Automatic Captions for Screen Recordings — Free, Offline

The speech-recognition model ships inside the extension. Transcription runs on your machine, works offline, and costs nothing.

One Click, Timed Cues

Record a tab with your voice, open the recording in the editor, and click Transcribe. The extension runs OpenAI’s Whisper model against the recording’s audio and produces caption cues: short lines of text, each with a start and end time. A progress bar tracks the work, and recordings longer than 30 seconds are handled in overlapping chunks.

The result is a caption track, not burned-in text you’re stuck with.

Automatic Is the Draft, Not the Final

A small on-device model will mishear things: product names, jargon, your accent on a bad microphone day. So the editor treats the transcription as a first draft and makes fixing it fast:

• Click any cue in the caption track to select it and jump the video to that moment
• Retype the text in place
• Drag start and end times so the line lands exactly on the speech
• Delete cues you don’t want, or re-transcribe to start over
• Toggle the whole track on or off for the preview and export

Cleaning up the misheard cues beats both raw auto-captions and typing subtitles by hand.

What Renders in the Export

The active cue draws as white text on a dark rounded pill, centred near the bottom of the framed video — standard subtitle styling that reads clearly over any content. Lines wrap automatically instead of running off-frame, and the caption size scales with the video, so text stays legible at any export resolution. When you export, the captions are rendered into the file along with the rest of the polish, locally.

Cloud Captioning vs On-Device, Plainly

That last row is the honest trade. Shipping the model inside the extension means shipping a small one — about 40 MB of quantized weights — and the English-specific variant is what fits and runs well everywhere. If you narrate in another language, this feature isn’t for you yet. If you narrate in English, you get captions that are private by construction rather than by promise.

Privacy

Your voice is the most identifiable thing in a screen recording. Here it is decoded, transcribed, and rendered entirely on your device. No account, no upload, no telemetry — the recording and its captions stay in the browser’s local storage until you export them yourself.

Questions? Reach out at support@superchargebrowser.com.