Providers

12 providers, 36 models, one schema. Prices below are vendor list prices — we add nothing.

• Soniox — 60+ languages with translation-grade accuracy; realtime bills wall-clock session time.
• AssemblyAI — Streaming turn model built for voice agents, plus strong async models.
• AWS — Amazon Transcribe streaming with per-request minimums.
• Azure — Azure AI Speech — enterprise realtime + fast transcription.
• Cartesia — Ink — low-latency STT built for realtime agents.
• Deepgram — The realtime workhorse — fast, cheap, excellent English.
• ElevenLabs — Scribe — high-accuracy STT from the voice company.
• Google — Google Cloud Speech-to-Text (Chirp).
• Groq — Whisper on LPUs — the fastest batch Whisper anywhere.
• Mistral — Voxtral — open-weights transcription via Mistral’s API.
• OpenAI — Whisper and the GPT-4o transcribe family — LLM-grade accuracy.
• Speechmatics — Broad language coverage; Melia-1 code-switches across 56 languages.

The live, machine-readable version: GET /v1/models.