Upload any audio (podcast, meeting, lecture), STT transcribes in source language, AI Inference translates, TTS generates audio in target language. Returns translated audio + aligned transcript.

Telnyx API Endpoints Used

• STT Transcribe: POST /v2/ai/transcribe -- ref
• AI Inference: POST /v2/ai/chat/completions -- ref
• TTS Generate: POST /v2/ai/generate -- ref

Architecture

  API Request
        │
        ▼
  ┌──────────────────┐
  │ Answer + Greet    │ ── TTS welcome message
  └────────┬─────────┘
           │
           ▼
  ┌──────────────────┐
  │ Gather Speech     │ ── STT transcription
  └────────┬─────────┘
           │
           ▼
  ┌──────────────────┐
  │ AI Inference      │
  │ • Translation      │
  └────────┬─────────┘
           │ ◄──── conversation loop
           │
           ▼
     JSON response

How It Works

1. Sends conversation to Telnyx AI Inference for processing
2. Converts response to speech via Telnyx TTS

Why Telnyx

Telnyx is an AI Communications Infrastructure platform - voice, messaging, SIP, AI, and IoT on one private, global network.

• Co-located inference - LLM runs on the same network as voice traffic. Sub-200ms round trips.

Environment Variables

Copy .env.example to .env and fill in:

Setup

git clone https://github.com/team-telnyx/telnyx-code-examples.git
cd telnyx-code-examples/ai-content-translator-python
cp .env.example .env
pip install -r requirements.txt
python app.py

Webhook Configuration

Set webhook URL in Telnyx Portal:

• Call Control Application -> https://<id>.ngrok.io/webhooks/voice

API Reference

POST /translate

Upload as multipart form:

curl -X POST http://localhost:5000/translate \
  -F audio=@lecture.mp3 \
  -F source=en \
  -F target=ja

Response:

{"job_id": "tr-a1b2c3d4", "status": "complete", "source": "en (English)", "target": "ja (Japanese)", "original_length": 1847, "translated_length": 923}

GET /health

curl http://localhost:5000/health

Troubleshooting

• Connection refused on port 5000: App isn't running. Run python app.py and check no other process uses port 5000.
• 401 Unauthorized: Your TELNYX_API_KEY is invalid. Generate a new one at portal.telnyx.com/api-keys.
• AI response slow/empty: Verify model name. See available models at developers.telnyx.com.

Related Examples

• run-llm-inference-python - Standalone inference
• build-voice-ai-agent-python - Voice AI agent

Resources

• AI Inference Guide
• Call Control Guide
• Telnyx Developer Docs
• Telnyx Portal