Location sound survives
Dialogue separation replaces narration while the ambience, score and field recording stay in place.
For documentary teams
Documentary has a split problem: scripted narration localizes cleanly, and field interviews carry a person’s testimony that should not be casually replaced. Octavia handles both, differently and deliberately.
The specific reasons this content usually stays in one language.
Replace the whole audio track and the location, the weather and the room all disappear with it.
Interviews recorded in wind, traffic and crowds are exactly where automated workflows usually fall apart.
Every territory wants a different combination of dubbed narration, subtitles and accessibility captions.
The parts of the platform that matter for this work, rather than the full feature list.
Dialogue separation replaces narration while the ambience, score and field recording stay in place.
The pattern most documentaries want: contributors keep their real voices, the narration carries the film.
Diarization keeps each interviewee distinct across a feature-length edit.
Searchable text across archive and interview material, which is a research tool as much as a delivery format.
Review transcript and translation before generation, which matters when a contributor’s words carry weight.
Same-language captions and transcripts alongside translated subtitles, per language.
Four steps from the file you already have to every language you need.
Plus stems if the mix has them; separation covers the archive material that does not.
Decide per element what gets dubbed and what stays subtitled.
Contributor speech is checked against source before anything is generated.
Dubbed narration, subtitle tracks and accessibility captions per territory.
One upload in, every format you need to publish out.
Including the ones where the honest answer is a limitation.
Subtitled, in most cases. A synthetic version of a real person’s voice speaking translated words is a meaningful step beyond what most release forms contemplate, and audiences read testimony as more credible in the speaker’s own voice. The common pattern is to dub narration and subtitle the people.
Yes, where the workflow separates dialogue from everything else rather than replacing the full mix. This matters more in documentary than almost any genre, because ambience is doing narrative work rather than sitting behind it.
Output quality is bounded by input quality, and heavy wind, crowd noise or overlapping speech will degrade results in any workflow. Run a two-minute sample through transcription before committing — if the sample is broadly accurate, the film is fine; if it is largely wrong, subtitles from a manual transcript are the realistic route.
Yes, and they are a separate deliverable from translated subtitles. Accessibility captions carry speaker identification and non-speech audio information, and the requirement applies to each language version — a Spanish dub needs Spanish captions too.
Practical guides from the Octavia editorial team.
The same engine, pointed at a different kind of work.
Start with narration in two languages and keep every contributor exactly as they were recorded.