Realtime voice model
A realtime voice model is a single AI model that listens to speech and replies with speech directly, instead of chaining separate speech-to-text, language model and text-to-speech steps.
A traditional voice agent pipeline transcribes the caller, sends the text to a language model and speaks the answer with a separate voice model. A realtime (speech-to-speech) model takes audio in and produces audio out, which removes hand-offs between models and can hear tone and emphasis that a transcript loses.
The trade-off is control: a pipeline lets you pick the best transcription, reasoning and voice provider separately and inspect the text at each step. Many platforms, VocalLabs included, let an agent run either way.
