Skip to content

feat: speaker diarization for transcription output #80

Description

@lovesahaj

Problem

SpeakType currently transcribes speech into text, but for recordings with more than one speaker, there is no way to tell who said what.

This makes the transcription less useful for meetings, interviews, calls, and conversations because speaker changes have to be added manually afterwards.

Expected behavior

It would be useful if SpeakType could optionally perform speaker diarization and include speaker labels in the transcription output.

For example:

Speaker 1: Let's start with the onboarding flow.
Speaker 2: Sure, I think the permission step is confusing.
Speaker 1: Agreed, we should explain why access to the microphone is needed.

Timestamps would also be useful, but generic Speaker 1, Speaker 2, etc. labels would already be enough for an initial version.

This could be an opt-in setting so the current fast dictation workflow stays unchanged by default.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or requestpriority:P3Lower priority backlog itemroadmapProduct direction or future exploration

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions