Speak into your microphone and watch several speech-to-text providers answer at the same time. You see each provider's own stream beside our reading of it, including the revisions — a transcript that only shows the final version hides the thing that actually differs between them.
Charged by the second, per provider. A session reserves what it may cost when it starts and refunds the rest when it ends, so you pay for the seconds you were actually connected. When the reserved amount runs out we stop the session for you — exactly as if you had pressed Stop, so nothing you said is lost.
Captured at 16kHz mono, sent every 250ms. Providers that want another rate are resampled on the server, so every one of them hears the same microphone.
8 models selected·
assemblyai—
awsus-east-1
azureeastus
deepgram—
elevenlabsus
googleus-central1
mistral-aiglobal
openaiglobal
speechmaticsus1
together-aiglobal
2 models do not serve auto-detect — they are held back from this session. aws transcribe, azure short-default
Missing a specific model? Ask us to add it — we usually add models quickly.
Each provider is billed separately, so choosing more of them buys fewer seconds with the same balance.