TTS Cost Comparison Shows Per-Minute Rates and Monthly Commitments

Five text-to-speech (TTS) models for voice agents list usage rates from $0.009 to $0.03 per minute, while three carry roughly $10 monthly spend commitments tied to each concurrent call.

Prices use a common conversion of 600 characters per minute to compare providers that bill by character. The figures also show a wide spread in time to first audio (TTFA), from 90 milliseconds to 395 milliseconds.

In a cascaded voice-agent system, speech recognition turns a caller’s words into text, a large language model processes them, and TTS generates the spoken response. The comparison measures the speech-generation stage using 600 characters as the assumed amount of speech per minute. At that rate, each per-million-character price is translated into a per-minute figure.

AssemblyAI-TTS-1 is listed at $15 per million characters, which works out to $0.009 per minute under the conversion. Its TTFA is 90 milliseconds, and it has unlimited concurrency with no spend commitment.

ElevenLabs v4 Turbo costs $40 per million characters, or $0.024 per minute. It lists a roughly $10 monthly spend commitment for each concurrent call and a 190-millisecond TTFA. Cartesia Sonic 3.6 is priced at $37.50 per million characters, equivalent to $0.0225 per minute, with the same approximate commitment and a 285-millisecond TTFA.

Rime Coda and Fish Drama 3 each list a price of $50 per million characters, or $0.03 per minute. Rime has unlimited concurrency without a spend commitment and a 395-millisecond TTFA. Fish carries a roughly $10 monthly spend commitment per concurrent call and has a 370-millisecond TTFA.

An unused commitment is nonrefundable. ElevenLabs, Cartesia and Fish carry commitments, while AssemblyAI and Rime list unlimited concurrency without them.

AssemblyAI has both the lowest per-minute rate and fastest TTFA among the five. Rime and Fish have the highest listed usage rate, though Fish’s TTFA is 25 milliseconds shorter than Rime’s. The comparison lays out three measures for voice-agent builders: usage price, time to first audio and monthly spend commitments linked to concurrency.