Blog · Which tool for which job
How FirstCut edits your video without uploading it
By Eran Strassburg · Updated 2026-10-10
The short answer
FirstCut never uploads your video. Your browser reads the file, extracts only the audio track, and uploads that for transcription, about 115 MB for two hours of typical AAC audio. The cut is a list of timecodes, and playback in the browser comes from the file on your own disk. The exported timeline points at the same originals.
What leaves your machine and what does not
| Uploaded? | Notes | |
|---|---|---|
| The video picture | No | It is read locally and never sent |
| The audio track | Yes | Extracted in the browser, uploaded for transcription |
| The transcript | Created on the server | Made from the audio and stored with your account |
| The cut | Stored as data | Beats, clips and timecodes, with no media in it |
Why a cut does not need the picture
FirstCut works from what is said. A clip is a source file plus an in point and an out point, chosen from the transcript. Nothing in that needs the pixels, so nothing in it requires the video to be anywhere but on your disk.
The cost of this design is stated plainly elsewhere on the site: FirstCut does not look at the picture, so it cannot choose B-roll or judge a shot.
How playback works from local files
To play a cut, the browser opens the original video on your disk and jumps to each clip's timecodes in order. A file you just imported is linked already. If you come back later, or open the sequence on another computer, you point FirstCut at the file or its folder again.
On Chromium browsers FirstCut can remember where the file is and restore the link with one click. Other browsers ask you to select the file each session.
What the transcription and AI providers receive
FirstCut is built on third-party providers. The speech-to-text provider receives the audio track and returns a transcript. The AI language model receives transcript text to assemble a cut, and never receives audio or video. The privacy page has the full description.
What this means for large camera files
Upload size depends on the length of the audio, not on the size of the video. A multi-gigabyte camera file uploads as its audio track only. Uncompressed camera audio is re-encoded to compact Opus in the browser first, so the upload stays small.
Because the picture is never copied, the exported timeline links to your originals at full quality.
Limits
- Some audio formats cannot be read in the browser and are refused when you pick the file: FLAC, ALAC, AC-3 and DTS, and files with no audio track.
- Each file can be up to 2 hours long.
- The browser's player is a preview. A file whose video codec the browser cannot decode will not play there, even though its audio imported.
- If you move or rename the video after importing, you will need to link it again.
The help page covers each of these.
Questions
Is the audio stored?
Yes. The extracted audio and the transcript are stored with your account so you can re-cut the footage. Re-cutting an imported file is free for three months from import.
Why did my file not import?
Its audio is probably in a format the browser cannot read, such as FLAC, ALAC, AC-3 or DTS, or it has no audio track. The help page lists what works.
Does the AI see my video?
No. The language model receives transcript text only.
Related guides
- Pre-editor or cloud video editor: export a timeline or render in the browser
- AI for documentary rough cuts: what it can and cannot do
- EDL, FCP7-XML, FCPXML and OTIO: which timeline format to use
- Who operates FirstCut
Try it on your own footage
FirstCut is free to try with no install and no credit card. Start a rough cut → · Or open the live demo, which needs no account.