Transcript to timed subtitles
Sync your transcript to the audio and get a timed SRT
You already have the words: a video script, an edited transcript, lyrics of a talk or an SRT whose timings no longer match. Upload the recording with that text and every line comes back with a start and end time taken from the speech, as SRT and VTT. Your wording is kept exactly; the tool only adds the timing. A recording of up to 10 minutes is free in full. A longer one costs $5, paid once, for up to 6 hours.
- One audio or video file up to 2 GB and 6 hours (12 hours on Unlimited)
- Paste the text, or upload it as TXT, DOCX, SRT, VTT, ASS or SBV
- Your exact words in the file, never rewritten by the AI
- Your line breaks kept; long lines split into subtitles of up to two lines
- An out-of-sync SRT re-timed from the audio, with its lines unchanged
- Word timings from speech recognition, matched to your text by code
- Up to 10 minutes free in full, first part of longer files free · whole recording $5, one payment
What you get back
Your own text as a numbered, timed subtitle file that follows the speech.
- SRT and VTT files where every cue is your text, with a start and end time taken from when those words are spoken.
- Subtitles of at most two lines of about 42 characters, split at sentence and clause breaks where possible.
- When you upload an SRT or VTT, the same cues with the same line breaks, only with new times.
- Cues in order, never overlapping, and on screen long enough to read.
- UTF-8 files with a byte-order mark, which older players and TVs need for non-Latin scripts.
- Lines you correct on screen are used in the files you download.
Doing it in a chatbot, and with this tool
A chat assistant can turn a transcript into SRT format. What it cannot do on a long file is hear when each line is said, so the timestamps it writes are guesses.
| What you need | A chatbot | This tool |
|---|---|---|
| SRT formatting for a few lines | Often fine if you give it the times | Also fine, and free in full up to 10 minutes |
| Times that match the speech | Invented or evenly spaced, drifting off the audio | Taken from the words as they are spoken |
| A one-hour transcript | Too long for one reply; stops partway | Up to 6 hours in one job |
| Your exact wording | Lines get reworded, merged or dropped | Your text kept letter for letter |
| Fixing an SRT that drifts | A guess at a fixed shift | Every cue timed again from the audio |
| A file YouTube or an editor accepts | Text to copy out of the chat | SRT and VTT downloads |
How to sync a transcript to audio
-
1
Add the recording and the text
Drop the audio or video file (MP3, M4A, WAV, MP4, MOV and more) and your text as a TXT, DOCX or SRT file, or paste the text into the box. A public Google Drive, Dropbox or direct link also works for the recording.
-
2
Set the language if you like
The spoken language is detected on its own. Setting it helps when the recording starts with music or with a few words in another language.
-
3
Check the first part, then download
The first part appears on screen as timed cues. If the timing looks right, pay $5 for the whole recording and download SRT and VTT.
Why a chat assistant cannot time your text
Turning a transcript into the SRT layout is easy for a chat assistant. The hard part is the timing: to know that a sentence starts at 12:41.3, something has to listen to the recording and find that sentence in it. On a long file, an assistant that only sees your text writes timestamps that are evenly spaced or made up, and they drift further from the speech with every minute.
This tool listens with a speech recognition model that gives a time for every word it hears. Your text is then matched to those words by code, not by the AI: phrases that appear once in both act as anchors, and the words between anchors are matched one by one. Where the recognition heard a word differently, wrote "13" for "thirteen" or missed a word, the time is taken from the neighbors. The words in the file are always yours.
What people use it for
- Video creators who wrote a script and want captions that match what they actually recorded
- Podcasters and journalists with an edited transcript who need SRT for a video version
- Typed transcripts of interviews, hearings or lectures that need timestamps for review
- Subtitles from another release or an older cut that drift, re-timed from this audio
- Teachers and language learners who want a text and a recording to play in sync
Getting a good result
The closer your text is to what is said, the better the timing. Small edits are fine: a removed "um", a corrected name, a sentence said slightly differently. A paragraph that is in the text but never spoken has no speech to attach to, so it is squeezed in next to its neighbors; delete such passages first. Long stretches of music, applause or another language are skipped and do not affect the lines around them.
If you want control over where each subtitle breaks, put one subtitle per line in your text; lines longer than about 84 characters are split at sentence or clause breaks. To keep an existing subtitle layout, upload the SRT or VTT itself: its cues and line breaks stay exactly as they are and only the times change.
Accuracy and limits
Timing follows the speech recognition, which works best on clear speech in widely spoken languages. For Amharic, Tigrinya, Somali, Dari, Kurdish, Burmese and Pashto the word times come from a general AI model and are less exact; check the free first part. Overlapping speakers and loud music under speech make some lines start a little early or late.
Each job takes one recording and one text. This tool keeps your words and adds timing; if you do not have a transcript yet, the long audio tool writes one, and the video subtitle tool makes subtitles straight from the audio.
Sync pages by use
Pages for common jobs, each opening the tool with the right options already selected.
Sync my script to my voiceover as SRT
You wrote the script, recorded the voiceover, and now you need captions that appear when you say each line. Up...
Read more →Turn a transcript into an SRT with timestamps
If your interview, podcast or lecture is already transcribed, by hand or by a service, all that is missing is...
Read more →Resync an SRT that drifts out of sync
Subtitles that start in sync and slowly drift, or that come from a release with different cuts, cannot be fixe...
Read more →Turn my YouTube script into captions
Automatic captions often get names, slang and product terms wrong, and fixing them line by line in YouTube Stu...
Read more →What it costs
Short recordings, first part
Free
A recording of up to 10 minutes is synced in full for free, downloads included, no account. For a longer recording, the lines of the first part are shown on screen with their times; a free account shows more of it.
Whole recording
$5
One payment for one recording of up to 6 hours with its text. SRT and VTT. No subscription.
Unlimited
$19 / week
Every tool on the site, recordings up to 12 hours per job. Renews weekly, cancel any time.
Questions and answers
How do I turn a transcript into an SRT with timestamps?
Upload the recording and paste the transcript or upload it as a TXT or DOCX file. Every line is timed from the speech; check the free first part, unlock the whole recording for $5 and download the SRT.
Can ChatGPT sync my script to the audio?
It can format a script as SRT, but on a long recording it cannot hear when each line is spoken, so the times it writes are guesses that drift. This tool takes word timings from speech recognition and matches your text to them.
Will it change my words?
No. The file contains your text exactly as you wrote it. Speech recognition is only used to find when each word is spoken.
Can it fix an SRT that is out of sync?
Yes. Upload the recording and the SRT or VTT. Every cue keeps its text and line breaks and gets new times from the audio, which also fixes drift that a single shift cannot.
What if my text does not match the recording exactly?
Small differences are fine; missing or changed words take their time from the words around them. Whole passages that are never spoken should be removed first, because there is no speech to time them by.
Which languages work?
The spoken language is detected automatically, and 77 languages can be chosen as a hint. Timing is most exact in widely spoken languages; check the free first part in yours.
How much does it cost?
A recording of up to 10 minutes is free in full. A longer one is $5, paid once, for up to 6 hours. Unlimited costs $19 a week, renews weekly and covers up to 12 hours per job.
Is my recording private?
Only you can open your job. The recording is sent to an AI model provider only to be processed and is not used to train models. Jobs started without an account are deleted after 7 days, and you can delete a job yourself at any time.
Updated October 25, 2026 · APPFACT STUDIO LTD