01
How much work is left for you?
Reading, mostly. The typing step is gone, and what remains is a pass through the text to correct names, product terms or anything spelled unusually. That is a minute of work on a clip rather than the twenty it used to take, and it happens before the clip is produced rather than after.
02
Three ways words get onto a video
Most viewers meet a clip with the sound off, which is why the first row is the only one that reliably works for short video.
| Method | Survives resharing? | Effort |
|---|---|---|
| Drawn into the picture | Yes, always | None here, it is what happens |
| A separate track uploaded alongside | Often lost | Moderate |
| The platform's own automatic captions | Shown only there | None, and no control |
| Typed by hand in an editor | Yes | High, per clip |
| Nothing at all | Not applicable | Loses the silent majority |
03
What to check during the read-through
Names come first. People's names, brand names and place names are where automatic text is most likely to guess, and a misspelled guest name is the error viewers notice and comment on. Scan for every capital letter in the text and ask whether it is spelled the way the person spells it.
Next, numbers and units. "Fifteen" and "fifty" sound alike in fast speech, and a wrong figure in a clip about money or health is worse than a typo. Finally, look for words that make sense in isolation but not in the sentence: a homophone that slipped through will usually read as nonsense when you take the line as a whole.
Finally, read the text once without the video. If a line makes sense only when you hear the speaker's tone, add the missing word or punctuation so it reads correctly on its own, because most people will read it on a muted phone.
04
Captions for people who cannot hear the clip
Captions exist first for viewers who are deaf or hard of hearing, and a few habits help them most. Keep the words in sync with the speech rather than running ahead. Do not trim what was said into a shorter paraphrase: someone who can partly lip-read will notice the mismatch.
If two people are talking, make it clear who is speaking, either by keeping each person's lines distinct or by cutting the clip so only one voice runs at a time. Sounds that carry meaning, a laugh that shows a line was a joke, a crowd reaction, belong in the picture too if they change how the moment is understood.
05
How much text a viewer can take in
Comfortable reading for most adults on a phone sits around three words per second. Fast speakers go well past that, and when every word they say is on screen at once, viewers fall behind and stop reading. Short lines that change with the speech keep the eye with the voice.
Place the words where they do not cover the thing people came to see. On a talking head that usually means below the chin; on a screen recording or a product shot it may mean higher. The best position is the one where a viewer can watch the face and read at the same time without moving their eyes very far.
06
Skip the typing
- Upload the video. Any common file, or one of the supported addresses.
- Read through the text. It arrives matched to the speech, ready to check.
- Correct and produce. Fix names and terms, then let it render.
Upload one video and read what comes back before deciding.
Questions people ask
- Do I type anything?
- Only corrections, and only if you want them.
- Can I change how it looks?
- Nine styles, twenty-one appearance effects, and position and size are yours.
- Do the words end up in the file?
- Yes, they are part of the picture.
- Does the video need captions already?
- No, that is the point of this.
- Which shapes can I export?
- 9:16, 4:5, 1:1 and 16:9, all on every plan.