Most online video editors work the same way: you upload the whole video, their servers do the work, and you download the result. capshak does it the other way around. The heavy work happens on your own computer or phone, inside the browser tab you already have open.
You don’t need to be an engineer to follow how this works. Here is the plain version.
What “editing in the browser” means
Modern browsers can do more than show web pages. They can decode video, draw frames with the graphics card and encode a new video file, all without any plugin. The technology behind this is called WebCodecs. It gives a web page access to the same video encoders and decoders your device already uses to play videos.
capshak uses it for every heavy step:
- reading and decoding your video,
- drawing the waveform on the timeline,
- the live preview with effects, color and captions,
- encoding the final MP4 (H.264 video, AAC audio).
Because all of that runs on your device, your video file is not uploaded to edit it or to export it. You also don’t wait for an upload before you can start, and you don’t wait in a server queue to get your export.
What is sent, and why
Some jobs need a server, because the AI models that do them don’t run in a browser. For each one, capshak sends as little as possible.
Captions
To turn speech into text, the browser extracts only the soundtrack and compresses it into a small mono audio file at 32 kbps. That is about 240 KB per minute of video, far smaller than the video itself. On older browsers it falls back to a WAV file.
That audio goes to capshak’s server, which forwards it to a speech-to-text provider (AvalAI, using ElevenLabs Scribe, or Groq Whisper). The server does not keep the audio file. What comes back is text with a time for every word.
Translation and proofreading
Translation and AI proofreading send caption text only, never audio or video. The text goes to Anthropic’s Claude, either directly or through AvalAI’s gateway.
The AI editor
If you use the chat-based AI editor, it sends your messages and a description of the edit, such as clips, captions and texts. Only when it decides it needs to “look at a frame” does it send a small JPEG snapshot of the preview, no larger than 768 pixels.
What stays local
Here is the short summary:
| Data | Where it goes |
|---|---|
| Your video file | Stays on your device (unless you turn on backup) |
| Soundtrack, for captions | A small 32 kbps audio file, not kept on the server |
| Caption text, for translation or proofreading | Sent to the AI model |
| Project documents (captions, translations, style, edits) | Saved with your account |
| The exported MP4 | Made on your device and saved there |
The optional backup
There is one good reason to put your video on a server: opening the same project on another device. For that, capshak offers an optional backup. The source video uploads quietly in the background while you work, so you are not blocked waiting for it.
A few rules apply to backups:
- They are deleted automatically after the retention period, 7 days by default.
- Playing a backup uses short-lived signed links, not a permanent public address.
- When a backup is gone, the editor simply asks you for the file again, and your project picks up where it left off.
Your project documents are saved with your account either way, so your cuts, captions and styles are never lost when a backup expires.
The trade-offs
Doing the work on your device is not free of cost. It just moves the cost.
- You need a capable browser. capshak works best in recent Chrome or Edge on a desktop. Recent Safari (including iOS 17 and later) works too. Firefox can edit, but it may not be able to export yet, because export needs the browser’s H.264 encoder; the export dialog tells you if that’s the case.
- Your device does the lifting. Exporting uses your processor and graphics. A newer laptop exports faster than an old phone.
- Keep the tab open while exporting. The export runs in the page, so closing it stops the export.
- Videos are short by design. capshak is built for short videos, up to 10 minutes each.
In return, you get something most editors can’t offer: your raw footage stays with you, and exporting costs nothing. There are no credits for export and no watermark.
Why we built it this way
Sending full videos back and forth is slow and expensive. Keeping them on your device makes editing faster to start, keeps your footage private, and lets us charge only for the AI work that really needs a server. That is why editing, styling, effects and export are free, and credits are only used for things like transcription and translation.
See it for yourself
Read the full picture on our privacy-first page, then create a free account and try an edit. Your video stays right where it is.