How to Add Captions to a Screen Recording (Automatic & Private)

The W3C explains that captions provide a synchronized text version of the speech and non-speech audio needed to understand a video. For prerecorded video whose audio carries meaning, captions are required at WCAG Level A.

For customer-facing video — onboarding, demos, tutorials — captions also signal care. A polished recording with clean subtitles reads as documentation; a silent wall of UI reads as an afterthought.

Manual captioning means transcribing the audio, deciding which meaningful sounds to include, and timing each line. It gives you direct control but can be tedious without dedicated software.

Cloud AI captioning can produce a useful first draft, but it requires uploading the recording or audio to a remote service. Check the provider's retention, privacy, pricing, and offline limitations before using it with sensitive footage.

On-device AI captioning runs a speech-recognition model like Whisper locally. You get the speed of AI transcription with none of the upload — which matters when recordings contain unreleased features, customer data, or internal tooling.

The W3C warns that automatically generated captions usually need editing before they meet user needs or accessibility requirements. Always review names, technical terms, punctuation, timing, speaker changes, and meaningful sounds before publishing.

WebVTT is the W3C format designed for time-aligned text tracks connected to HTML video. SRT is also widely accepted, but platform support varies, so check the destination before choosing an export format.

Here is the on-device workflow in Clapio. It works the same on macOS, Windows, and Linux, and it works offline.

Burned-in captions are part of the pixels — they show everywhere the video plays, including chat previews and GIF-style embeds, but viewers can't turn them off and there is exactly one language.

Sidecar SRT or VTT files keep captions as data: platforms like YouTube let viewers toggle them and pick languages, and you can ship several translations with one video file. The practical rule: burn in for Slack, social, and docs; ship SRT/VTT for YouTube, LMS platforms, and help centers — or do both from the same recording.