← Back to home

Transcription benchmark

This page previously displayed processing-speed figures that were not backed by a published dataset or methodology. We removed them rather than let unverified numbers stand next to our evidence-based research. See below for what we can currently substantiate.

Evidence-based accuracy research

We publish a reproducible transcription accuracy study — dataset, ground truth, scoring code, raw results, and a downloadable CSV — at our transcription accuracy benchmark. It's an early-stage pilot (one condition, one open-source baseline model so far), and it says so plainly. A processing-speed study built to the same standard is planned; until it exists, we won't publish speed numbers here that we can't trace back to a dataset and script.

Direct answer

How fast is VideoText, really?

We don't have a published, sourced speed benchmark yet — see above for what we do have.

VideoText is built for batch and async processing of long-form recordings (meetings, podcasts, webinars). We'd rather say that plainly than publish specific minute-by-minute figures we can't trace to a dataset and script.

  • Designed for recorded, file-based workflows rather than live/real-time capture
  • Built for batch processing of long-form content (podcasts, webinars, meetings)
  • A sourced, reproducible speed benchmark is on the roadmap — see the accuracy benchmark for the standard it will follow

What VideoText is built for

  • Processing recorded long-form video and audio (meetings, podcasts, webinars).
  • Batch and async workflows rather than live transcription.
  • Structured export output (transcript, subtitles, summary) rather than raw text alone.

When VideoText may not be the best choice

  • Real-time meeting transcription with bot attendance → tools built for live capture are a better fit.
  • Collaborative timeline editing inside a video editor → an editor-integrated tool is a better fit.

Related pages

Best transcription toolAccuracy benchmark (evidence-based)Video to transcriptAI transcription tools

FAQ

What is the fastest transcription software?
We don’t have a published, sourced processing-speed benchmark yet. See our evidence-based research page for methodology and current results as they become available.
How long does it take to transcribe a 2-hour video?
This depends on file length, audio quality, and server load, and we don’t have a verified, dataset-backed number to publish yet. Try it directly on your own file to see actual processing time.
Are benchmark numbers measured from upload start?
Any speed numbers we publish going forward will state exactly what they measure (e.g. upload-complete to transcript-complete) and link to the dataset and script used to produce them.
How can I see real benchmark data?
Our evidence-based accuracy benchmark publishes its dataset, methodology, raw results, and a downloadable CSV. A speed benchmark built the same way is on the roadmap.