Soniox is a speech AI platform that provides speech-to-text, text-to-speech, and speech translation through one unified API. All three services cover 60+ languages, and the platform is designed for real-time voice applications where transcription, synthesis, and translation happen while a person is still speaking.
Speech-to-text runs in real time over a streaming connection and handles multilingual and language-switching conversations, multiple speakers, alphanumerics such as phone numbers and identifiers, and domain-specific vocabulary with a single model across languages. Text-to-speech (TTS v2) generates speech in 60+ languages with instant voice cloning from short reference clips, low-latency streaming that starts before a sentence is complete, and expressiveness controlled through audio tags. Speech translation covers 3,600 language pairs and emits translated output before the source sentence finishes.
The platform lists in-region processing for data residency and SOC 2 Type 2, ISO/IEC 27001:2022, HIPAA, and GDPR compliance.
Features
- Real-time speech-to-text: streaming transcription across 60+ languages with sub-200 ms latency
- Multilingual and code-switching: one model handles language switching within a conversation
- Speaker identification: multi-speaker conversations are separated by speaker
- Alphanumeric accuracy: phone numbers, IDs, names, and domain vocabulary
- Text-to-speech v2: 60+ languages, voice cloning from short clips, cross-language voices, character-level timestamps, and audio tags for expressiveness
- Streaming synthesis: audio starts before the sentence is complete
- Real-time translation: 3,600 language pairs with low-latency output ahead of sentence boundaries
- Unified API: speech-to-text, text-to-speech, and translation behind one API, with a companion app
- Compliance and residency: in-region processing, SOC 2 Type 2, ISO/IEC 27001:2022, HIPAA, and GDPR
