SRT in · speech on the cue

Turn an SRT file into speech.
Keep each cue’s window.

Each numbered cue is spoken on its own, starting at the timestamp already in the file. If the words do not fit the window they speed up, up to 1.5×, and a line that is still too long is played in full instead of being chopped.

10 free minutes · No credit card required

From subtitle cue to spoken line

  1. 1

    Upload the .srt

    Drop in the subtitle file. The cue numbers, timestamps, and text become an editable transcript.

  2. 2

    Fix the words that will be said

    Change a name, delete a caption that should stay silent, or translate the lines before any audio is made.

  3. 3

    Generate speech into the windows

    Pick a voice. Each cue is synthesised for its own start-to-end slot and laid onto one timeline.

How an SRT cue is spoken

An SRT cue is a number, a time window, and the words. Recastr speaks the words and keeps the window. The same timing rules apply when you start from VTT or ASS.

Cue index
The number on its own line is only a label. Speech is placed by the timestamp underneath it, so a file that skips or repeats numbers still lines up.
Comma timestamps
SRT writes the window as HH:MM:SS,mmm, with a comma before the milliseconds. Both ends of that window are kept, to the millisecond.
Speed cap
When the chosen voice cannot say the line inside the window at its normal pace, the whole line speeds up. The combined speed-up stops at 1.5×. Local fast and slow spans are accounted for so they do not stack on top of that.
Lines that still do not fit
The leftover words are not deleted. The line plays into the gap after its cue, and the next cue starts later only if the line reaches it. Cues that already overlapped stay overlapped.
Gaps
A gap stays quiet when the line before it fits. A line that runs long speaks into that gap instead of being cut.
File size
Subtitle uploads are read up to 1 MB. The spoken result is limited by your voiceover minutes and by a 30-minute project cap.

Reasons to hear the cues

Check a translation by ear

A translated SRT keeps the original windows. Hearing it shows which lines feel rushed before you mux the track onto picture.

Download that read as MP3

A spoken version of captions you already wrote

Captions that were proofread for the picture are a finished script. Recastr can translate them into 57 languages and speak the result in the same windows.

Hear your own voice on the cues

Clone a voice from a 15–120 second clip you have permission to use, then assign it to the transcript. An account can keep 10 clones, on the free plan as well as paid.

Spot lines the window cannot hold

A cue that hits the 1.5× cap is a writing problem: split it, or cut words, then generate again. The editor is the place to do that, before the line crowds the next one.

See what a minute costs

A cue clock, not one continuous read

Pasting every subtitle into a text-to-speech box produces a single read at the voice’s own pace, and that read walks away from the picture line by line. Recastr synthesises each cue separately and puts the audio back between that cue’s start and end.

The words are editable before they are spoken

Hearing-impaired captions, speaker names, and [MUSIC] markers are text like everything else. Delete a cue and its window stays silent. Change a word and only the speech changes; the timestamps stay.

One voice, or one voice per named speaker

A plain SRT is one speaker. When cues begin with a consistent Name: prefix, those names can be split into speakers and given different voices. WebVTT voice tags and ASS name columns are handled on their own format pages.

Watch it work

Frequently asked questions

Will the speech start when the subtitle appears?
Each cue is aimed at its own start time, and the end time is the window the voice is fitted to. A gap stays silent when the previous line fits. If that line runs long, it occupies the gap.
What happens when the line is longer than the cue?
The line speeds up, up to 1.5× combined across the voice’s pace control and a final time-stretch. Past that cap the words are not cut. The extra audio fills the following gap, and a later cue moves only if the line reaches it.
Can I translate the SRT and still keep the timing?
Yes. Translate into any of 57 languages. Timing tags that encoded the original language are dropped, because those pauses belonged to different words, and the new text is fitted to the same windows.
Do I get one audio file or one file per cue?
One file. Every cue is mixed onto a single timeline at its timestamp. MP3 is the default download; the file format is covered on the SRT-to-MP3 page.
How is this different from the SRT-to-audio page?
Same app and the same cue timing. This page is about the speech: how a cue is fitted, what happens when it does not fit, and how to edit the words before they are said. The SRT-to-audio page is the general workflow.

Hear your SRT spoken

10 free minutes included · No credit card required