The short answer
Audio to text
Captions communicate speech and meaningful non-speech sound to viewers who cannot hear the soundtrack clearly.
Visuals to narration
Audio description communicates important visual information to viewers who cannot see the picture clearly.
Subtitles usually translate or transcribe dialogue. They may not include speaker identification and meaningful sound cues. A transcript presents the audio in a separate text form, but it does not automatically make visual action available during playback.
A simple test
Mute the video. If the message disappears, captions are needed. Now listen without looking. If names, actions, graphics, or on-screen instructions disappear, audio description is needed.
What belongs in captions
- Spoken words, including relevant false starts or emphasis.
- Speaker identification when the speaker is not visually obvious.
- Meaningful music and sound, such as “[door slams]” or “[quiet piano music]”.
- Timing that follows the actual speech without appearing before it.
What belongs in audio description
- Essential action and scene changes.
- Intentional on-screen text not spoken aloud.
- Speaker or character identity.
- Visual demonstrations, charts, expressions, and spatial relationships needed to follow the content.
Do not merge the editing jobs
Caption timing follows speech on the original video. AD timing follows safe narration windows and may extend the video with pauses. Editing both in one shared timeline invites errors. Keep separate caption and AD panels, validate each track, then build the publishing files.
Read the full audio description workflow or explore VoxDescribe video accessibility outputs.