Add Subtitles to Video

Captions turn spoken content into visible text that can be reviewed, corrected, and timed. Automatic transcription provides a starting draft from the actual audio. It is most useful when followed by a listening pass, especially for names, numbers, specialist terms, accents, and overlapping voices that a model may misunderstand.

Style should make the words easier to read without taking attention away from the picture. Use a clear font, enough contrast, and a placement that does not cover an important face or diagram. Check a small-screen preview as well as the larger editing monitor; a subtitle that seems spacious on a laptop can crowd a phone frame.

When you add subtitles to video, test the longest caption against both a bright and a dark frame. A contrasting outline can help the same text remain readable as the scene changes. Move the caption above an existing lower-third label when necessary, and check that the new placement does not cover a speaker’s mouth or an essential diagram.

Choose a video, generate or import caption text, then review and export. Listen across each caption boundary and edit the timing when a line appears too early or lingers after the speech. Correcting the transcript inside the editor should update both the on-screen preview and the downloaded subtitle or burned-in result.

Readable captions can reveal a pause, repeated phrase, or unnecessary introduction worth trimming. If you make a duration change after captioning, update the subtitle timeline with it. Alternatively, finish the picture edit first and transcribe the resulting clip, which gives the caption pass a stable timeline to follow.

Take a short instructional clip with three sentences and one technical name. Generate a transcript, then listen sentence by sentence. Correct the name before styling the text; attractive typography cannot fix an incorrect instruction. Break a long line where a viewer can understand the phrase, rather than forcing an entire paragraph into one screenful.

Move a caption’s start close to the spoken phrase and its end after the phrase finishes, leaving enough reading time without carrying it into an unrelated scene. Check whether punctuation matches the meaning. On a vertical crop, keep lines shorter than in a wide tutorial. Preview the longest line and a line over a bright background to test size and outline. Save the SRT if you may revise or translate the text later, and use burned-in output when the receiving player will not load a separate subtitle track.

Play the exported video with no subtitle file loaded if you chose burned-in captions; the text should remain visible because it is part of the picture. For SRT, load the file into a compatible player and check the first, middle, and last cues against speech. Confirm that your corrections, not the initial draft, are present. Review edited time ranges for overlaps, negative starts, or captions extending beyond the video. Test accented characters and the selected language for missing glyphs. If a clip has no speech, a truthful empty transcript and manual-entry option are more useful than invented sentences.

A project record and a downloaded media file serve different purposes. The record keeps the chosen tool and editing settings so you can return to recent work. The media file is the actual result you can play or share. Browser storage can retain media when permission and available space allow it, but it is local to that browser and may be cleared. Keep the original source and download important versions instead of treating the recent-project list as a permanent archive.

If a saved record asks you to reselect the source, choose the matching original rather than an unrelated file with a similar name. Check the restored parameters against its duration and dimensions. When continuing in another tool, use the actual finished result if the next operation should build on the changes already made. Starting from the original again would discard those changes unless you deliberately apply them again.

If a task cannot finish, first read the specific message and preserve the source. An unsupported codec, invalid range, model-download failure, and insufficient working storage need different responses. Retry only after correcting the likely cause. For a large or unfamiliar recording, try a short representative section that includes the difficult motion or audio; this checks the real operation without requiring a complete long export.

A failed or cancelled process is not a successful result and should not consume the daily allowance. Keep any previous usable download while retrying. After changing the source or parameters, confirm that the preview identifies the current file and the next export uses the latest settings. An old result should never silently become the answer to a new request.