Client & confidential work
Subtitle footage you are not allowed to upload.
Unreleased product footage, client rushes, embargoed interviews, internal recordings. The file stays where it is.
LumaCaption subtitles confidential footage without uploading the video file — only the audio is used to work out the words and their timings, so unreleased, embargoed and NDA-covered video never leaves the device it is on.
Every mainstream captioning tool works the same way: you upload the video to their servers, it is processed there, and you download the result. For most videos that is unremarkable. For some it is a policy violation, a contract breach, or an unannounced product sitting on somebody else’s infrastructure.
Agencies hit this constantly — a client’s footage under an NDA that names permitted processors, and a captioning tool that is not one of them. So do in-house teams with an unlaunched product, journalists with an embargoed interview, and anyone whose recording contains something that has not been said publicly yet. The usual answer is to caption it by hand in an editor and lose the afternoon.
The video file is never uploaded
Only the audio is used to work out the words and their timings. The video itself stays where you put it.
Nothing queues on a server
There is no processing queue holding your footage, no rendered copy sitting in someone’s bucket waiting on a retention policy.
Exports are unmetered
Re-render as many versions as review demands. Only transcription minutes are charged, so a fourth round of client notes costs nothing.
Restrained styles included
The Clean lane’s 37 styles and the Cards lane’s 19 lower thirds are built for work that has to look like a deliverable.
What "never uploaded" does and does not cover
Being precise about this is more useful than being reassuring about it, because the distinction is what you would have to explain to a client or a security reviewer.
The video file is not uploaded. That is the claim, and it is the one that matters for footage: nobody else receives your frames, your faces, your unreleased screens, your unlaunched packaging.
The words have to come from somewhere, and that is the audio. Candy Lite handles transcription without your audio leaving; Candy Ultra is the premium engine and is where the accuracy on hard audio comes from. If your confidentiality obligation extends to the spoken content and not just the picture, that is the distinction to check against your own policy before you start — and it is a question worth asking of any tool, not only this one.
- Your footage
- Never uploaded, on any plan. No frames, no faces, no on-screen content leaves the device.
- Your renders
- Produced where you are, downloaded directly. There is no rendered copy on a server to expire.
- Your transcript
- Editable and downloadable as SRT, VTT or plain text. What you correct is what you keep.
- Worth checking yourself
- If your obligation covers the spoken content and not just the picture, choose the engine accordingly. Any honest tool should let you ask this question and get a straight answer.
The agency workflow this is actually for
The shape of the problem in an agency is not one video, it is a round of revisions. A client sends rushes, you cut, you caption, they send notes, you re-caption, they send more notes. If captioning costs money or minutes per render, that loop is where the budget goes.
Here transcription is the metered thing and rendering is not. One pass over the audio, then as many styled exports as the review takes — different styles for different platforms, a version with the logo colour corrected, a re-cut after the legal read — none of which draws down minutes on any plan.
The transcript is also where client-specific vocabulary gets fixed once. Product names, executives’ surnames, internal acronyms: correct them in the transcript and every clip cut from that recording inherits the correction rather than repeating the error twelve times.
How it works
Open the footage
Client rushes, an unreleased cut, an internal recording. The file does not leave the machine it is on.
Fix the vocabulary once
Names, products, acronyms. Every clip cut from this recording inherits the corrections.
Export as many versions as review needs
Burned-in in any of 240+ styles, plus SRT or VTT for a client who wants the file. Exports are free and unlimited.
Common questions
Is my video uploaded to a server?
No. The video file is never uploaded on any plan — only the audio is used to work out the words and their timings. There is no queue holding your footage and no rendered copy sitting on someone else’s storage waiting for a retention policy to delete it.
Can I use this for NDA-covered client work?
That depends on the NDA, and the honest answer is to read it against what actually happens: the video file does not leave your device, and the transcription engine you choose determines how the audio is handled. If the obligation covers the picture, this clears a bar most tools cannot. If it also covers the spoken content, pick the engine accordingly — and ask the same question of every other tool in the workflow.
Does re-rendering after client notes cost extra?
No. Transcription minutes are the only metered thing; exports are free and unlimited on every plan including free. A fourth round of revisions costs the same as the first: nothing.
What styles suit a client deliverable?
The Clean lane — 37 restrained styles built on editorial typography with contrast measured against the actual footage — and the Cards lane, whose 19 lower thirds can be pinned to a client’s exact palette and reused across a campaign.
Is there a watermark on the output?
No, on any plan, including free. Nothing is added to your export.
