Generate timed subtitles from audio or video up to 300 MiB and 20 minutes. Choose the spoken language, review each cue, then download SRT or VTT. The browser downloads Whisper and processes the recording on your device.
Choose a recording, then check its spoken language before generating subtitles.
Transcription runs in this browser
Try a short recording before your own file
Listen to a public English, Spanish or Arabic sample, load it into the tool, then compare your result with the dataset reference. These short read-speech clips do not represent noisy meetings or long recordings.
Se recomienda enfáticamente a los viajeros que se informen sobre cualquier riesgo de clima extremo en el área que visitan, dado que ello puede afectar sus planes de viaje.
The reference is a dataset transcript, not a promise of the tool's output.
Audio and reference: FLEURS contributors / Google, CC BY 4.0. Original recordings are unchanged. FLEURS · CC BY 4.0
From a recording to a subtitle file
Open the original recording. If it is a video with a supported audio track, you can generate captions directly. Listen to the result before downloading, especially names, numbers and lines that flash by too quickly.
Choose the recording
Select supported audio or video and choose its spoken language. MP3, WAV and M4A recordings, and video formats your browser can decode, use the same subtitle workflow.
Generate and review
Run transcription, then listen back while checking the subtitle text. Correct names, numbers and the start and end of each cue. Resolve invalid times before exporting.
Download the subtitles
Choose SRT for a separate subtitle file or VTT for a web caption track. Check the downloaded copy in the player or editor where you will use it.
An actual subtitle output to inspect
This SRT is the original Whisper output for the public English sample, generated with Tiny Q8 and WebAssembly on 3 October 2026. Use it to practice checking a cue against its recording and reference text.
The dataset reference says “the slow communication channels.” Listen to the recording and compare it with “this little communication channels” in the generated cue. Then check whether the cue starts with the speech and stays on screen long enough to read.
Limits and local processing
Use this page for supported media up to 300 MiB and 20 minutes. For a longer local recording, use large-file transcription, up to 1 GiB and 1 hour, then export its SRT for review. File extensions do not guarantee decoding; the audio codec must also work in your browser.
The browser downloads Whisper before it can generate subtitles. Processing speed depends on the device, model and recording. Selected media stays on this device. Subtitle edits are a working copy: download them before leaving; they do not replace the transcript saved in local history.
The site still loads application resources. Optional analytics follow the Privacy Policy.
Common questions
Can I convert MP3 or video to SRT?
Yes. Choose a supported audio or video file and generate a timed transcript. Review its words and timestamps before downloading SRT. You do not need to extract an MP3 from a video first.
What are the file and duration limits?
This page uses the standard transcription limit of 300 MiB and 20 minutes. The separate large-file tool accepts local files up to 1 GiB and 1 hour and also exports SRT and VTT.
Does my audio leave this device?
Selected media is decoded and transcribed in your browser. The browser still downloads the application and Whisper model. Optional analytics follow the site's Privacy Policy.
Will these subtitles be perfectly timed or accurate?
No. Speech recognition can miss words, and cue boundaries can need adjustment. Listen to the source, check names and numbers, and review captions in the final player. Timing checks do not verify what was spoken.
Can I use a YouTube page link or burn subtitles into a video?
No. Use a local recording or a direct media URL that allows browser access. The result is a separate SRT or VTT text file; it does not render captions into the video image.
Allow Google Analytics and Plausible to measure visits and tool usage? Optional analytics stays off until you agree. Your files and transcripts are not included in product events. The tools work either way. Privacy Policy