Audio to Text with Timestamps (SRT/VTT Subtitles)

Audio to Text with Timestamps (SRT/VTT Subtitles)

A plain transcript is useful for notes. Subtitles need times. FlashKit’s private transcriber keeps the chunk times the Whisper pipeline returns and writes them as SRT or WebVTT. That is the “audio to text free with timestamp” path: one run, three downloads (TXT, SRT, VTT).

How the cues are built

The worker asks Whisper for return_timestamps: true (or 'word' if you tick that box). Each chunk has a start and end in seconds. The page formats those seconds as HH:MM:SS,mmm for SRT and HH:MM:SS.mmm for VTT. If a fractional part would round to 1000 ms, the time carries into the next second — including 59.9996 s → 00:01:00,000 and the hour boundary.

This post is the subtitle how-to. The no-sign-up video walkthrough is transcribe video to text free. Privacy limits stay in privacy-first interview transcription.

Export SRT or VTT

Phrase-level cues by default. Word-level is a checkbox. The file is not uploaded.

Open the transcriber

Using the file as subtitles

  1. Download the .SRT or .VTT next to the media file.
  2. In VLC, Subtitle → Add Subtitle File. In many editors, import SRT onto the timeline.
  3. For a web player, point a <track> element at the .vtt.
  4. If you need another language, that is a new decision: the subtitle translator sends text to a translation API after consent. Do not send an NDA transcript.

When the times look wrong

  • A silent intro still occupies time. Cut dead air with the audio cutter before you transcribe if you want cues to start at 00:00:00.
  • Word-level mode writes one cue per word. Players will flicker. Use it for alignment, not for a cinema watch.
  • The model does not burn captions onto pixels. It only writes a sidecar file.

Languages and accuracy tips are in Indic audio to text and better transcription accuracy.

Frequently asked questions

What is the difference between SRT and VTT?

Both are timed cue lists. SRT uses a comma before milliseconds (00:00:04,120). WebVTT starts with WEBVTT and uses a decimal point (00:00:04.120). FlashKit writes both from the same chunks.

Can I get a timestamp on every word?

Yes. Tick word-level timestamps before you run. The file becomes denser. Use phrase-level cues for a reading transcript.

Why would a time show 1000 milliseconds?

Older builds could print 00:00:04,1000 when a stamp rounded to .9995 s. The formatter now carries that into the next second (00:00:05,000), including minute and hour roll-over.

Related Tools