Soniox

Soniox is a voice AI platform that gives developers one API for speech-to-text, text-to-speech and speech translation in more than 60 languages. It is built for real-time use, such as voice agents, live captions and dictation, as well as batch transcription of audio files. Soniox also offers a consumer app for transcription, translation and voice typing that runs on its own API.

What It Is

Soniox builds its own speech models, inference engine and cloud infrastructure rather than adapting general-purpose models. Developers connect through REST and WebSocket APIs with SDKs, and manage API keys, usage, logs and billing in the Soniox Console. A single model covers every supported language, so apps do not load separate models per language.

Speech-to-Text

The speech-to-text API transcribes word by word as people speak, or processes uploaded files. It handles speakers who switch languages mid-sentence without manual configuration, separates speakers, identifies languages, and is designed to capture names, numbers and reference IDs exactly as spoken. Speaker diarization, language identification and smart formatting are included rather than sold as add-ons. STT v6, released in November 2026, is the latest model; Soniox describes it as the first speech recognition model that knows what a voice agent is saying and filters out the agent's own voice, and says it improves accuracy for voice typing, meetings and translation.

Text-to-Speech and Translation

The text-to-speech API streams expressive speech in 60+ languages with low latency, starting audio from the first few words. It supports audio tags for control over delivery, accurate reading of alphanumerics, language switching within a sentence, instant voice cloning from a few seconds of audio, and a library of more than 200 built-in voices. Speech translation is built into the speech-to-text API: the same call returns the transcript and the translation as the speaker talks, across 3,600 language pairs, in one-way mode for captions and broadcasts or two-way mode for conversations.

Soniox App

The Soniox App brings the same models to individuals on iOS, Android, macOS, Windows and the web. It includes Smart Scribe for transcription, a Translator, and Voice Typing, along with a personal workspace, shared projects, transcript sharing and offline recording.

Privacy and Deployment

Soniox says audio stays in memory during real-time processing and is never stored, and it lists SOC 2 Type 2, ISO/IEC 27001:2022, HIPAA and GDPR compliance. The same models and API can run in specific regions for latency and data residency, including a regional deployment in India in limited early access. Soniox names customers including Perplexity, Samsung, LG, Agora, Krisp, LiveKit, Pipecat, Fireflies.ai, Retell AI and Wispr Flow, and runs a public comparison tool for testing its speech-to-text, text-to-speech and translation against other providers on your own audio.