AutoSub › Guides › Private subtitle generator

Subtitle generator that doesn't upload your video

Most online subtitle tools start by uploading your whole video. For unreleased footage, client work, internal recordings or anything personal, that's a lot to hand over just to get captions. AutoSub is built the other way round: the video stays on your computer and only what's needed for transcription leaves it.

Open AutoSub free30 free minutes when you sign up. Works best in Chrome or Edge.

What happens to your file

A one-hour 4K video can be several gigabytes. The speech-only audio from the same recording is a small fraction of that, which also makes the upload much faster on ordinary connections.

How to use it

1. Open AutoSub. Go to autosub.isroot.in in Chrome or Edge. There is nothing to download, though you can install it as an app from the address bar if you'll use it often. Use an up-to-date Chrome or Edge, since local decoding relies on WebCodecs.
2. Pick the spoken language. Choose the language people speak in the recording, or leave it on auto-detect. Picking it explicitly helps with short clips and heavy accents.
3. Drop in your video. AutoSub reads the file directly from your disk, decodes it in the browser and extracts the audio. Silent parts are dropped and the speech is compressed into small chunks; only those chunks are uploaded for transcription. Sign in when asked; new accounts get 30 free minutes.
4. Review and edit. The subtitles open in the editor. Click any line to jump the video to that moment, fix words, adjust start and end times, split or merge cues, and use find & replace for names the model misheard. Undo and redo cover every change.
5. Export or burn in locally. Export SRT, VTT, ASS or TXT, or burn captions into a new MP4. Rendering happens on your computer, so the finished video is never uploaded either.

How this compares with typical online tools

To be clear about the trade-off: AutoSub is privacy-conscious, not offline. Transcription uses a server-side model because running a large speech model well in a browser on an average laptop is still slow. If your situation forbids sending any audio off the machine, use a local tool instead.

When this matters

• Agencies and freelancers working on client footage under NDA.
• Teams captioning internal training, all-hands or product demos.
• Creators preparing videos before release.
• Anyone with a large file and a slow upload connection.

Pricing starts with 30 free minutes. After that, Pro is $9.99 a month for 1,000 minutes, Pro Annual is $79.99 a year, or a 300-minute pack is $7 one-time. Any purchase also removes the small watermark from burned-in exports.

Frequently asked questions

Does any data leave my device?

Yes, a small amount: compressed audio of the speech in your file is uploaded so it can be transcribed. The video frames, the original file and the finished MP4 never leave your device. The uploaded audio is deleted within 24 hours.

Where is the audio transcribed?

On Cloudflare's infrastructure, using the Whisper large-v3-turbo speech recognition model. See the privacy policy for full details on processing and retention.

Can I edit and export without uploading anything at all?

Yes. If you already have an SRT or VTT file, importing it, editing, styling, burning in and exporting all run in your browser with no upload and no account.

Is it fully offline?

No. Transcription needs an internet connection because it runs on a server. If you need a completely offline workflow, an open-source local tool such as whisper.cpp is a better fit, at the cost of setup and a capable computer.