Deepgram Nova-3
Only active realtime speech-to-text path for dictation and live recording.
dictationrealtime
WaiComputer Dictation Arena
Synthetic fixtures track repeatable accuracy and latency for the locked speech stack: Deepgram Nova-3 for live text and ElevenLabs Scribe v2 for full transcripts.
File STT check
Record once, verify full transcription
RecordTranscribeReviewReady
0:00
One audio sample.ASingle active file STT model.
Speak naturally. Stop when the phrase is complete.
Language
Private round. Audio is used for this request only.
Active model
AElevenLabsstandby
Leaderboard
WaiComputer Active STT Stack
Current production speech-to-text provider contract
| Rank | Model | WER | CER | Latency | Speed |
|---|---|---|---|---|---|
| #1 | ElevenLabs Scribe v2elevenlabs / scribe_v2 | 21.9% | 10.0% | 1437 ms | 4.92x |
Model fit
Which model belongs to which task
ElevenLabs Scribe v2
Only active full-session file transcription model.
filefull transcript