Add subtitles to your video
The speech becomes subtitles you can correct, style and burn into the video. Free, and nothing is uploaded.
mp4, mov, m4v, webm, mkv, avi. Up to 60 minutes and 2 GB.
Open source. It writes your subtitles on your own computer.
- Only the first time on this device
- Nothing to install
- Your video is never uploaded
Premium will write your subtitles in the cloud: much faster, more accurate, and nothing to download.
CutNoodle's subtitle tool turns the speech in a video into subtitles, lets you correct the text and the timing, style them, and exports the video with the subtitles burned in. It runs entirely inside your browser tab: the video is never uploaded, and it is free with no account.
Before you start: one download per device
The transcription is done by a speech model, Whisper, running on your own computer. That is why the tool is free and why nothing is uploaded, and it has one cost you should know about first: the first time you use the tool on a device, your browser downloads the model. It happens once. The model is then kept in the browser, and from the second video on the tool starts right away, even with no connection.
| Model | Download when the browser can use the graphics card | Download when it cannot |
|---|---|---|
| Tiny | 114 MB | 39 MB |
| Base | 197 MB | 73 MB |
| Small | 559 MB | 238 MB |
- Each model is a separate download. If you used Tiny and then pick Base, Base is downloaded too.
- It is per browser, not per account. Another computer, a phone, another browser on the same computer or a private window each download it again.
- Clearing the site's data removes it. So can the browser itself when the disk is nearly full. The next run downloads it again.
- On a mobile connection it uses your data. If that matters, start with Tiny or wait for Wi-Fi.
The exact size for your browser is written on the tool before you start. If you import an .srt or .vtt file instead of transcribing, nothing is downloaded at all.
How it works
- Choose a video. MP4, MOV, M4V, WebM, MKV or AVI, up to 60 minutes and 2 GB. It is opened from your disk, not uploaded.
- Create the subtitles. Pick a model and start. The audio is read, the model writes the words with the moment each one is spoken, and the words are grouped into short lines.
- Correct them. Every subtitle is a line you can retype. On the timeline you drag a block to move it and drag its edges to change when it starts and ends. Split, join, add and delete are one click, and everything can be undone.
- Style them. Eight styles, ten fonts, your own colours, outline, background, size, position and animation. The preview shows exactly what the export will contain.
- Export. The video is encoded again in the browser with the subtitles drawn on every frame, and you download an MP4.
What you can change
- Words per subtitle. Two, four or six. Two suits fast vertical clips; six reads like a normal caption line. Changing it rebuilds every subtitle, and one undo brings the old ones back.
- Style. Clean, Social, Minimal, Word Pop, Karaoke Pill, Yellow Lift, Headline and Kinetic Color. Some show a whole line, others light up or bring in one word at a time as it is spoken.
- Animation. Keep the motion the style comes with, or choose None, Fade, Pop or Rise for any style.
- Text. Ten fonts, text colour, the colour of the spoken word where the style has one, a thin or thick outline, capital letters, and a background box with its own colour and opacity.
- Size and position. Drag the subtitle on the video, or use the sliders. It stays inside the frame.
Work in progress is saved in the browser without the video. Close the tab, come back, choose the same file again, and the subtitles and the style return as you left them.
How long it takes
Transcribing in a browser is slower than a cloud service, and that is the trade for keeping the file on your machine. In our measurements with Chrome, on a computer whose graphics card the browser can use, the Base model needed between a fifth and two fifths of the length of the video. Tiny is somewhat faster and less accurate; Small is the most accurate, the slowest, and by far the largest download.
Without graphics acceleration the work falls on the processor and takes considerably longer. The tab also slows down when it is in the background, so keep it open and in front. The export is a second wait: roughly as long as the video, often less.
What it does not do
- It uploads nothing, so there is nothing to delete afterwards.
- It does not let you choose the language, and it does not translate.
- It does not tell speakers apart.
- It uses one style, one animation and one position for the whole video. You cannot style a single subtitle differently.
- An imported .srt or .vtt file carries the timing of each line, not of each word, so the word by word styles spread the words evenly across the line.
- It handles one video at a time, up to 60 minutes and 2 GB. No batch mode.
- It writes MP4 with H.264 and AAC only. The export uses the browser's own video encoder, which current Chrome and Edge have; a browser without it says so instead of exporting.
- It does not cut or trim. To take the pauses out first, use the silence remover and add subtitles to the result.
For the choices behind readable subtitles, such as line length, timing and placement, see how to add subtitles to a video.
Common questions
Do I have to upload my video to add subtitles?
No. CutNoodle opens the file from your disk and does everything inside the browser tab: the transcription, the preview and the export. Neither the video nor its audio is sent anywhere, and there is no account.
Why does it download something the first time?
The speech is turned into text by a speech model that runs on your own computer, and your browser has to fetch that model once. It is between about 40 and 560 MB depending on the model you choose and on your browser. It is then kept in the browser, so the next video starts right away and works offline. A different device, a different browser or a private window downloads it again.
How long does it take to create the subtitles?
In our measurements with Chrome, on a computer whose graphics card the browser can use, the Base model needed between a fifth and two fifths of the video's length, so a 15 minute video took a few minutes. Without graphics acceleration the work falls on the processor and takes considerably longer. Keep the tab open and in front while it runs.
Which languages does it understand?
The speech models are multilingual and recognize the language on their own. You cannot pick the language, and there is no translation: the subtitles are in the language that is spoken. Accuracy drops with strong accents, background noise, music and people talking over each other, which is why every line can be corrected before you export.
Can I use subtitles I already have?
Yes. Choose the video, then import an .srt or .vtt file instead of transcribing. Nothing is downloaded in that case. You can then restyle the subtitles and burn them into the video like any others.
Are the subtitles part of the picture or a separate track?
Part of the picture. They are drawn onto every frame, so they show on any platform and cannot be switched off by the viewer. If you need a track that can be switched on and off, download the .srt or .vtt file from the Export tab and upload it next to your original video.
What file do I get back?
An MP4 with H.264 video and AAC audio, at the resolution of your original, named after it with a -subtitled suffix. From the same tab you can also download the subtitles alone as .srt or .vtt.

