ScribeBench
Get Featured
Independent measurement bench

We measure transcription accuracy.
Vendors don't get to.

Every Speech-to-Text API below is scored by running the same public-domain audio benchmark through it and computing the Word Error Rate ourselves. The numbers are first-party measurements — and sponsors cannot edit them.

Method: LibriSpeech test-clean (CC BY 4.0) · scored with jiwer WER · latency = median of 5 runs · v1

8.0%
2700ms
10 tools
Try

Get your API featured on ScribeBench

Featured placement puts your tool at the top of the index with your logo and pitch — in front of developers actively choosing a transcription API.

Claim or feature your listing →

Placement is for sale. Your measured WER is not — every featured tool shows the same unedited benchmark number.