Guides

Working notes on building voice agents, with every number pulled from the same catalog that prices your calls. Updated when the models or the prices move.

• How much does an AI voice agent cost per minute?

  What a phone or web voice agent costs in 2026, line by line: speech-to-text, language model, text-to-speech, telephony and platform fees at list price.

• Best speech-to-text for voice agents in 2026

  Streaming STT models compared on time to final transcript, built-in end-of-turn detection, word error rate, languages and price per minute.

• Best text-to-speech for voice agents in 2026

  Real-time TTS models ranked by arena quality, time to first byte and cost per spoken minute, with the trade-offs that matter on a phone call.

• Voice agent latency: the 700 ms budget

  Where the milliseconds go between a caller finishing a sentence and the agent's first sound: turn detection, STT, LLM first token, TTS and the network.

• Vapi vs Retell vs Bland: pricing compared

  Platform fees, telephony, concurrency charges and model lock-in on Vapi, Retell and Bland, with monthly totals at 1,000, 10,000 and 100,000 minutes.

• Speech-to-speech vs cascaded pipelines

  When a native speech model beats STT, LLM and TTS chained together, and when it does not: latency, cost, task completion and tool calling in 2026.

• Build a phone agent with the Voiceflint API

  A working phone agent in five API calls: create the agent, attach a number, place a call, read the receipt, receive the webhook.