📞 Voice AI Cost Calculator

System Architecture Overview

Simple Modular Chain:

• Telephony Provider → makes call
• Speech-to-Text (STT) → Converts audio to text
• Simple Server → Prompt management + basic RAG + orchestration
• Large Language Model → Conversation processing
• Text-to-Speech (TTS) → Converts response to audio
• Telephony Provider → Delivers audio response

📊 Research-Based Assumptions (July 2025)

Token Usage: 420 input + 280 output tokens per minute (including context & memory)

Voice Processing: STT charges for customer speaking time (40% of call), TTS charges for AI speaking time (60% of call)

Volume Discounts: LLM: up to 50% (Batch API), Voice: up to 20%, Telephony: up to 30% (Twilio enterprise rates)

Telephony: Twilio Germany rates - Local $0.0174/min, Mobile EEA $0.042/min, Mobile Std $0.35/min

Infrastructure: $200 base + $0.002 per call (hosting, monitoring, compliance)

Pricing Sources: OpenAI API, Anthropic, Deepgram, ElevenLabs, Twilio official pricing pages