Turn a finished transcript into an SRT file with real timestamps
If your interview, podcast or lecture is already transcribed, by hand or by a service, all that is missing is the timing. Upload the recording with the transcript and every line comes back as a timed SRT and VTT cue in your exact wording, speaker labels included. Recordings of up to 10 minutes are free in full; longer ones cost $5 once, for up to 6 hours.
- Transcript as TXT or DOCX, or pasted; typed by you or delivered by a transcription service
- Spelling, punctuation and speaker labels such as "Host:" kept exactly as written
- Each line timed from the moment it is spoken in the recording
- Long paragraphs split into readable cues at sentence and comma breaks
- SRT and VTT in UTF-8 with a byte-order mark, for players that need one
You already have the words
A transcript that someone checked carefully is worth keeping as it is. Running the audio through a fresh transcription just to get an SRT throws that work away: names come back misspelled, your corrections disappear, and you proofread a second time. This tool keeps your text and only adds the clock. Speech recognition finds when each spoken word occurs, and code matches your transcript to those words.
A plain transcript gives a player nothing to work with. Even one with a timestamp every few minutes falls short, because subtitles need a start and an end for every few seconds of speech.
Preparing the transcript
- Keep one paragraph or speaker turn per line; your line breaks become cue boundaries
- Speaker labels stay in the cue text as written, so "Dr. Patel:" will show on screen; delete them first if you want only the spoken words
- Old timestamps typed into the text, such as [00:14:02], are treated as words, so remove them or they will appear in the subtitles
- Notes such as [inaudible] or [crosstalk] stay in the cue where you put them
- Passages that were transcribed but later cut from the recording should go, since there is no speech to time them by
Where the speech and the text differ
Clean verbatim transcripts leave out "um", false starts and repeated words. That causes no trouble: heard words with no match in your text are skipped, and your tidier line keeps its place. Where the recognizer hears "twenty-five" and you typed "25", or it misses a mumbled word, the time is taken from the words on either side.
Two people talking over each other, or a guest far from the microphone, can make a cue start slightly early or late. For Amharic, Tigrinya, Somali, Dari, Kurdish, Burmese and Pashto, the timing comes from a general AI model and is less exact.
What a timed transcript is good for
- Captions for the video version of a podcast episode
- Subtitles for a recorded lecture or webinar on a course platform
- Interview footage in an editor, with each answer placed on the timeline
- Accessible captions for a talk you transcribed months ago
How the tool works, what it costs and how your files are handled is on the Sync Transcript to Audio page; prices for every tool are on the pricing page.
Last updated
Questions and answers
Can ChatGPT add timestamps to my transcript?
It can arrange a transcript as SRT and will write in times if you ask. Without hearing the recording, those times are spread evenly or guessed, so on an hour-long interview they wander further off with every minute. On long text it also tends to shorten or merge lines. Here the times come from the audio and the text is left alone.
Will speaker names appear in the subtitles?
Yes, if they are in your text, because labels are kept exactly as written. Remove them before uploading if you only want the spoken words on screen.
My transcript leaves out the ums and false starts. Is that a problem?
No. Extra sounds in the recording that are not in your text are passed over, and your cleaned-up lines are still timed from the words around them.
In which formats can I upload the transcript?
Paste it, or upload a TXT or DOCX file. If you already have a subtitle file, SRT, VTT, ASS, SBV and LRC are read as well, and their cue layout is kept.
How much does a timed transcript cost?
Up to 10 minutes of recording is free in full. Anything longer is $5, paid once, for up to 6 hours. Unlimited at $19 a week covers up to 12 hours per job.
Who can see my transcript?
Only you. Your files are sent to an AI model provider only to be processed and are not used to train models. Jobs started without an account are deleted after 7 days, and you can delete a job yourself whenever you like.