By Robert J. Corn · Founder, AetherCut
You can add subtitles to video free without watermark using client-side tools that process footage in your browser: AetherCut exports clean, watermark free video on every tier with no outbound video bytes, and CapCut, Clipchamp, Kapwing, VEED, and Descript each have free paths that avoid burned-in stamps, though most require sending footage to their servers. This guide shows the exact steps for each, verified frame-by-frame.
Short answer: none of the five mainstream browser editors guarantee a watermark-free free tier across the board, and each imposes different export limits. Descript's free plan, for example, restricts export quality and watermarks video on its lowest tier, per Descript's pricing page. For CapCut, Clipchamp, Kapwing, and VEED, confirm the current watermark and export terms on each vendor's pricing page before you commit footage to it, because these terms change frequently.
Don't trust a marketing page. Export a ten-second clip with subtitles, scrub it frame by frame, and check three things: a corner watermark, a burned in outro card, and whether the subtitle styling was altered on export. Do this on the free tier with your own test footage, never client material.
| Tool | Free-tier watermark | Typical free limits to verify |
|---|---|---|
| CapCut | Historically no watermark on exports | Resolution caps, feature gating |
| Clipchamp | Historically watermarked at 1080p | Export resolution tiers |
| Kapwing | Watermark on free exports | Length caps per export |
| VEED.io | Watermark on free exports | Length and resolution caps |
| Descript | Watermark on free video export | Export quality tiers |
AetherCut applies no watermark on any tier, and you can verify the client side workflow yourself in DevTools: see free online video editor with no watermark.
Open AetherCut in Chrome or Edge and drop your file onto the timeline. Playback and decoding run through WebCodecs in the browser, so the footage stays on your disk. You can confirm this yourself: open DevTools, switch to the Network tab, and watch for outbound traffic while you scrub the timeline. No video bytes leave the machine.
Click Generate Captions on the toolbar. AetherCut runs a Whisper model client side, producing a timestamped subtitle track in a minute or two for a typical deposition length clip. The transcription details are covered on the auto-captions page.
Click any caption line to edit the text directly on the timeline, which matters when Whisper mishears a name or a term of art. Adjust font, size, colour, and screen position on the subtitle track. Legal and medical users will want to review every line manually before export.
Export burned in subtitles or an SRT file at up to 4K, with no watermark on any tier. The render also runs client side, so the Network tab stays empty end to end.
CapCut's free tier includes auto-captions, but the tool runs in your browser with footage processed on its infrastructure, and some caption styles sit behind the paid plan. For regulated footage, that processing path is the disqualifier, not the watermark.
Clipchamp, owned by Microsoft, offers auto-captions on the free tier with no watermark on exports. Clips are processed on Microsoft's servers, so the same confidentiality concern applies.
Kapwing's free tier adds subtitles and auto-captions, but stamps a watermark on exports unless you upgrade. Its collaboration model depends on shared server state, so your source clips live in its cloud.
VEED's free tier auto-generates subtitles and exports with a watermark. Footage is processed in its cloud, which rules out deposition or patient video.
Descript is the strongest free-tier subtitle workflow of the five: transcribe, edit text, export SRT or burn in, and the free plan carries no watermark. Its help center documents the auto-caption workflow in detail (Descript support docs). But transcription happens on Descript's servers, so privileged footage still leaves your machine. If that's a dealbreaker, the pattern to look for is local processing: AetherCut handles captions through the same no-upload pipeline described at editing video without uploading, where you can watch the Network tab and confirm zero outbound video bytes. The trade-off is honest: fewer cloud conveniences, and you supply the patience for local 4K encoding.
If your footage contains patient identifiers, deposition content, or source material, handing it to a third party can breach HIPAA, attorney-client privilege, or an NDA the moment the transfer happens. It does not matter whether the tool deletes files afterward; the exposure already occurred. That is why CapCut, Clipchamp, Kapwing, VEED, and Descript are non starters for regulated work: each requires sending footage off your machine to process it.
A verifiable claim beats a promised one. Open the Network tab, load your clip, and watch the outbound traffic. If no video bytes leave the browser, processing is client side and the footage never crosses your network boundary. AetherCut is built so this test passes by default, running WebCodecs and WebGPU locally with no server in the path (https://aethercut.app/offline-browser-video-editor).
