate STT/TTS plugins

Mix and match providers — STT → LLM → TTS with full control

Model Gemini's native multimodal speech model

Voice Built-in Gemini voice character

Speech-to-Text STT Realtime streaming speech recognition

STT Languages DETECT

മലയാളംml

Englishen

हिन्दीhi

தமிழ்ta

తెలుగుte

ಕನ್ನಡkn

Auto

Select 1+ — ml+en default. Auto = 90+ detection (may hallucinate CJK). List sent as stt_language.

Language Model LLM Cerebras/, Sarvam/ → direct · Other / → OpenRouter · Plain → Google API

Text-to-Speech TTS Cartesia/ prefix → Cartesia · Others → ElevenLabs

⚡ Preemptive Reply Start LLM while user is still talking — saves ~0.5-1s per turn

🗣️ Backchannel Sounds Say "mm-hmm", "I see" while user speaks — fills the turn gap

Voice

Rachel 11Labs

Riya Cartesia

Latha Cartesia

Katie Cartesia

✏️ Custom

Voice ID

idle

Room —

Agent —

Duration 0.0s

Microphone off

Response Latency —

💬 Conversation Log

🎙️ Tegain Voice Agent