Captions and transcription
Most video on a phone is watched without sound, at least at first. Captions are the primary channel there, not a courtesy.
Tools → Captions.
Transcribing
Transcription runs on your own machine. Nothing is uploaded and no service is called.
Choose the clip, set the Offline transcription settings, and start. The job appears in the queue with progress, and Cancel Transcription stops it.
The settings control how much of your processor it uses. More finishes sooner and makes everything else feel slower while it runs.
Transcription gets names, places and technical terms wrong, every time. Read every cue before you ship it. A misspelled name is the one error the subject always notices.
Fixing the result
The caption editor sits under the waveform, so you can see the speech you are captioning.
- Add Caption makes a new cue.
- Split Caption breaks one in two, for a cue that runs too long to read.
- Merge with Next joins two, for when transcription broke a sentence in an odd place.
- Import Captions takes an existing SRT, which is how you bring in somebody else's transcript.
Drag the start and end edges on the waveform to fix timing, and use the zoom to get close enough to be accurate.
Line length
Caption lines wrap at 42 characters, on word boundaries, so a long line becomes two readable ones rather than one that runs off the frame.
Two lines on screen at once is the maximum that reads comfortably. Break at natural phrases rather than wherever the count ran out. "The county commissioners voted / three to one on Tuesday" reads. "The county commissioners voted three / to one on Tuesday" does not.
Style
Caption font, Caption font size (pixels), Caption position, Caption margin (pixels), Caption outline (pixels) and Caption box opacity.
An outline or a background box is what makes text readable over moving footage. Without one, the caption disappears the moment the picture behind it goes light.
Thin fonts vanish under compression, because the encoder treats a one pixel stroke as noise and smooths it away. Use a weight heavy enough to survive.
On vertical platforms, keep captions out of the bottom fifth of the frame, which the platform's own interface covers.
Burned in or a separate file
Caption output decides which.
A separate SRT is better wherever the platform takes one. The viewer can turn it off, the platform can translate it, and you can fix a typo in a text editor in ten seconds without re-exporting anything.
Burned in is necessary where the platform will not take a file, and it is what you want when the styling is part of the look. The cost is that it is part of the picture forever.
Caption video format sets the output format when burning in.
A transcript is worth keeping
Once the captions exist, the transcript costs nothing extra. It makes the content searchable and it is the only version that works for a screen reader.