How to Get a Transcription With Timestamps or Timecode

To get a transcript with timestamps, upload your recording to WhisperAI, turn on Timestamps and export from the recording's page. Pick the format for your next step: TXT or DOCX to read and quote, SRT or VTT for captions and video editors, JSON for your own code. To match video timecode, add the timeline's start time to every timestamp.

How do you transcribe audio or video with timestamps?

1. Upload the file. Drop audio or video on WhisperAI's audio to text page or video to text. Files up to 5 GB work on every plan, free included. When it's done, you'll see the transcript with speaker labels.
2. Turn on Timestamps above the transcript. Then pick a view. Paragraph puts one time at each change of speaker. Segments puts one in front of every sentence. Click any time and the audio jumps there, so you can check a line by ear.
3. Export from the recording's page. The export follows the view on screen, so what you see is what lands in the file.
4. Pick the format. TXT, DOCX and PDF carry the times when Timestamps is on. SRT, VTT and JSON always carry a start and an end for each line. Get JSON from the Export menu on your recordings list.

What's the difference between a timestamp and a timecode?

A timestamp counts hours, minutes and seconds from the start of the file: [01:23] means one minute 23 seconds in. A timecode is the video world's clock, written HH:MM:SS:FF, where FF is the frame. It also runs from the camera's or the timeline's own start time, not from the start of the file. A semicolon before the frames, as in 01:12:34;14, marks drop-frame timecode, used at 29.97 frames a second.

Most transcription software shows HH:MM:SS without frames, and so does WhisperAI: every export counts from the start of the file.

What does each export format look like?

Here are the same two spoken lines as WhisperAI writes them in each format.

[01:23] Speaker 1: So where are we on the budget?

[01:26] Speaker 2: About sixty percent, two months to go.

One time per line, start only. Under an hour it's minutes and seconds; past an hour it adds hours, as in [01:02:15]. Seconds are rounded down, so the time lands at or just before the words, never after them.

1
00:01:23,240 --> 00:01:25,900
So where are we on the budget?

2
00:01:26,100 --> 00:01:29,750
About sixty percent, two months to go.

WEBVTT

00:01:23.240 --> 00:01:25.900
So where are we on the budget?

00:01:26.100 --> 00:01:29.750
About sixty percent, two months to go.

Both caption formats give every line a start and an end to the millisecond. SRT puts a comma before the milliseconds. VTT opens with the word WEBVTT and uses a period. Give your tool the one it asks for. YouTube takes both.

{
  "start": 83.24,
  "end": 85.9,
  "text": "So where are we on the budget?",
  "speaker_label": "Speaker 1"
}

JSON gives start and end in seconds, with the speaker on each line, so 83.24 is one minute and 23.24 seconds into the recording.

How often should a transcript be stamped?

A time on every sentence is what you want for cutting a clip, and clutter when you're reading an interview.

Here's the clip-and-chapter job in practice. A podcast studio owner on Reddit had a 28-hour audiobook to split by chapter. In their words: "If it's transcribed, I can literally do a search for when they say 'Chapter 10' and then get the timecode." With the Segments view, every sentence has its own time, so the search lands on the line that says "Chapter 10" and the time in front of it is your cut point.

For YouTube chapters, paste the times into the description. YouTube wants the first one at 00:00, at least three in order, and each chapter 10 seconds or longer.

Why doesn't my transcript line up with my video timeline?

Because the two clocks start in different places. Your transcript counts from the first moment of the file. Camera cards and tapes often begin at 01:00:00:00, and editors start the program there, with a lead-in before it. Some files carry a start timecode in their metadata too.

The fix is one sum: timeline time = the timeline's start time + the transcript's time.

To land on the exact frame, take the milliseconds from the SRT, VTT or JSON export. At 25 frames a second, 0.560 seconds is 0.560 × 25 = 14 frames. So 00:12:34,560 in the SRT becomes 01:12:34:14 on a timeline that starts at 01:00:00:00. At another frame rate, multiply by that rate instead.

The same goes for a clip you trimmed before transcribing it. Its transcript starts at 0:00 where you cut, so add the time the clip started in the full recording.

Can you fix a word without losing the timestamps?

Yes. Before you export, use WhisperAI's editor to fix names, technical terms and misheard words, and to rename speakers. When you export the edited text, WhisperAI pins it back to the original timings, using the words you didn't change as anchors. So correct the transcript first and export once. It beats fixing an SRT by hand and nudging every caption.

Have a transcript you typed somewhere else and want times on it? Upload the audio, and copy your corrections into the new, timestamped transcript.

How accurate are AI timestamps?

Close enough to find and quote, worth a check before a frame-exact cut. AssemblyAI, the speech engine that transcribes files uploaded to WhisperAI, puts word times within about 400 milliseconds. At 25 frames a second that's up to 10 frames.

Before you cut, click the time on a line near the start and one near the end, and listen. If the early one matches and the late one doesn't, the video has been edited since you transcribed it, or it has a variable frame rate. Transcribe the final export instead.

What should you check before you export?

• Timestamps is on, in the view you want in the file.
• Speaker labels are on for an interview, off for captions.
• Download SRT and VTT within 30 days of uploading. TXT, DOCX, PDF and JSON keep their times after that, and a timestamped TXT turns into an SRT in the free TXT to SRT converter.

The free plan has every format and 5 minutes a month, enough to run a short clip through and see which export your editor takes. For a backlog like that 28-hour audiobook, Business Pro gives you unlimited minutes for $24.99 a month.

Sign up free and upload a file