What is Gladia?
Gladia is a speech-to-text API for real-time and recorded audio, built for developers making voice agents and meeting tools. The Paris-based company says more than 300,000 developers and 2,000 enterprise teams use it, including HeyGen, VEED and Recall.ai, and that it has transcribed more than 26.5 million hours of audio.
Gladia is tuned for multilingual and real-time use: it supports 100+ languages with code-switching mid-sentence, returns live transcripts in under 300 milliseconds, and adds diarization, translation, sentiment and entity extraction in the same call. It has native integrations with voice agent frameworks such as LiveKit and Pipecat and telephony like Twilio, and offers EU data residency and a no-training guarantee. Its Solaria models cover a universal multilingual option and a model tuned for English and European languages.
Main Features
- Transcribes live audio with sub-300 millisecond latency.
- Transcribes recorded files in batch.
- Detects and switches between 100+ languages mid-sentence.
- Labels speakers with diarization and adds word-level timestamps.
- Translates, summarizes and extracts entities from transcripts.
- Integrates with LiveKit, Pipecat, Twilio and Retell.
- Offers EU data residency and SOC 2, HIPAA and GDPR compliance.
- Offers Solaria models tuned for multilingual or English and European audio.
How to Get Started with Gladia
- Create an account at gladia.io to get the free credit.
- Copy your API key from the dashboard.
- Send an audio file or open a real-time WebSocket session.
- Turn on diarization or translation in the request options.
Gladia Pricing
New accounts get a one-time 50 EUR credit, roughly 80 hours of recorded or 60 hours of real-time transcription. The pay-as-you-go Starter plan costs $0.61 per hour for async and $0.75 per hour for real-time. Growth goes as low as $0.20 per hour async and $0.25 real-time with volume, and Enterprise is custom.
Who Should Use It?
- Developers building AI voice agents.
- Meeting note and recording app makers.
- Companies serving multilingual customers.
- European teams that need EU data residency.
- Contact centers analyzing calls.
Gladia Alternatives
- Deepgram: popular speech API with strong English accuracy
- Speechmatics: accurate multilingual transcription for enterprise
- Recall: API for meeting bots that capture calls
Frequently Asked Questions
Is Gladia free?
Yes, to start. New accounts get a one-time 50 EUR credit, then pay as you go from $0.61 per hour.
How many languages does Gladia support?
Gladia supports 100+ languages and can switch languages within a sentence.
Gladia vs Deepgram: which is better?
Gladia is strong for multilingual audio and EU data residency. Deepgram is a common choice for English-heavy, high-volume workloads.
Does Gladia train on my audio?
No. Gladia contractually guarantees it does not train on customer audio.