[Feature Request] Support for Multi-Track Audio Transcription in MP4 Containers

[Graham42]

First off, thank you for building this awesome tool with cross platform support!

Currently, when importing a video file (MP4/MKV) with multiple discrete audio tracks (e.g., separate tracks for different speakers in a podcast or gameplay recording), Buzz appears to only transcribe the first audio stream.

It would be great to have a way to select which audio track(s) to transcribe from a multi-track file, with options for:

• Track Selection: Upon importing a file with multiple audio streams, provide a dropdown or checklist to select specific tracks.
• Batch Processing: The ability to transcribe multiple selected tracks from the same file simultaneously (producing separate SRT/TXT files for each).
• Automatic Labeling: If multiple tracks are selected, automatically use the track name or index as a prefix for the exported filenames (e.g., video_track1.srt, video_track2.srt).

The current workaround is to manually extract tracks using ffmpeg -map 0:a:X before importing them into Buzz.

This would be a huge win for editors using DaVinci Resolve or Premiere Pro. Having separate SRTs allows for 100% accurate speaker separation without the extra compute time or potential errors of AI diarization. It also makes it much easier to style speakers differently (like using different colors or screen positions) in the final edit.

[raivisdejus]

@Graham42 Can you share some example file. Best if it is something short but larger file would also be fine.

I see us implementing this feature at some point, seems useful.

[ElsAr4e]

@raivisdejus Not sure if the OP meant the same feature like I do. But I would need this feature:

I have a stereo recording of a conversation where the participants are physically separated by channel (Speaker 1 on the Left, Speaker 2 on the Right).

I would specifically need this to handle the following scenarios:

Physical Channel-to-Speaker Mapping: An opt-in mode to treat Channel L and Channel R as separate speakers automatically.

• Assign Channels to Speakers: Map Channel L to Speaker 1 and Channel R to Speaker 2.
• Visual Distinction: Automatically separate the transcript in the UI and exports (e.g., using different colors or clear labels like "Speaker L" / "Speaker R").
• Parallel Transcription: Process both channels independently to ensure no words are lost during overlaps.

Overlapping Speech: Since the signals are physically separated, the transcript should be able to show both parties speaking simultaneously with their respective timestamps.

Advanced Multi-Person Detection (Intra-Channel): Even if one channel contains more than one person (e.g., a group on a speakerphone on the left side, and a single person on the right), keeping the channels separate first would drastically increase accuracy. Ideally, there would be a secondary opt-in mechanism to perform diarization within a specific channel to automatically detect and label if more than one person is talking on that side (e.g., "Speaker L1", "Speaker L2").

Visual Distinction: It would be great if the export (PDF/SRT) or the UI could reflect this separation (e.g., via labels like "Speaker L" / "Speaker R" or color-coding).

Currently, the workaround of manually splitting files and merging SRTs by timestamp is very time-consuming. Having a "Transcribe channels separately" option would be a game-changer for anyone with high-quality stereo recordings.