Skip to main content
STT converts spoken audio into text that your AI Voice Agent can understand.

​
What is STT?

Speech-to-Text (STT) is the process of converting spoken words into written text. In Kuralit, STT:
  • Streams audio continuously - Process audio in real-time
  • Provides interim results - Show transcription as user speaks
  • Delivers final transcripts - Complete utterance transcriptions
  • Supports multiple languages - Various language codes

​
How STT Works

STT plugins process audio through this flow: 
Audio Input (WebSocket)
    ↓
STT Plugin
    ↓
Interim Transcripts (as user speaks)
    ↓
Final Transcripts (complete utterance)
    ↓
AI Agent (LLM)

​
Configuration

​
Basic Configuration

from kuralit.server.agent_session import AgentSession

# Using Deepgram
agent = AgentSession(
    stt="deepgram/nova-2:en-US",
    # ...
)

# Using Google Cloud STT
agent = AgentSession(
    stt="google/en-US",
    # ...
)

​
Environment Variables

# Deepgram
DEEPGRAM_API_KEY=your-deepgram-api-key

# Google Cloud STT
GOOGLE_STT_API_KEY=your-google-stt-key
# OR
GOOGLE_STT_CREDENTIALS=/path/to/credentials.json

​
Available Providers

  • Deepgram - High accuracy, real-time streaming, multiple languages
  • Google Cloud STT - Google ecosystem integration, high accuracy
View all STT providers →

​
Language Codes

Common language codes:
  • en-US - English (United States)
  • en-GB - English (United Kingdom)
  • es-ES - Spanish (Spain)
  • fr-FR - French (France)
  • de-DE - German (Germany)

​
Sample Rates

STT plugins support various sample rates:
  • 8000 Hz - Telephone quality
  • 16000 Hz - Standard quality (recommended)
  • 44100 Hz - High quality

​
Next Steps