Gladia
Gladia is an audio API to transcribe and enrich speech in 100+ languages, with real-time and async Solaria models; part of OVHcloud since 2026
Overview
Gladia describes itself as "the end-to-end audio infrastructure to record, transcribe and enrich audio through a single API". It offers async and real-time speech-to-text in 100+ languages with code-switching, speaker diarization, word-level timestamps, translation in the same call, and audio intelligence such as summaries, sentiment, named entities and PII redaction. Its Solaria-1 model handles real-time and async with broad language coverage, while Solaria-3 (June 2026) is a cheaper pre-recorded model tuned for English, French, German, Spanish and Italian. The Paris company, founded in 2022, raised a $16M Series A led by XAnge in October 2024, and the European cloud provider OVHcloud completed its acquisition of Gladia on 31 July 2026. Hosting is offered in the EU and US.
The Verdict
Who Should Use Gladia?
Best For
- Meeting, call and voice-agent products that need multilingual real-time transcription
- European teams that want an EU-owned vendor with EU hosting
- Startups that want a generous free credit and simple per-hour pricing
Not Ideal For
- Teams needing on-prem deployment (not offered in the sources we found)
- Very high real-time volumes on a tight budget: there is no listed real-time volume discount
- Less widely spoken languages where reviewers saw weaker accuracy
What's Great
- Accuracy and multilingual real-time transcription are the top G2 praise points
- Responsive support, including direct access to the team
- Easy API setup and clear documentation
- €50 free credits with no card, roughly 312 async hours on Solaria-3
- Diarization, translation and audio intelligence in one API call
Watch Out For
- Cost at high volume is the most common G2 complaint
- Accuracy drops with noisy audio, heavy accents and less-spoken languages; ITdaily's Dutch test was slow and error-heavy
- ITdaily found no MP4 support and slow transcription on some files
- Solaria-3 is pre-recorded only and drops code-switching
- The OVHcloud acquisition may change roadmap and terms
Pricing
View all features & details
Key Features
- Async and real-time speech-to-text, under 300ms real-time latency (vendor claim)
- Solaria-1 (100+ languages, code-switching) and Solaria-3 (five European languages, async)
- Speaker diarization with word-level timestamps
- Translation in the same API call
- Summaries, sentiment, named entities and PII redaction
- Audio-to-LLM pipeline with built-in or your own model
Use Cases
- Meeting assistants and notetakers
- Speech input for voice agents
- Call analytics with PII redaction
Gladia's own benchmark puts Solaria-3 at 6.4% WER on Earnings22, ahead of AssemblyAI and Speechmatics; that is a vendor-run test. Sources: Solaria-3 launch · OVHcloud acquisition · Sifted (Series A) · Docs
How It Compares
| Feature | Gladia | AssemblyAI | Speechmatics |
|---|---|---|---|
| Entry async price | €0.16/hr (Solaria-3) | $0.15/hr | $0.24/hr |
| Languages | 100+ (Solaria-1) | See vendor | 55+ |
| On-prem | Not offered in sources found | See vendor | Yes |
| Free start | €50 credits | See vendor | $100 credits |
Sources: Gladia pricing · AssemblyAI page · Speechmatics page
Is this your tool? Add a badge