Gladia is a speech-to-text API, a service that another program calls over the internet to turn spoken audio into written text. Gladia SAS makes it.
A program sends a recorded file through a REST upload (asynchronous, or batch, transcription) or streams live audio over a WebSocket (real-time transcription), and gets a transcript back. Common formats such as MP3, WAV, FLAC and Opus are accepted, in more than 100 languages with automatic language detection. Optional audio intelligence features enrich the transcript: speaker diarization (labelling which speaker said each sentence), translation, subtitles, named entity recognition, removal of personal data (PII redaction), sentiment analysis and summaries. SDKs exist for Python and Node.js, and the API connects to voice-agent tools such as Pipecat, LiveKit and Twilio.
Gladia is a hosted service billed per hour of audio with prepaid credits and no subscription; data can stay in the EU. Enterprise customers can buy an annual plan with custom models.