How to add captions to a screen recording without uploading it

To add captions to a screen recording without uploading it, make them on your own computer. Focal, a screen recorder for Windows, does this with a speech model that runs on your computer: record with your microphone on, make captions in the editor, correct any line that came out wrong, pick a look, and export them drawn into the picture, as a track viewers can switch on, or as an .srt file. The model is downloaded once, about 190 MB, and your voice and your screen stay where they were recorded.

My name is Vince. I make Focal, and between demos and tutorials I talk to my monitor more than to most people I know. This is how I get all those words onto the video without sending them anywhere first.

Why caption a screen recording

Most social feeds start a video muted, and many people never turn the sound on. A demo without captions spends its first seconds as a silent film about a settings page. With captions, those same seconds can be read.

Captions are also how deaf and hard-of-hearing viewers follow you at all. They help anyone watching in a second language, anyone who finds your accent hard to follow, and anyone in an open office without headphones. And captions are text. Text can be searched where a voice cannot, by a site and by you, months later, looking for the one video where you explained a setting.

Screen recordings have one more reason of their own. They are full of words that are hard to hear and easy to read: function names, menu items, product names. Captions put those words on screen at the moment you say them, which is the moment the viewer needs them.

The usual ways, and what they send where

There are three common routes. Each one works, and each one has a catch.

A captioning service

You upload the video, or at least its audio, and get captions back. Some services use people, some run a speech model on their own servers. If you need captions checked by a person for a legal or broadcast requirement, a human service is the right call and worth the upload. The catch is that your voice, and usually your screen, now sit on someone else’s servers under someone else’s retention policy, a genre of document I have personally never read to the end. For a public launch video that may be fine. For a recording of an internal dashboard, a customer’s account or code that has not shipped, it often is not.

YouTube’s automatic captions

Upload the video to YouTube and it can caption it for you at no cost. If the video is going to YouTube anyway, this is a fine route and I would not talk you out of it. The catch is that the upload is the price, and the captions live in YouTube’s player. If the video is headed for a help center, a pull request, a sales email or a social feed, you have put it on a platform it was never meant for just to get some text back.

Typing them by hand

A subtitle editor lets you type each line and set when it starts and ends. Nothing leaves your computer, which is the good news. The rest is a loop: play a few seconds, pause, type, set the times, and go again, for the whole video. It takes many times longer than the recording itself. I tried it once, and by the end I knew my own voice better than I ever wanted to. That loop is why so many screen recordings go out with no captions at all.

I wrote more about why the upload matters in your screen recording should stay on your computer.

How to add captions in Focal

Focal takes a fourth route. The speech model runs on your computer, so captioning needs no upload and no account. The model is downloaded the first time you make captions, and after that nothing about captioning leaves the machine. The privacy note lists the few things Focal ever sends anywhere. It is a short list, and your audio is not on it. The captions section of the home page shows the result. Here is how it goes.

1. Record with your microphone on. Press the recording shortcut, Print Screen by default, then drag a region, click a window or press Enter for the whole screen. In the small panel that opens next, turn the microphone on. Focal keeps that choice for next time.
2. Talk as you work, then stop. The editor opens with zooms already made.
3. Make captions. The first time, Focal downloads the speech model, about 190 MB. The captions come a phrase at a time, following your voice.
4. Read them through and correct any line that came out wrong, right there in the editor.
5. Pick a look: Clean, Subtitle, Lines, Shorts, Karaoke or Minimal. Or set your own font, weight, case and colors, with a box behind the text, a box per line, an outline, a shadow or nothing. Captions can dim the words ahead, or color or box the word being spoken, so they highlight word by word. They can show a few words at a time on one or two lines, sit at the top, middle or bottom, and fade or pop in.
6. Choose how they leave the editor: drawn into the picture, as a track viewers can switch on, or both. Save an .srt file for any other player or site. Then export the MP4.

Which to choose? Draw them into the picture for social feeds, where a clip gets downloaded and reposted and a separate track is easily lost along the way. Use a track for a player with a captions button, so people who do not need them can switch them off. Save the .srt when the video goes to a site that takes its own caption files.

Once you like a look, save it in a style along with your background and cursor, and set that style as the default. Every new recording then gets the same captions, and you no longer have to remember which shade of yellow you settled on last month.

How long captions take

On a computer with a dedicated GPU, captioning is quick. Measured on 2026-10-31 on my test machine, an RTX 3090 with a Ryzen 9 9950X3D, ten minutes of speech was captioned in about 14 seconds, which is shorter than some of my opening sentences. On the CPU alone, without the GPU, the same ten minutes took about 38 seconds. That is a fast desktop CPU, so expect longer on a smaller one. Other machines differ, and the first run adds the model download.

Getting names and technical terms right

General speech models do well with everyday words and badly with yours. A product name, a library, a function called useEffect: these are the words a screen recording is about, and they are exactly the words a model tends to hear as something else. Focal reads the words on your screen and gives them to the model as hints, so a term that is visible, like useEffect or the name of the product you are showing, comes out spelled the way the screen spells it.

Focal also leaves out the phrases speech models make up over typing and clapping. Give a model a few seconds of rattling keys and it can invent a sentence nobody said, with the calm confidence of someone who was not listening. Those sentences do not end up in your captions.

If a word still comes out wrong, correct that line in the editor before you export.

Captions in other languages

Focal captions 26 languages, detected from your speech or chosen by you. It can also write speech in any language out in English, which helps when you explain things best in your own language and your viewers read English. Japanese and Chinese are read without spaces, the way they are written. Languages other than English use a second model of the same size, about 190 MB, downloaded the first time you need it.

Edit the video by its transcript

Once a recording has captions, the captions list is also a way to edit. Select words in it and cut them from the video. Cut every “um” and “uh” in one click, a click that does a lot of work on my own recordings. Tighten long pauses, while the pauses where you click or type stay, so the video still shows what you did.

This is the quickest way I know to turn a first take into something worth sending: talk through the demo once, then cut the false starts by reading them. The guide to recording a product demo video on Windows covers the rest of that workflow.

Captions for vertical video

A vertical cut for Shorts, Reels or TikTok needs captions more than any other version, because it usually starts muted on a phone. Change the aspect ratio to 9:16 and Focal re-aims the zooms and the captions for it. Start from the Shorts look, and set the captions in the middle rather than at the bottom, where those apps put their own buttons and text, and they got there first. Then export with the preset for Shorts and Reels. The guide to turning a screen recording into a vertical video goes through the rest.

Captions are in the free trial with every other feature, with no time limit, no card and no account. Exporting a video needs a license, and until you enter a key the editor’s preview shows a “Made with Focal” watermark. Download Focal, record two minutes with your microphone on, and make captions to see how close the first pass comes.