Transcribe speech to text without uploading files

How to transcribe audio or video

Use a recording with audible speech in a format your browser can decode. The tool reads the audio track of a video without changing the original file.

1. Click Choose an audio or video file, or drop it onto the dashed area. Wait until its duration and size appear.
2. Set Language to Auto-detect, Spanish or English. Keep Whisper Base for a quick draft, or choose Whisper Small for a more careful result.
3. Click Transcribe. If the selected model is missing, its download starts with a progress bar; Cancel stops it and leaves the form ready to retry.
4. Read the transcript in the scrollable result, then use Download TXT, Download SRT or Download VTT to save it.

Choosing Whisper Base or Small

The first model you request also needs a shared runtime of about 27 MB. Completed model files are cached in this browser, so they do not need to be fetched for every transcript.

• Whisper Base is the default and the lighter choice: roughly 80 MB plus the shared runtime when needed. It is faster, but its Spanish transcripts can be noticeably rough.
• Whisper Small is roughly 250 MB plus the shared runtime when needed. It needs more storage, memory and processing time, but is the better choice for Spanish and difficult recordings.
• Auto-detect works well when the opening speech is clear. Choose Spanish or English yourself if the recording starts with music, silence or another language.
• Only the model files come from the network. Your audio and the resulting transcript are processed and kept on your device.

What the TXT, SRT and VTT downloads contain

All three formats come from the same recognition result. TXT contains the readable words, while SRT and VTT also use each word’s start and end time to build subtitle cues.

• TXT is useful for notes, articles and editing the text in another application.
• SRT is a widely supported subtitle format with numbered cues and comma-separated timestamps.
• VTT uses the WebVTT format for compatible web players and browser-based video workflows.
• Subtitle lines are grouped at sentence endings, pauses and practical length limits. Review both wording and timing before publishing them.

Getting a cleaner transcript

Clear, close microphone recordings with little background sound usually work best. Very short clips, overlapping speakers, strong noise and specialist vocabulary can reduce accuracy.

• Choose the correct language instead of relying on detection when you already know it.
• Use Small for important Spanish text, and always proofread names, numbers and technical terms.
• Long recordings require enough free memory because the browser decodes and processes the audio on your device. Split a very large file if your device cannot handle it.
• If no speech is recognized, the tool shows a message instead of empty downloads. Try a clearer section or check that the file contains an audible voice.