Burn Subtitles Into Video

It means the text becomes part of the picture — the same pixels as the video itself. There is no separate track, so there is no captions menu to find, nothing for a player to fail to load, and no font for it to substitute. They are also called hardcoded or open captions. The trade is that they are permanent: a viewer cannot switch them off, change the size, or pick another language, and you cannot edit them afterwards without re-rendering from the original.

Burn them when you do not control the player. Instagram, TikTok, most embedded players and anything shared as a file will either ignore a sidecar SRT or bury it behind a menu nobody opens — and the majority of social video is watched with the sound off, so captions that are optional are captions that are missed. Keep them separate when you do control the player, when you need more than one language, or when accessibility guidance applies: a viewer who needs larger text can resize a real caption track and cannot resize a burnt one.

No. It is decoded, captioned and re-encoded entirely in your browser, and no part of it is transmitted. There is no file size limit for the same reason — there is nothing to upload.

SRT and WebVTT, either as a file or pasted straight in. Both are handled by the same parser, since they differ in almost nothing that matters: a header line, a comma or a full stop before the milliseconds, and optional cue settings after the arrow. Blocks that cannot be read are skipped rather than rejected, because a file with one malformed cue in the middle is still a file worth using.

It is a readability check. Subtitling standards cap reading speed — Netflix uses 17 characters per second for English, and lower for children’s programming — because a subtitle that is accurate but gone before it can be read is no more useful than a wrong one. Automatic transcription produces these constantly, since it times cues to the speech rather than to reading. The tool flags them before you burn, which is the only moment the fix is cheap.

Because only one of them will appear. If a cue starts before the previous one has ended, this tool shows the earlier one and the later one is simply lost for the overlapping stretch. Most players behave the same way. It is a common artefact of hand-edited timing and of merging two subtitle files, and it is invisible until the exact moment it matters.

If they are consistently early or late, yes — the timing offset shifts every cue together, up to ten seconds either way. Scrub to a line of speech and nudge it until the text lands. What the offset cannot fix is drift, where the error grows through the video: that usually means the subtitles were written against a different frame rate or a differently edited cut, and no single shift will line them up.

Slightly, as with any re-encode, and you control how much. Every frame is decoded, captioned and encoded again, so the output is a second generation. The bitrate is derived from your source file rather than from a fixed number, and High or Maximum keeps that generation from showing. What the tool never does is record playback into a recorder — that approach drops frames under load, and a dropped frame here can mean a subtitle that flashes or never appears at all.

Yes, and that is the point of flagging them. Cues that read too fast can be extended into the silence that follows them, in one click; the tool reports honestly how many it could not help, because a cue with another right behind it has nowhere to grow and needs splitting or shortening instead. Overlapping cues can be separated by trimming the earlier one, which is the safer repair since the later cue is the one still being spoken. Cues carrying more text than fits in two lines can be split, at a sentence end where there is one in the middle third and at the nearest word boundary otherwise, with the time shared in proportion to the text on each side. Every repair can be undone, and you can also edit any cue text or timing by hand with the reading speed updating as you type.

Yes — save them as SRT or VTT at any point. Repairs and hand edits are included, so the work is not trapped inside one render. It is worth doing before you burn: the corrected file is the one you will want next time, and for uploading alongside the video where a real caption track is accepted.

No. Render a ten-second sample from wherever the playhead is, at the real settings, and check it before committing. It starts at the playhead rather than at zero because the opening of a video is usually a title card with nothing to caption. A twenty-minute burn is an expensive way to discover the text is too small.

They shade the part of the frame the platform covers with its own interface. TikTok stacks a caption, a username, a music credit and a button rail over roughly the bottom fifth and right sixth of the frame; Reels and Shorts are similar; broadcast uses a traditional ten per cent margin. Subtitles placed there are not clipped, they are simply covered — and you will never see it happen in your own player, because your player shows none of that furniture. The guides are preview only and never burnt in, and the tool warns when your margin would put the text underneath.

Yes, ten ways. Fade, pop, slide up, slide in and blur in animate the cue as a whole. Typewriter reveals letters across the cue. Word by word and word pop bring in one word at a time, which is the look most short-form social video uses. Karaoke highlight shows the whole line from the start and moves a colour across it as each word is spoken. Every one is a pure function of where the frame sits inside its cue rather than of wall-clock time, so the animation is identical in the preview, in the export, and again if you re-render at a different frame rate.

They spread the words evenly across the cue. Neither SRT nor WebVTT carries word-level timestamps — only the start and end of each cue — so an even distribution is the honest approximation, and it is a good one: the animation starts and finishes exactly with the cue, which is the part a viewer notices. Keeping cues short makes it more accurate, which is what you want for readability anyway.

It depends on where it is going. For film, television, documentary or anything long, no — broadcast subtitles do not move because movement pulls the eye away from the picture, and over an hour that is exhausting. For short-form social video, yes: word-by-word captions measurably hold attention, most of that audience is watching with the sound off, and the clip is over in thirty seconds. The looping preview is there so you can judge it rather than guess, and a still frame will not tell you — it shows one instant of an entrance.

It balances the lines rather than filling them. The obvious approach is to pack the first line until it will not take another word and drop the rest onto the second, which is how you end up with a full line followed by the single orphan word "letter." — the text was not too long, it was badly divided. So the number of lines is settled first, and then the narrowest width that still fits in that many lines is used, which pushes the break towards the middle at no cost: the same words, the same number of lines, spread evenly. On a real example the difference between the two lines went from 806 pixels to 31. Line breaks already present in your subtitle file are treated as deliberate and kept, since whoever wrote them usually broke at the sense.

Because a subtitle in points or pixels means something different at every resolution — 24 pixels is comfortable on a 480p clip and nearly invisible on 4K. Expressed as a percentage of frame height, one setting looks the same on every video, and you can change the output resolution without restyling.

An outline follows the shape of the letters and covers almost none of the picture, which is why broadcast subtitles use one. A box is more legible over busy, bright or fast-moving footage but hides more of the frame. The band is a compromise for talking heads, where the bottom of the frame is rarely doing anything. Over unpredictable footage a heavy outline is usually the safest single choice.

Bottom, with a margin, for almost everything. Keep them well inside the edge: players, phone interfaces and social apps all overlay controls, progress bars and usernames along the bottom, and text flush to the edge disappears behind them. For vertical social video the Social preset places them nearer the middle for exactly this reason. Move them to the top only when something important sits at the bottom of the shot.

Yes — our video transcription tool generates them from the speech, on your own machine, and exports SRT or VTT you can load straight in here. It is the natural pairing: transcribe first, correct the text, then burn it in.

It depends on the length, resolution and frame rate, but it is faster than the video plays, because frames are pulled and encoded as fast as the machine allows rather than in real time. A minute of 1080p at 30 fps is a few thousand frames, each decoded, drawn and encoded. Lowering the frame rate or the output size is the quickest way to speed it up.