Observed arrival · 2026-10-10
TranscribeBench puts transcription claims on the record
A set of browser-based subtitle and transcription calculators sits alongside a benchmark of speech-to-text accuracy on openly licensed recordings.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People evaluating transcription and subtitle workflows
- Worth noticing
- Its 33.6-minute baseline spans clean speech, restaurant noise, multilingual video calls, and meetings; commercial vendor tests have not yet run.
Field notes
The current benchmark runs Whisper medium.en against openly licensed recordings with human transcripts, then scores errors with published code. Its 33.6 minutes cover clean read speech, restaurant noise, relaxed video calls between speakers of different first languages, and four-person meetings. The site says no commercial vendors could be tested in the first round, so the published results are a baseline rather than a vendor ranking.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
$PaidCommerce or pricing visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue