Gladia is a speech-to-text and audio-intelligence API built for developers who want to add transcription, diarization, translation, and audio understanding to their own products, rather than a ready-made meeting-notetaker app for end users. It transcribes in 100+ languages through its own Solaria models, over a single API that handles both real-time and asynchronous audio. Pricing is usage-based per hour of audio rather than per seat, with 10 free hours every month on the entry Starter plan and volume discounts on the Growth plan.
Who builds it
Gladia builds its own proprietary speech-to-text models rather than reselling an off-the-shelf engine. Its lineup includes Solaria-1, aimed at the broadest possible language coverage (Gladia says 42 of its 100+ supported languages are unique to its platform), and the newer Solaria-3, tuned for noisy, real-world business and call-center audio, which Gladia claims is 26% more accurate than Solaria-1 on real English customer calls. The company markets itself as audio infrastructure for teams building voice agents, meeting tools, and call analytics, not as a consumer app.
Core features
- Single API for both real-time (streaming) and asynchronous (uploaded file) transcription
- 100+ supported languages (101 listed in its documentation), with automatic language detection and mid-sentence code-switching for nearly all of them
- Speaker diarization built on pyannoteAI, which Gladia markets as the top-ranked speaker-detection accuracy available
- An audio-intelligence layer for running summarization and extraction-style analysis directly on transcribed audio
- Two selectable STT models: Solaria-1 for broad multilingual coverage, Solaria-3 for higher accuracy on real business/customer-call audio in English, French, German, Spanish, and Italian
- Compliance coverage for GDPR, HIPAA, SOC 2 Type II, and ISO 27001, plus EU data residency
Pricing
- Starter — pay-as-you-go: $0.61/hour asynchronous, $0.75/hour real-time, with 10 hours included free every month and no credit card required to start; up to 25 concurrent async / 30 concurrent real-time requests
- Growth — commitment-based discounts down to about $0.20/hour async and $0.25/hour real-time (Gladia says up to 67% less than Starter), flexible concurrency, custom volume discounts
- Enterprise — custom annual pricing with unlimited concurrent requests, SLAs, a dedicated account manager, and custom/fine-tuned models
- Verified on gladia.io/pricing; check the current pricing page, since usage-based rates and discount thresholds can change.
Who it's for
The 10 free hours per month on Starter are enough for developers prototyping a transcription feature or evaluating accuracy before committing budget. Growth fits products with steady, higher-volume transcription that want lower per-hour rates and more concurrency. Enterprise suits companies that need custom or fine-tuned models, guaranteed SLAs, or fully unlimited concurrent throughput. Because Gladia ships as an API rather than a finished application, it is built for engineering teams integrating speech-to-text into their own product, not for people wanting a ready-to-use meeting assistant.