Soniox is a speech AI platform that offers real-time speech-to-text, text-to-speech and speech translation through one API. It supports more than 60 languages, handles speakers who switch languages mid-sentence, and delivers transcription with sub-200ms latency. Its text-to-speech produces expressive audio with audio tags and instant voice cloning, and translation covers 3,600 language pairs. Billing is pay-as-you-go by tokens, which works out to about $0.10 per hour for async transcription, $0.12 per hour for real-time transcription and $0.70 per hour of generated speech.
Developers building voice agents, dictation tools and live captioning are the main users, along with teams in call centers, meetings and healthcare. Product managers on wearables and translation apps choose it for low-latency multilingual output. Enterprises with data-residency needs can use regional deployment, including a launch in India. Companies that do not want to stitch together separate vendors for recognition, synthesis and translation benefit from a single platform, and a consumer app is also available for hands-on use.
Typical applications include voice agents that serve non-English markets, meeting note-takers that show captions as people speak, and contact center software that transcribes and translates calls in real time. Automotive voicebots rely on accurate recognition of customer IDs and alphanumerics, and dictation products turn speech into text in noisy environments. Perplexity integrated Soniox to power its voice experience, and customers named in its customer section use it for meetings, government projects and phone-call transcription. The company also publishes side-by-side comparisons with OpenAI and Deepgram and a pricing calculator.
