Transcribe the source
FFmpeg prepares the audio and multilingual Whisper turns the original speech into timed subtitles.
Case study · 02AI-assisted localisation
Russian
Dubbing
A Finder-based workflow that transcribes source speech, adapts it into concise Russian and assembles a timed voice track into the final video.
Original speech to timed Russian dub
00:48
Problem → solution
Translate meaning.
Respect the timing.
A literal Russian translation often becomes too long for the original speech slot. The pipeline adapts language, timing and generated voice as one coordinated process.
Adapt for the slot
Russian lines preserve meaning, use a controlled glossary and shorten selectively when timing demands it.
Fit and finish
Voice segments are timed, joined, mixed, normalised and attached to the final video.
Technical notes Open details
Validation & reruns
Subtitle structure is validated, overlong cues receive targeted retries, and per-line audio caching avoids rebuilding finished segments.
Responsible use
Intended for material I own or have permission to adapt. The Russian track uses a synthetic AI voice and is presented as such.