What does a voice agent really cost?
A production voice agent is three billed layers stacked together: the platform that runs the call, the speech-to-text that hears the caller, and the text-to-speech that answers. Pick one of each to see the combined cost per minute, built from verified published rates, not estimates.
This combines three verified layers: the platform’s orchestration and telephony per minute, streaming speech to text, and text to speech synthesis (converted from its per character rate at about 950 characters per minute of speech). Language model tokens are not included here: they vary by model and conversation, and most platforms pass them through at cost. Every rate links to its primary source on the tool’s page.
How this number is built
Each layer’s rate is normalized to dollars per audio minute by the same deterministic cost engine that powers every comparison on the site: speech-to-text stays in dollars per minute (streaming where a vendor publishes it), and text-to-speech converts from its per character price at about 950 characters per minute of speech. The three add up to a conversational minute. Language model tokens vary by model and conversation and are excluded; most platforms pass them through at cost. See the methodology for how every rate is sourced and verified.