Oral history subtitles that respect the original voice.

Keep access conditions beside the recording

Oral history is more than an audio file. Begin with the permissions and access conditions agreed for the recording, including any limits on sharing or translation. The Oral History Association’s principles emphasize respect for narrators and ongoing participation. Use the project’s own review process to decide what can become a publicly accessible subtitle track.

Keep an unchanged source recording and work from a clearly identified copy. Polyglot processes media and subtitle text locally after model setup, but your own backups and sharing choices still determine who can access them. Save the project in an appropriate location and avoid putting personal details into filenames when a project identifier is enough.

Find your way in Polyglot

Keep each permitted recording in a separate project. Files brings its original and translations together without changing your external permission and archival records. Review uncertain names against the attached recording before exporting.

Preserve unfamiliar words rather than normalizing them

Create the original-language draft and compare it carefully with playback. Old place names, family names, dialect and historical vocabulary are easy places for recognition to produce a more familiar but incorrect word. Use verified project notes as a reference. Do not rewrite a speaker’s phrasing simply because the generated alternative sounds more polished.

For example, a narrator may use an older street name that differs from a current map. Record the spoken name faithfully after checking it; explain contextual information in the archive’s accompanying description if appropriate. If a passage remains unclear, follow the project’s transcription convention and keep a review note rather than presenting an invented reconstruction as certain.

Decide what the subtitle track is for

A timed subtitle file supports listening and viewing. It is not automatically a complete research transcript with speaker metadata, editorial annotations and archival catalog fields. Polyglot exports SRT and does not automatically identify speakers. Plan any additional transcript or metadata deliverable separately so the archive receives what it actually requires.

Speech recognition supports 25 European languages; dialect or multilingual material may need substantial human review, and translation options do not imply broader recognition support. When preparing another language, involve a reviewer who can assess the narrator’s meaning and the agreed access conditions. Keep uncertainty and nuance visible in the translation instead of smoothing them away.

Package files for the archive’s actual system

Ask the receiving archive which subtitle format and naming convention it accepts. Polyglot can export SRT or WebVTT from the reviewed track. Do not rename an extension and assume the contents are converted: choose the format in Export. Other formats may need a separate tool. Keep the source export so subsequent conversions can be checked against a known reviewed version.

Use a stable identifier such as OH-014-edit02.en.srt and record the source recording, reviewer and language in the handoff notes. Test the track with its matching media, including a quiet passage and a section near the end. Review needed speaker context and meaningful non-speech information separately. Preserve the original, the working project and the approved subtitle export according to the archive’s own retention and access process.

Keep exploring.