Gladia

Gladia is a speech-to-text API built for developers shipping voice features into production — real-time meeting tools, contact center AI, voice assistants, transcription products. The technical pitch is they took Whisper, rebuilt it for streaming, and shipped Solaria-1, their proprietary model that covers 100+ languages with native code-switching (the speaker flips mid-sentence between English and Spanish, the transcript follows).

Real-time latency lands around 300ms for final transcripts, and async processing runs at roughly 60 seconds per hour of audio. Every plan — even the free 10-hours-per-month tier — includes diarization, translation, sentiment, NER, summarization, custom vocabulary, and code-switching.

That bundling is unusual; competitors often unbundle these and bill per feature. Pricing scales from $0.61/hour async on Starter down to $0.20/hour on Growth with annual commits, and live streaming runs slightly higher.

The API speaks WebSocket for streaming, REST for upload, and accepts the usual audio formats. The product is not for end users — there's no polished web app where a journalist drops in an MP3 and edits a transcript.

That's by design. Best for developer teams building voice products, AI meeting assistants, contact center vendors, anyone who needs Whisper-quality at production reliability with enterprise features baked in. Wrong fit for non-technical podcasters who want a click-and-edit transcription experience — Sonix, Happy Scribe, or Otter cover that.