Caption the mix your viewers hear in an OBS recording

Give the editor three moments to listen for

Imagine a recorded demo of a map-making app. At 00:08 the presenter says “Save the route.” At 00:19 the app speaks the confirmation “Route saved.” At 00:32 a producer gives a direction on a separate talkback microphone while the presenter silently adjusts the map. The viewer should hear both of the first two lines. At the third moment, neither the direction nor a caption for it belongs in the release.

In OBS, the editor can assign a source to several recording tracks in Advanced Audio Properties and choose which tracks to record in Output settings. The OBS guide illustrates a full mix on track 1; that is an example, not a guarantee about the file you received. Have the recording owner identify the intended sources and audition the tracks in an editor. If the public mix already includes the talkback, selecting it by number will not remove the producer.

Export a reference with only the approved sound

Use an owned demo or get permission to process the voices, screen content and app audio included in the release. Ask the editor for a review export if you are not cleared to receive the isolated microphones. A raw recording that happens to start on the public mix can still carry other selectable audio.

In the editor, solo the approved mix or build it from the cleared sources. Do not add a presenter’s isolated microphone on top of a mix that already contains it. Export the entire locked picture with one audio track in a supported format such as an H.264/AAC MP4. Keep the same first frame, pauses and ending as the delivery. Changing the container alone does not establish which sources are audible.

Play the export. If “Route saved” has disappeared, the presenter-only track sounded plausible but missed the app. If you hear the direction, send the reference back for a new mix. Removing words from an SRT cannot remove a voice from a movie. Keep the original recording with the authorized editor.

Review the two voices against the picture

Open the approved reference in Polyglot on an Apple Silicon Mac. After model setup, transcribe locally, or open an existing SRT with the same picture for review. Listen through the save action: “Save the route” is an instruction and “Route saved” is confirmation. A repeated instruction in both cues changes what a viewer thinks happened. If the voices are hard to distinguish without sound, identify the app in the wording where helpful.

Review the whole cut before exporting the SRT. If the publisher requests another language, keep the approved source and check the distinction between instruction and confirmation in the translation. Name the files for the edit and language, for example RouteDemo-v4-en.srt. The three listening points are spot checks; they cannot clear the rest of a recording.

Test the exact files that will leave the edit room

Open the delivery movie with captions switched off. Confirm the two intended voices are audible, the private direction is absent and no unapproved audio track is available to select. Inspect the delivered file’s audio streams in a media inspector as well: a player menu can hide a track. If the file contains another audio stream, identify it before release, even if playback starts on the correct one.

Load the delivered SRT. With sound muted, watch the save action and its confirmation, then the silent adjustment. The two distinct lines should appear at their respective moments; the private direction should never appear. Check a late line for sync, too. If speech was removed by trimming the picture rather than remixing the audio, request the revised cut and review the downstream cue times before sending the files.

Keep exploring.