Speech-to-text: what it is and how an AI voice agent uses it
Speech-to-text is the technology that converts a person's audio into text while they talk. It is the first step of an AI voice agent: it listens to the call, understands the request and answers in natural speech.
What speech-to-text is
A speech-to-text engine receives the call audio and transcribes it as the caller speaks. The text goes to the language model, which decides the reply based on the conversation so far.
How an AI voice agent uses it
1. It listens to the call and transcribes it in real time.
2. It works out the intent: a booking, a pricing question, a complaint or a request to speak with a person.
3. It finds the answer in your information through a knowledge base (RAG) and replies by voice.
4. It follows the flow you designed and saves the call data.
InfiVoice agents support 30+ languages, so they transcribe and reply in the language the caller uses.
Speech-to-text vs. a traditional IVR
- Traditional IVR: recognizes only keypresses or a few predefined keywords.
- AI speech-to-text: understands full sentences, even when the caller does not use the exact words.
To see how it fits into the whole call, read how AI voice agents work and how to prepare a knowledge base for an AI voice agent.
How to get started with InfiVoice
Create an agent, load your business information, design the flow and assign a local number. Plans are at pricing.
Price
The Free plan includes 100 credits. The Pro plan costs $79 per month with 250 credits. One credit equals 60 seconds of call time.
Frequently asked questions
Is speech-to-text the same as an IVR?
No. An IVR only recognizes fixed options. Speech-to-text converts any sentence into text so the AI can interpret it.
Does it work in multiple languages?
InfiVoice supports 30+ languages, so the agent transcribes and replies in the caller's language.
How much does it cost to try?
The Free plan includes 100 credits, which equal 100 minutes of call time.