Audio
Speech-to-text
Transcription, diarization and real-time captions.
What it is
Speech-to-text (STT/ASR) transcribes audio into text, optionally identifying speakers, timestamps, languages and punctuation, in batch or in real time.
How pricing works
Billed per minute or hour of audio processed. Real-time streaming and extras such as diarization may cost more.
We compare everything in one unit: USD per hour of audio.
Typical uses
- Meeting and call transcription
- Live captions
- Voice agents
- Searchable media archives
What to compare
- Word error rate (WER)
- Real-time latency
- Speaker diarization
- Languages
Watch out for
- Accuracy drops a lot on noisy audio, accents and jargon; test with your own recordings.
All 8 APIs, cheapest first
Free to browseTick 2 to 4 APIs to compare them side by side.
| API | $ / audio hour | Real-time streaming | Speaker labels | Languages | Quality | Price data | Compare |
|---|---|---|---|---|---|---|---|
| Soniox (async) Soniox | $0.10 | Yes | Yes | 60 | – | Verified Oct 6, 2026 | |
| Whisper large-v3 (Groq) Groq | $0.11 | No | No | 99 | 82 | Unverified sample | |
| Universal-2 AssemblyAI | $0.15 | Yes | Yes | 99 | 89 | Verified Oct 4, 2026 | |
| Rev AI Reverb Rev AI | $0.20 | Yes | Yes | — | – | Verified Oct 6, 2026 | |
| ElevenLabs Scribe v2 ElevenLabs | $0.22 | Yes | Yes | — | – | Verified Oct 6, 2026 · aging | |
| Speechmatics Melia 1 Speechmatics | $0.24 | Yes | Yes | — | – | Verified Oct 6, 2026 | |
| Nova Deepgram | $0.26 | Yes | Yes | 36 | 88 | Unverified sample | |
| Gladia (Starter, async) Gladia | $0.61 | Yes | Yes | 100 | – | Verified Oct 6, 2026 |