I spent an afternoon once tracing a production bug back to a single missing field in a system log — not by reading the whole log top to bottom, but by finding the exact moment in the trace where the request went wrong and working outward from there. A transcript timestamp does the same job for a recording. It's the difference between re-listening to an hour of audio and jumping straight to the eleven seconds that actually mattered.
Short answer: a timestamp marks the exact point in an audio or video file where a specific word or sentence was spoken, usually written as [HH:MM:SS]. You can't add real ones to plain text after the fact — they have to come from aligning the text against the actual audio during transcription. Here's what that means in practice, and how to get a transcript that has them.
TL;DR: Timestamps let you click a moment in the transcript and jump to that exact point in the recording. They can't be faked or added to text alone — the timing data comes from the transcription process itself, matching words to the waveform. Manual transcription needs a dedicated timestamp tool; AI transcription generates them automatically as a byproduct of how it works.
Why you can't just "add" a timestamp to existing text
This trips people up constantly: you have a transcript already, and you want timestamps on it now. The problem is that a timestamp isn't a label you attach to a sentence — it's the result of matching that sentence's audio waveform to its position in the recording, in seconds. If you only have the text and not the alignment data from the original transcription pass, there's no way to reconstruct exactly when each word was spoken. You'd have to re-transcribe from the audio, using a tool that generates timestamps as part of that process, not add them after the fact.
That's the one thing worth knowing before you go looking for a "timestamp my existing transcript" tool. It doesn't really exist, because the audio is the only source of that information.
How automatic transcription gets you timestamps for free
When an AI transcription model processes audio, it isn't just guessing at words — it's aligning each word or phrase to a specific position in the waveform as it goes. The timestamp isn't a separate step; it's a natural output of how the transcription happens. That's why every AI-generated transcript can include timestamps by default, while a transcript typed out by hand needs someone deliberately noting the time as they go.
This is also why exporting a transcript as SRT — a subtitling format built entirely around timestamped segments, per the Library of Congress's own digital-format documentation — works cleanly from an AI transcription tool. The timing data was already there from the start.
Where it gets genuinely hard: overlapping speakers
Here's the part most explanations of timestamps skip entirely. A single-speaker recording is straightforward — one voice, one clean timeline, timestamps line up exactly where you'd expect. A multi-speaker meeting is a different problem, and it's more common than you'd think: research on speaker diarization puts the overlap ratio — the percentage of a conversation where more than one person is talking at once — at roughly 10 to 30 percent in typical meetings. When two people talk over each other, a naive system has to decide which speaker's words get which timestamp, and a system without real diarization will often just merge overlapping speech into one confused block with one timestamp covering both people.
This is the same investigative instinct as tracing a production incident: the trace is only useful if it tells you what actually happened, in order — not a rough approximation. A transcript where three overlapping seconds of a meeting get flattened into a single timestamp isn't giving you the moment you actually need; it's giving you the general neighborhood.
Getting this right requires speaker detection and timestamp alignment working together, not as separate features bolted onto each other afterward.
What this actually looks like in practice
Once you have a transcript with real timestamps, the practical use is simple: you turn on a timestamps view, and every line in the transcript shows the exact moment it happened. Click on any line, and the audio player jumps straight there — no scrubbing through fifty minutes of a recording hoping you land close to the right spot. Combined with speaker labels, you get "at 14:32, Karim said..." instead of a wall of undated text.
The same timing data is what makes an accurate SRT export possible for anyone turning a recording into subtitled video content.
When you don't actually need timestamps
If you're transcribing a five-minute voice memo you're going to read once and delete, timestamps add nothing — you'll read the whole thing anyway. They earn their value specifically on long recordings you expect to revisit: a two-hour meeting you'll need to fact-check weeks later, an interview you're pulling quotes from, a lecture you're studying from repeatedly. Don't treat timestamps as a feature you always need; treat them as the thing that saves you time on recordings long enough to need saving time on.
What this costs
Flat pricing, timestamps included on every plan that generates a transcript:
- Free ($0/mo): 2 transcriptions a day, up to 20 minutes per file, with timestamps and SRT export included.
- Starter ($5/mo): Up to 45 minutes per file, 20 hours a month.
- Pro ($10/mo): Up to 2 hours per file, 45 hours a month, plus speaker detection and the Vocabulary Library for names and terms.
- Unlimited ($15/mo): Up to 6 hours per file, unlimited transcription.
Every new account includes a 3-day Pro trial.