Burned-in or separate subtitles
There are two ways to put subtitles on a video, and they solve different problems.
Separate (soft) subtitles live in their own file, usually SRT or VTT, or as a track inside an MKV or MP4. The viewer can switch them on and off, pick a language, and the text stays sharp at any size. The catch is that the player has to support them, and many places where short videos are watched do not read a subtitle file you supply alongside the video.
Burned-in (hard) subtitles are drawn into the frames themselves. They are just part of the picture, so they appear everywhere the video plays: in a messenger preview, on a social feed, on a phone in silent mode, embedded on a site with a minimal player, or on a TV that ignores subtitle tracks. They cannot be turned off, translated by the player or restyled later.
Burned-in subtitles are the practical choice for social clips, where a large share of viewers scroll with the sound off, and for any video you hand over to someone else without control over how it will be played. For a film archive or a multilingual library, separate files are better.
What happens to your video
This tool runs entirely on our servers in one job with two stages:
- Speech recognition. The audio is transcribed with the Whisper speech model into timed segments, exactly as the transcription tool does when you choose SRT output. Silence is skipped and punctuation is added automatically.
- Rendering. The segments are drawn onto the frames and the video is re-encoded to H.264 MP4 at a quality setting (CRF 23) that keeps the picture close to the original. The sound is re-encoded as AAC at 128 kbps, and the file is arranged so it starts playing before fully downloaded.
The resolution and frame rate stay as they were. MOV, MKV and WebM inputs come back as MP4, which is what social platforms and messengers expect. The finished file is named after the original with -subtitled added.
Because the whole picture is re-encoded, the output size can differ from the input. A clip recorded in HEVC on a recent iPhone may come back somewhat larger, since H.264 needs more bits for the same quality. If size matters, run the result through video compression afterwards.
How the subtitles look
The style is fixed and designed to stay legible on any background:
- White sans-serif text, centred at the bottom of the frame.
- A semi-transparent dark box behind each line instead of an outline, so the words read over bright skies and busy scenes alike.
- The text size scales with the video height, so a 720p and a 4K version of the same clip look the same proportionally.
- A bottom margin of roughly a tenth of the frame height, which keeps the text clear of progress bars and the caption area that phone apps overlay.
Each subtitle corresponds to one recognised phrase. Fast or continuous speech produces longer phrases, which wrap onto two or more lines; this is more noticeable on vertical 9:16 video, where the frame is narrow.
Getting a clean result on the first try
Since the text cannot be edited before it is burned in, the quality of the recognition is what you get. A few things raise the odds:
- Set the spoken language instead of leaving detection on when the clip opens with music or a jingle. The subtitles are written in the language that is spoken; picking a different one does not translate them.
- Use the cleanest audio you have. Background music under speech is the most common reason for missing or garbled lines.
- Watch for names and numbers. Brand names, people's names and figures are where errors cluster. If they matter, preview a short test clip first.
- Keep the original. Burned-in text is permanent, so store the clean version in case you need it without subtitles or with corrected text.
If you need full control over the wording, generate an SRT with transcription, correct it, and load it as a subtitle track in your video editor instead.
Time, limits and privacy
Processing time depends mostly on the length of the recording, since recognition runs at roughly real time, so trimming a clip to the part you will actually post with trim video saves both waiting time and quota. Free use covers one video a day, up to 10 minutes and 500 MB, without an account. Premium removes the daily limit and accepts recordings up to 90 minutes. The uploaded video is deleted once processing finishes, and the result expires automatically within about an hour.