How to fix a transcript before making subtitles

Automatic transcripts are a useful starting point, but a wrong name or missing word can change what a subtitle says. Fix the text before you split it into captions, then check each caption against the audio so the words and timing match.

If you need a draft, Speech to Text can transcribe an audio or video file and save the result as an SRT subtitle file. If you already have a transcript, you can skip transcription and correct the text first.

Correct the transcript before formatting subtitles

Start with a draft. Open Speech to Text, choose an audio or video file, and select the language if prompted. The tool can save the transcript as text or as an SRT file. It does not turn pasted text into timed subtitles, so a transcript you already have still needs to be timed in a subtitle editor.
Listen while you read. Play the recording and follow the transcript line by line. Pause or replay any section that sounds wrong. A transcript can look plausible while leaving out a short word, repeating one, or assigning speech to the wrong person.
Check names and specialized terms. Verify people, places, product names, acronyms, and technical words against reliable source material or the speaker's own spelling. Speech recognition often changes unfamiliar names into ordinary words.
Correct words using the audio, not guesswork. If a sentence is unclear, replay it at a slower speed if your player allows. Do not fill in a word just because it would make the sentence sound more natural. Mark a genuinely unclear passage for a second listen rather than inventing what was said.
Fix punctuation and sentence boundaries. Add periods, commas, question marks, and capitalization so the meaning is easy to follow. Use the speaker's pauses as clues, but do not put a full stop after every pause if the sentence continues.
Decide how much speech to keep. For a verbatim transcript, retain meaningful repetitions and false starts. For edited subtitles, you can remove a repeated filler or a false start when it adds no meaning, but do not remove a qualification, correction, or detail that changes the message.

Keep a copy of the original transcript before making edits. That gives you a way to check a disputed phrase or restore wording if an edit changes the speaker's meaning.

Break the corrected text into readable captions

Subtitle text should be easy to read while the video continues. Divide the transcript into short units that match the speaker's thought, rather than pasting a whole paragraph into one caption.

Keep phrases together. Break at a natural pause or between clauses. Avoid separating a person's name from a title, a number from its unit, or a short verb from the words it depends on.
Use one or two lines. A caption with a couple of short lines is usually easier to read than a long line stretching across the screen. If a caption looks crowded, divide it at a sensible point.
Do not split a word. Keep each complete word on one line, and avoid breaking a short phrase in a way that makes the reader pause in the wrong place.
Keep the wording concise without changing the point. Remove unnecessary repetition only if the subtitle style allows it. If every spoken word matters, keep the wording and make more captions instead of cutting information.

For captions meant to be accessible, include important sounds that are not speech, such as a doorbell or an alarm, when viewers need them to understand what is happening. Label a speaker when it is not clear who is talking. Do not add sound descriptions that are not supported by the video or audio.

Match the captions to the recording

Correct words are only part of a good subtitle file. Each caption also needs an appropriate start and end time. Use a subtitle editor or video editor that supports timed captions to adjust the timing, then watch the scene with the captions visible.

Start each caption when the words begin. Avoid showing a line well before the speaker says it. If the speaker pauses mid-sentence, you may need separate captions for the two parts.
Leave enough time to read. A dense caption needs more screen time than a single short word. If it disappears too quickly, shorten the text without losing meaning or split it into two captions.
Check the handoff between captions. Make sure a caption does not cover the next speaker's line, linger through an unrelated shot, or leave an awkward gap in the middle of a continuous sentence.
Watch the full video once. Read the captions while listening, not just by scanning the transcript. This catches timing drift, missed words, confusing speaker changes, and line breaks that looked fine in an editor.

If you are working with an SRT file, a caption block normally has a number, a time range, the caption text, and a blank line before the next block. A time range looks like 00:00:04,200 --> 00:00:06,000. Keep the time format and blank lines intact when editing the file in a text editor. For major timing changes, use a subtitle editor rather than changing timestamps by hand.

Fix common transcript problems

A word is missing or replaced

Replay the audio and check nearby words as well as the word itself. Recognition can mistake a word because of an accent, background noise, or a speaker talking over someone else. If you cannot hear the word clearly after replaying it, do not confidently substitute a guess.

The transcript is in the wrong language

Check that the selected language matches what is spoken, especially if the recording switches languages. Run a new transcription with the appropriate language setting if needed, then review it again. A language mismatch can produce fluent-looking text that does not match the recording.

Speaker names or labels are inconsistent

Choose one clear label for each person and use it throughout. If you cannot identify a speaker confidently, use a neutral label or rely on the scene context rather than guessing a name.

The SRT file shows the timing as text

Check that the file is saved with an .srt extension and that the timestamps keep the comma before milliseconds and the arrow between start and end times. If an editor has changed the punctuation or line breaks, restore the SRT structure or export the subtitle file again from a subtitle editor.

Keep the file private while you work

When you transcribe an audio or video file with Speech to Text, the file is processed in the browser and is not uploaded. Live dictation works differently: its live text uses the browser's speech service. In Chrome and Edge, that service works through the browser maker's servers, and live text is not available in Firefox. For confidential material, use file-based transcription and review the final subtitle file before sharing it.