Caption Family Videos and Voice Messages Locally

Why this works

Captions make family videos usable for relatives who are deaf or hard of hearing, and they make voice notes searchable years later. The transcript stays on your computer.

Step by step

1. Extract the audioUse ffmpeg to pull the audio track: ffmpeg -i video.mp4 -ar 16000 -ac 1 -c:pcm_s16le audio.wav.
2. TranscribeRun whisper.cpp with the model you downloaded. Ask for timestamped output so each line has a start time.
3. Turn the transcript into captionsConvert the timed text to an SRT subtitle file using a plain-text editor or a small script, then check the timings.
4. Burn in or attachLoad the SRT as a sidecar file in most video players, or add it to the video with ffmpeg if you want it always visible.
5. Correct namesFamily names and place names are the most common errors. Fix them before you share.

Copy-paste prompt

Clean up this timed transcript for subtitles: keep each caption under two lines and 42 characters per line, correct obvious mishearings only when the context is clear, mark unclear words as [unclear], and keep the timestamps unchanged.

TRANSCRIPT:
<paste>

Check your result

• Captions are readable at the video's speed
• Family and place names are correct
• Unclear speech is marked rather than guessed

Pitfalls to avoid

• Speech models struggle with overlapping voices and music; expect to edit.
• Share private family recordings only with people you choose.