Problem / Motivation
Soniox identifies speakers well, and users want speaker labels in transcripts, especially for uploaded files and longer conversations. This came up as follow-up feedback on PR #703 after the Soniox v5 plugin update.
Proposed Solution
Add an opt-in Soniox setting for speaker diarization, pass the Soniox diarization option in async transcription requests, parse speaker metadata into TypeWhisper's existing speaker fields (PluginStructuredTranscriptionSegment.speakerLabel / speakerConfidence), and preserve labels through the existing transcript/export paths.
Alternatives Considered
Keep using providers that already expose diarization, such as AssemblyAI, until Soniox support is implemented.
Problem / Motivation
Soniox identifies speakers well, and users want speaker labels in transcripts, especially for uploaded files and longer conversations. This came up as follow-up feedback on PR #703 after the Soniox v5 plugin update.
Proposed Solution
Add an opt-in Soniox setting for speaker diarization, pass the Soniox diarization option in async transcription requests, parse speaker metadata into TypeWhisper's existing speaker fields (
PluginStructuredTranscriptionSegment.speakerLabel/speakerConfidence), and preserve labels through the existing transcript/export paths.Alternatives Considered
Keep using providers that already expose diarization, such as AssemblyAI, until Soniox support is implemented.