Make a simple SRT file

What to do

Write a numbered block, a start and end time separated by an arrow, then the caption text. Leave a blank line between blocks.

Build and test two cues first

Use a plain-text editor and save as UTF-8. Add a cue number, a timestamp line, and the first caption. Leave a blank line before the next block. Use a second cue with a later time so you can confirm the order. Load the short sample in a compatible player before writing an entire track. Check that the filename ends in .srt rather than .srt.txt.

Example or check

Use 00:00:01,000 --> 00:00:03,500 for a cue that appears from one second to three and a half seconds.

Related checks for captions and accessibility

Start with a matching video version

Use captions that belong to the exact cut you plan to publish. A different trim, changed introduction, or alternate frame rate workflow can affect the relationship between text and playback. Save a copy of the original timed file before editing. Note the first line, a line near the middle, and a line near the end so you can check synchronization after any change. If a video has no usable timed track, transcribe the speech first and assign times against the actual audio. Formatting a file cannot infer missing words or repair timings that were never written.

Read the structure of a caption cue

A timed cue has a start, an end, and the text that should appear during that interval. In SRT, a common pattern is a sequence number followed by comma-separated milliseconds in the timestamps. WebVTT uses a header and dot-separated milliseconds, and it can carry settings and metadata that do not map cleanly to SRT. Keep blank lines between cues, use UTF-8 for accented characters, and inspect the saved file in a plain-text editor when troubleshooting. If you move between formats, compare cue counts and check that the player reads the beginning and end.

Distinguish offset from drift

Choose two audible words far apart and compare the cue times against those words. If both cues appear a similar amount early or late, a uniform offset may solve the issue. If the error grows over time, investigate the cut, edit speed, or source timing rather than applying a single shift to everything. At a video boundary, a negative shift can move a cue partly or fully before zero. Review what happened to the first visible line and make sure the resulting file still makes sense to the viewer. Always retain an unshifted copy.

Keep styling expectations realistic

A basic format conversion can preserve speech text and timings, but players do not render the same layouts or formatting. WebVTT settings may disappear in a simpler SRT file, and some platforms ignore text styling. If an overlay or two-position subtitle track is important, keep the original and use the supported format for that destination. Preview the actual file in a suitable player. For a video with large labels near the bottom edge, test whether captions cover an instruction. Move important on-screen text or adjust the design before assuming the player will position captions differently.

Troubleshoot an upload methodically

Start with a small known-good file containing two timed cues. If it uploads correctly, compare the failing file's extension, encoding, header, timestamp separators, and blank lines. Look for a cue whose end is earlier than its start or whose timing arrow is missing. If the short file fails too, verify the chosen language, upload mode, and account permissions in Studio. After any successful upload, inspect the rendered captions on a video preview. A file can be accepted and still contain a spelling mistake, poor timing, or a wrong language assignment.

Document what a conversion changed

When exporting a new subtitle file, note the source format, output format, language, and any timing offset applied. If the output is missing position information or comment blocks, retain the original instead of manually guessing how those details were represented. Spot-check any character outside the ordinary English alphabet. A collaborator should be able to tell which file is the current review copy and which one is the untouched input. If you later discover a timing issue, the record lets you undo a shift or convert again without compounding the earlier changes.

Edit text for the viewing experience

Read each cue while watching the image. Look for names, numbers, negatives, and units because small transcription mistakes can change the instruction. Use line breaks at a phrase boundary when possible. A pause, interruption, or speaker change may call for a separate cue even if the full sentence looks short on paper. Sound cues should explain meaningful off-screen events when the viewer needs that information. Avoid describing every background sound. A good check is to mute the video and ask whether a viewer can still follow the action and understand who is speaking.

Check short and long cues together

Pick a few quick exchanges and a few long sentences when reviewing readability. A quick cue needs enough time to recognize who is talking; a long sentence may need splitting so it does not block the image or disappear before it can be read. In a technical video, put a value and its unit together where possible. Check line breaks after translation because phrases can grow in length. If captions overlap in an unintended way, compare the start and end times of neighboring cues rather than correcting only the one that looks wrong.

Manage caption files for revisions

Name the caption track with the video version and language, such as a short identifier plus en and the edit number. When you receive a corrected version, compare changes before replacing the earlier track. Check that the editor changed the intended line without altering its time by accident. If the video was recut, use a dedicated working copy for retiming. Keep private transcripts on your own system. A browser-based conversion can use local files, but publishing captions in Studio makes the selected track available with the video according to its visibility.

Use the right check after publishing

On the watch page, turn captions on and choose the intended language. Check the first line, one section in the middle, the last spoken line, and any unusual character or sound cue. Switch devices if the video is likely to be watched on a phone. Look for conflicts with graphics and product labels. When requesting review, share both the time and the exact correction you need. A note such as “at 1:18, change 15 volts to 50 volts” is actionable and lets the editor check the source audio before editing.