We publish our processing stats, accuracy benchmarks, and technology choices publicly. These are numbers we measured — not marketing claims. Updated monthly.
These figures are VideoText’s own operations. For third-party industry statistics on captions, hearing loss, and speech recognition, see transcription statistics.
Last updated: March 14, 2026
Measured on our test set of 200 clips (total ~18 hours). Word Error Rate converted to accuracy percentage. Test conducted with Whisper large-v3, March 2026.
* These are real internal benchmarks, not vendor-reported figures. Whisper performance varies by language, audio quality, and content type.
Measured from upload-complete to transcript-complete. P50 = median; P90 = 90th percentile (worst typical case). Paid plans have queue priority during load.
We publish our stack so you know exactly what processes your content. No black boxes.
Most transcription tools don't publish accuracy benchmarks or processing speeds publicly. We do, because we think you should know what you're buying before you commit.
