Subtitles and captions
Words on screen that hold their shape at every sentence length, accurate wherever the audio carries meaning, and the judgement about when to burn them in and when a caption track serves the viewer better.
The short version
Is this for you? For captions burnt on too early and lines that break mid-thought. It is as much the judgement about when to burn them in and when a caption track serves the viewer better as the craft of getting them right. Not if silent viewing is not a factor for your audience; where everyone watches with the sound on, burnt-in captions cost more than they return.
What leaves your plate. Subtitling begun only once the cut is locked, so the text never chases a moving edit, reading speed treated as the real constraint, and the preset tested short, typical and long before it ships, at a two-line ceiling.
What starts working. Words that hold their shape at every sentence length and clear of every platform's own furniture, accurate wherever the audio carries meaning, and burned in or supplied as a track on purpose rather than left to a default.
The smallest way in. A free audit: three fields, and a written reading back, usually within about three working days. It commits you to nothing.
What to read next. The five levels to see where this fits, or the casework to see it done.
When captions help, and when they get in the way
Subtitling starts only once the cut is approved
The locked timeline goes out as audio alone and comes back as an SRT. What travels with it is deliberate: the file name exactly as it reads on the tracking sheet, the folder the captions are to land in, and notes on any term or name that will otherwise be misheard. Caption a cut that is still moving and you will do it twice, then spend a week hunting the version that got away.
One partner gets no burned-in captions at all
On a recipe channel whose owner adjusts her wording just before posting, the captions are typed natively in the app instead. The reason is hers and not ours: typing them in the app keeps her last-minute wording under her own control, where burning them in would take that final edit away from her. Choosing not to burn them in is a service decision, and it is one of the few places a caption preset genuinely loses. Either way an accurate caption of the meaningful audio still exists; what changes is whether it is painted into the picture or supplied as a track the viewer controls, never whether the words are there at all.
Reading speed is the constraint, so lines break where a sentence breathes
Nobody chooses the pace a caption is read at; the speech sets it. So lines break at clauses instead of at the edge of the frame, two lines is the ceiling, and a sentence that will not fit gets split where a person would have paused. A caption broken mid-phrase gets read twice, and the second reading is taken out of the picture.
The preset is tested at three sentence lengths before it ships
Short, typical and long. A caption style that looks right on four words and breaks its own margins on fourteen is not a style; it is a screenshot. The face, the size, the contrast and the safe margins come from the partner's own type law, so captions read as part of the brand once they are already accurate and legible, and not as something the editing software supplied on the night.
How the words get onto the screen
The routing, in order
Cut approved. Audio exported on its own. The subtitler receives the file name, the destination folder and the notes. The SRT comes back and is laid against the locked timeline. On the short vertical cuts that same SRT is the source, so the wording agrees between a long video and every clip taken out of it.
Sizes, safe areas and the corner nobody looks at
Vertical captions sit inside the platform's safe area, clear of the interface furniture along the top, the bottom and the right. Where an inset window and a piece of notation share a frame, each gets a fixed corner and keeps it, so a viewer learns within one video where to look for which.
The notes travel with the audio
Anything a transcriber cannot get from the recording alone goes with it: a technical term said quickly, a name spelled unusually, a passage where two people overlap. Those are precisely the errors an automated pass leaves behind, and precisely the ones an expert audience spots first and mentions in the comments.
Captions are a decision. On one channel the answer was to type them into the app rather than burn them in, and that was the correct piece of subtitle work.
Knowing when a caption track serves the viewer better than a burned-in one is as much of the craft as fitting them on.
What changes for you
Words that stay legible at every sentence length and inside every platform's furniture, and an accurate caption of the meaningful audio always present, whether painted into the picture or supplied as a track the viewer controls. Either way it was decided rather than defaulted.
Related: Titles and motion · Long-form editing · Shorts and Reels · Vertical video editing
Send the channel and we will read the edits themselves: pacing, cuts, consistency across recent videos. The audit names what post-production would change first, in writing, before anything is committed.
Request your free audit.
Three fields, and a written reading back: a fit snapshot, two or three prioritised opportunities, and a recommended next step.
A person reads the channel and writes the audit by hand: a considered read typically takes three working days. That is the usual shape, not a promised turnaround. We use these details only to reply to you: no lists, no lurking.
What you will get
A fit snapshot: where your channel stands, and whether we are a match.
Two to three opportunities: specific, prioritised, yours to keep.
A recommended next step, even if that step is not us.
The audit is free and commits you to nothing: nobody follows up with a call you did not ask for.
The casework shows all of this done.
The case studies are the same work with the numbers, the sources and the dates attached.
See the casework →