Gladia

Gladia is a speech-to-text API that gives developers real-time and batch transcription plus audio intelligence features through a single integration. It is built for products that listen, such as voice agents, meeting note takers and contact center tools. Its two selling points are multilingual coverage through its Solaria models and per-hour pricing priced in euros that undercuts many better-known APIs.

What is Gladia used for?

Gladia sits behind other software. A typical team sends audio to it, receives a transcript with timestamps and speaker turns, and then feeds that text to a language model or a database.

• Streaming transcription for phone and web voice agents built on Pipecat, LiveKit or Twilio
• Meeting recording and notes through its native bot for Zoom, Google Meet and Microsoft Teams
• Batch transcription of uploaded audio and video with webhook delivery
• Summaries, sentiment and named entity extraction on the same API call
• PII redaction for call recordings before they reach storage or analytics
• No-code pipelines through Zapier, Make and n8n connectors

Who Gladia is best for

Gladia fits startups and product teams building voice features who want to start on self-serve credits and scale without a sales call. European companies that prefer an EU-headquartered vendor will find that useful too. Teams that need a complete transcription app for staff, rather than an API, should look at finished tools instead.

Gladia pricing

As of October 2026, Gladia's self-serve plan is pay as you go. Asynchronous transcription starts at 0.16 euros per hour on Solaria-3 and 0.27 euros per hour on Solaria-1, with rates falling at higher monthly volumes. Real-time transcription starts at 0.68 euros per hour. New accounts get 50 euros in credits valid for one year, and self-serve concurrency is capped at 25 async and 30 real-time requests. Add-ons are billed per hour, for example summarization at 0.03 euros and PII redaction at 0.07 euros. Enterprise plans are annual and custom priced, with unlimited concurrency, fine-tuned models, zero data retention, SLAs and a dedicated account manager.

Limitations to keep in mind

• On self-serve plans, model training on customer audio is enabled by default and must be switched off
• Language coverage differs by model, so check that Solaria-3 supports your languages before choosing it
• Diarization and enhanced punctuation are marked async only in the plan comparison
• Prices are in euros, which adds currency risk for US budgets

Gladia vs alternatives

Deepgram is the better-known choice for voice agents and also offers text-to-speech from the same account. AssemblyAI has a broader set of speech understanding features and extensive documentation. OpenAI Whisper is free if you are willing to host it, though you give up streaming convenience and managed scaling. Pulse: Speech-to-Text by Smallest AI is another low-latency option aimed at agent builders. Choose Gladia when you want low per-hour costs, an EU vendor and summaries or redaction bundled into the same request.

Tips for getting started with Gladia

Turn off model training in your account settings before sending real customer audio. Use the free credit to compare Solaria-1 and Solaria-3 on your own languages and accents. If you are building a voice agent, start from the Pipecat or LiveKit integration instead of writing the WebSocket client yourself, and add custom vocabulary for product names early.

Gladia FAQ

Is Gladia free?

Gladia gives new accounts 50 euros in free credits valid for one year, which covers roughly 300 hours of async transcription. After that, usage is billed per hour.

Does Gladia support real-time transcription?

Yes. Gladia offers streaming transcription over WebSockets, priced from 0.68 euros per hour of audio on the self-serve plan.

Is Gladia safe to use with company data?

Gladia lists GDPR, HIPAA and SOC 2 Type II compliance. Self-serve accounts should opt out of model training, and enterprise plans offer zero data retention.

What is the best alternative to Gladia?

Deepgram and AssemblyAI are the most common alternatives, while OpenAI Whisper suits teams that want to self-host a free model.