How this calculator works
This AI Voice Agent Cost Calculator estimates what a spoken AI agent costs to run, both per minute and per month. It stacks the three services a voice agent needs on every call: speech-to-text (STT) to transcribe the caller, the LLM to decide what to say, and text-to-speech (TTS) to voice the reply. You enter your per-minute price for each layer, then your calls per month and average minutes per call. The per-minute figure is simply STT plus LLM plus TTS, and the monthly total scales that by your total talk-time, so volume and call length are the largest drivers of what you pay.
Because cost scales with total minutes, small per-minute differences compound quickly at scale, so it pays to compare providers for each layer separately rather than picking a single bundled vendor. The LLM estimate is the least predictable input: longer conversations and more verbose responses raise real usage above a flat per-minute guess. Trimming average call length, or routing simple calls to a cheaper model, often lowers the bill more than shaving cents off STT or TTS.
Frequently asked questions
What does an AI voice agent cost per minute?
A voice agent stacks three costs per minute: speech-to-text (~$0.006/min), the LLM that generates replies (token-based, a few cents per minute of conversation), and text-to-speech (~$0.015/min or more for premium voices). Total is typically $0.05-0.20 per minute depending on the models.
How do I cut AI voice agent costs?
Use a cheaper LLM for routine turns, cap response length, pick a mid-tier TTS voice, and keep calls short. STT is usually the smallest line; the LLM and premium TTS dominate.