Text to speech with subtitles

How the subtitle timing is worked out

Your text is split into short parts of a sentence or two each, and each part is generated as audio. The SRT uses the real length of every generated part, so each part's subtitles start and end where that audio does. Within a part, the timing of individual subtitle lines is estimated from text length, which is close but not frame-exact.

That has a practical consequence: shorter sentences give tighter captions. A script written in short, spoken sentences produces subtitles that track the voice closely. A long, winding sentence is broken into several subtitle lines whose timing within the sentence is estimated, so the middle lines can sit slightly early or late.

What the SRT file looks like

Each subtitle is at most two lines of up to 42 characters. You do not need to break your script into caption-sized chunks yourself; write it as normal paragraphs and the file is laid out for you.

SRT is a plain text format that video editors and many video platforms accept as a caption file. You can open it in a subtitle editor to adjust a line's wording or nudge timing before publishing, or even in a basic text editor for a quick wording fix.

Getting audio and captions in sync in your editor

Download the audio and the SRT from the same generation, without changing the text in between, so both describe the same recording. Place the audio at the very start of the timeline, or at the exact point where you want the first subtitle to appear, and import the SRT as a caption track starting at that same point. If you later slide the voiceover to make room for an intro, shift the whole caption track by the same amount.

If you edit the script after downloading, press Generate again and download both files fresh. Only the changed parts are regenerated, so this is quick, and it keeps the captions matched to what is actually said. For voiceovers you plan to cut and mix, take the audio as WAV; text to WAV explains why.

Writing a script that captions well

Write the script the way you want it to appear on screen, because the captions come from your text. One catch: if you respell a hard name to fix its pronunciation, such as 'Win' for 'Nguyen', that spelling ends up in the subtitles too. Correct those lines in the SRT after downloading; it takes a minute and viewers will see the proper name while hearing it said correctly.

Use [pause 1s] between ideas and a blank line between paragraphs. Pauses give viewers time to finish reading a caption and give you natural places to cut to new footage. Leave notes to yourself and stage directions out of the text box, since anything in the box is both spoken and captioned.

Short vertical videos and longer channel videos

For vertical short videos, captions are often the main way people follow along, and the two-line, 42-character limit keeps each caption compact on a narrow screen. The TikTok text to speech page covers pacing and a slightly faster voice for shorts. For longer videos with B-roll and music, see text to speech for YouTube.

Frequently asked questions