Burn in subtitles when the platform strips sidecar files or your audience watches muted. Export an SRT when you want editable captions, accessibility compliance, or platform managed caption rendering. AetherCut supports both, and because decoding runs client-side through the File System Access API and WebCodecs (W3C), neither export path sends footage anywhere.
YouTube accepts SRT uploads and renders captions natively, so a sidecar file is usually the better choice there. TikTok and Instagram favor burned-in text: both apps often discard sidecar captions, and muted autoplay means visible subtitles carry the message. One caveat worth quoting: AetherCut's burned-in export renders through the browser's canvas, so complex styling is simpler than an SRT's plain-text format but slower on long 4K timelines. For offline or restricted-network workflows, see AetherCut's no internet editing guide.
Free tiers usually cap export length, not editing length, so the fix is trimming before you caption. Cut your footage down to the segment you actually need first, then run captions on that clip only. In AetherCut, trimming happens locally in the browser, so you can slice a 40 minute deposition into a 3 minute highlight without moving a byte off your machine. Fewer minutes to caption also means faster transcription.
Set in and out points on the timeline, duplicate the clip if you need multiple excerpts, and delete the dead footage. Whisper runs on the trimmed clip, so a 3 minute excerpt transcribes in seconds instead of minutes.
AetherCut has no watermark on any tier, but 4K export in a browser depends on your hardware: WebCodecs and WebGPU do the encoding locally, so a laptop with a discrete GPU handles 4K comfortably while an older machine may stall. If 4K stalls, export at 1080p, which is fine for most subtitle deliverables.
Whisper-based transcription on clean, single-speaker audio typically lands in the mid-90s percent word-accuracy range, per OpenAI's Whisper documentation. That is good enough for YouTube subtitles, not for a deposition.
Legal and medical speech breaks auto-captions in predictable places: party names, drug names, Latin phrases, and overlapping speakers. Whisper has no way to know "Doe" is "Dough" or that "MI" means myocardial infarction, so expect to correct every proper noun.
Practical workflow: generate captions automatically, then read the SRT against the audio at 1.5x speed. Twenty minutes of footage usually needs ten minutes of editing. For anything going into a court filing or patient record, manual review is not optional, it is the compliance step.
AetherCut keeps every byte client side, but that constraint costs you. Rendering happens through WebCodecs and WebGPU in your browser, so a 4K export on a five year old laptop will be slower than Premiere Pro on the same machine with a discrete GPU. There is no render farm behind the tab to bail you out.
Collaboration is the bigger gap. Shared projects need shared state, and shared state needs a server, which is exactly what AetherCut refuses to have. Two editors cannot work on one timeline the way they can in Descript or VEED. Teams that need coediting on confidential footage should use an SRT handoff workflow instead: one person exports subtitles, the other imports them.
Finally, no mobile app and no offline install. If your machine loses the browser, you lose the editor.
Historically, yes: CapCut exports have carried no watermark on the free tier, though resolution caps and gated caption styles apply. The terms change frequently, so run the frame by frame test on a ten second clip before committing real footage. The bigger issue for regulated work is that CapCut processes footage on its servers regardless of watermarking.
Yes. AetherCut runs decoding, Whisper transcription, and rendering entirely in the browser through WebCodecs and WebGPU, so no video bytes leave your machine. Verify it yourself: open DevTools, watch the Network tab while scrubbing and exporting, and confirm zero outbound traffic. The five mainstream alternatives all require sending footage to their servers.
Most do. Descript's free plan supports SRT export, and AetherCut offers both SRT and burned-in output at up to 4K with no watermark. Choose SRT for YouTube or when you need editable, accessible captions; burn text in for TikTok and Instagram, which often discard sidecar files and rely on visible subtitles for muted autoplay.
It depends on the tool. Kapwing and VEED cap export length on free plans, while AetherCut imposes no length cap because rendering happens locally on your hardware. The practical fix for capped tools is trimming before you caption: cut to the segment you need, then run transcription on that clip only. Shorter clips also transcribe faster.
No. Whisper based transcription depends entirely on the audio stream, so a silent clip produces nothing to transcribe. Your options are burning in text manually as styled captions on the timeline, or generating an SRT by hand. If the clip has audio but Whisper mishears terminology, expect to correct every proper noun before export.
Thanks for rating!