What is this site?
WhisperX (whisperx.app) is an independently operated, unofficial hosted service that runs the open-source WhisperX pipeline in the browser—no local Python or GPU upkeep. Upload English audio or video, get a transcript with word-level timestamps and optional speaker labels, then review it in a browser editor and export TXT, SRT, VTT, or JSON. The site states it is not affiliated with the upstream authors or OpenAI.
Core features
- Hosted WhisperX: transcription, word alignment, and optional speaker diarization
- Browser editor: playback linked to the transcript; jump to words, edit text, rename speakers
- Keeps model output separate from human edits; editing a sentence does not create fresh alignment
- Export TXT, JSON, SRT, VTT for original or edited copies; re-export does not consume more minutes
- Public samples for preview; processing your own files requires an account
- Charged by media duration: one-time 5-minute trial for verified accounts; subscriptions and minute packs available
Common use cases
- Interviews or podcasts that need reviewable English transcripts with speaker turns
- Meeting recordings turned into readable notes before export
- SRT/VTT for video, then checked in an editor at the start, middle, and end cues
- Using the WhisperX pipeline without standing up a GPU environment
- Trying public samples for timing and speaker labels before uploading private files
Who is it for?
- People processing English recordings who need word timestamps or speaker labels
- Individuals or small teams who want install-free review and export in the browser
- Users who know open-source WhisperX but do not want to maintain a local stack
- Poor fit: non-English-first work, fetching YouTube/external links, or burn-in/translate/summarize all-in-one workflows
- Also poor fit: compliance settings that treat machine transcripts as final without human review
How does it compare to similar tools?
Choose by output needs:
- Real WhisperX pipeline with word timing/speakers and no self-hosting: the hosted site is simpler
- Full control of models and data: install the open-source project locally
- Generic speech-to-text without alignment/diarization emphasis: other ASR products may fit
- Machine transcripts and speaker labels can be wrong; the site’s own samples show what to check
FAQ
Q: Is this the official WhisperX website?
A: No. It is an independent unofficial hosted service; the open-source project is m-bain/whisperX on GitHub.
Q: How does billing work? Does editing cost more minutes?
A: Charges follow measured media duration. Starter is about $9/month for 120 minutes; Creator about $29/month for 600 minutes; one-time minute packs are also sold. Editing and re-exporting an existing result do not consume more minutes.
Q: Which languages and upload methods are supported?
A: Hosted transcription currently targets English recordings. Upload your own files; the service does not fetch external media links.
Q: How long do the trial and results last?
A: Verified accounts get a one-time 5-minute trial. Trial results last about 7 days; paid results about 30 days (see the site for current terms).
