The short answer
Create a plain UTF-8 text file with a numbered block for each caption: start and end timestamps, the words to display, then a blank line. Save it with a .srt extension and test it with the matching video. A transcript gives you the words; you still need to decide when each caption appears and disappears.
1. Start with the exact video you will publish
Keep the final video edit and the subtitle file together. Watch the video once and write down the spoken words before adding timings. Check names and numbers against the recording. If you change the edit after timing the captions, check the whole subtitle file again: even a short opening trim can move every line.
For this walkthrough, use the same 24-second espresso demonstration as our subtitle-burning guide. It combines edited Mixkit footage with an original script and synthetic narration. The downloadable two-cue starter covers only the first two sentences, ending at 8.720 seconds. The complete six-cue file below covers the rest of the narration; the starter is not a complete caption track for the full clip.
2. Build the four parts of each SRT block
An SRT block contains a sequence number, a timing line, caption text and a blank separator. Start with 1 and number the following blocks in order. Write the timing as hours:minutes:seconds,milliseconds, using a comma before the three millisecond digits and an arrow between the start and end. The end must be later than the start.
The example is taken from our actual demonstration file. Its first caption begins 760 milliseconds into the video and ends at 4.760 seconds. The second begins at 5.160 seconds. Those values belong to this recording, not to every video with the same words. Replace both the words and the times when adapting the template.
A line break inside the caption creates a second text line; a blank line ends the block. In this sample, each sentence is divided into two readable lines. Keep the break at a sensible phrase boundary and check the result at the size people will watch. File syntax alone cannot tell you whether the text covers an important control or is comfortable to read.
1
00:00:00,760 --> 00:00:04,760
This short demonstration shows
an espresso machine in use.
2
00:00:05,160 --> 00:00:08,720
Coffee flows from the spout
into a metal cup below.
3. Save plain text with the right filename
Use an editor that can save plain text, choose UTF-8, and save a working copy such as coffee-demo.en.srt. Avoid a word-processing document with fonts, margins or rich-text markup. On macOS, switch TextEdit to plain text before saving if you use it; in other editors, use the equivalent plain-text mode.
Check the actual extension after saving. A file named coffee-demo.en.srt.txt is still a text-file download to tools that detect the format by its extension. Reopen the saved file and confirm that it contains the numbered blocks you expected. Renaming a Word document or an untimed paragraph to .srt does not add the required structure or timestamps.
UTF-8 is a practical delivery choice, and YouTube specifically requires it for basic SRT uploads. That does not mean every platform preserves every SRT feature. Keep this first file simple: ordinary text and timing. Review the destination’s own format requirements before relying on colors, placement or inline styling.
4. Treat TXT-to-SRT timing as a draft
If you already have a transcript, Recast’s free converter can turn non-empty text lines into SRT blocks. Its current TXT conversion assigns three seconds to each line, starting at zero. It does not listen to the video or align words to speech. Put one intended caption on each line before converting, and treat the resulting timestamps as placeholders.
We checked the two lines below against the converter code used by the site on November 4, 2026. The output starts the first line at 0.000 seconds and the second at 3.000 seconds. Our actual sample needs starts of 0.760 and 5.160 seconds. Conversion therefore creates a useful editable structure, but these two lines are not synchronized until you correct their timings.
The three-second value is a tool default, not a caption-reading recommendation. A long sentence and a short phrase should not automatically receive the same display time. Listen to the final recording and adjust each cue, or use a subtitle editor with a video timeline. This converter is not a timing editor or speech transcription service.
This short demonstration shows an espresso machine in use.
Coffee flows from the spout into a metal cup below.
1
00:00:00,000 --> 00:00:03,000
This short demonstration shows an espresso machine in use.
2
00:00:03,000 --> 00:00:06,000
Coffee flows from the spout into a metal cup below.
5. Check timing against the picture and voice
Play the file with the exact video, checking the first cue, a middle cue and the ending before reviewing the complete playback. Read at normal speed. A caption should accompany the relevant speech without flashing briefly or remaining on screen after the explanation has moved on. Listen for any missing words rather than judging only whether the text looks tidy.
If all captions are shifted by the same amount, investigate an opening trim or other fixed offset. If they gradually drift, check that the subtitle file belongs to this version and playback speed. If only one caption is wrong, inspect that block. Converting SRT to VTT changes the file format while preserving cue times; it cannot fix a timing problem.
For a caption track intended to convey the audio to someone who cannot hear it, include meaningful sound information and speaker identification where needed. A speech transcript alone can miss that information. Have someone review the saved file in its destination player, including on a small screen, before treating it as finished.
6. Deliver a selectable track or a captioned MP4
Keep the reviewed SRT as your editable source. Upload it as a separate track if the destination supports SRT. For a web player that needs WebVTT, convert it and check the new file in that player. SRT contains text and timestamps; it does not contain video or audio, and you cannot make a playable MP4 from the subtitle file alone.
If the placement needs words permanently visible in the picture, provide both the clean video and the reviewed SRT to the subtitle burner. That produces a different deliverable: a video whose text cannot be switched off. Save the clean master too, because correcting a burned caption means making another video export.
The converter and burner both run locally in the browser and need no account. They do not translate the transcript. Recast Studio is a separate workflow for supported video translation and requires sign-in and credits; creating an SRT template does not require starting a Studio job.