WHISPERX GUIDE
How to use WhisperX online
The shortest route from a recording to a WhisperX result is to upload, run, and download. Use the editor when the recording needs a little more attention.
By the WhisperX team · Reviewed October 15, 2026
1. Try the sample before you upload
The public sample lets you play audio, move between segments, change text, and explore exports. Sample changes stay in your own session. They do not alter another visitor’s copy or start a paid inference task. Restore the sample whenever you want to return to its original result.
2. Choose a file and sign in
Use a recording you own or have permission to process. A real task belongs to your account, so sign in before upload and processing. If signing in returns you to an empty file picker, choose your file again; a browser cannot safely restore a local File object across every redirect.
3. Keep the options simple
Pick the language when you know it. Enable speaker labels for conversations and leave them off when you only need a solo transcript. Speaker labels describe voices in this recording, not verified identities. You do not need to provide a Python environment, a GPU, or your own model access token.
4. Check the duration and confirm
The charge uses the recording’s confirmed duration, including pauses. Review the quote before starting. Processing reserves the required balance; completed output settles that reservation. A failed processing task releases it. Purchasing more minutes is a separate payment action.
5. Listen at the places that matter
Check proper nouns, numbers, technical language, quiet passages, and interruptions. Click a segment to return to its position. Edit the words directly or rename speaker labels. The original model output remains a useful reference. A text edit does not create new measured word times.
6. Export the result for your next step
TXT is for reading; JSON keeps structured transcription data; SRT and VTT are subtitle formats. Export your latest saved copy after editing. Download before the retention period shown on your task ends, and check caption timing in the player or editor where you will use it.
See the minute packs and retention terms ↗