Diarization

​

Overview

• Meeting transcripts — Attribute statements to participants
• Interviews — Separate interviewer and interviewee
• Podcasts — Identify hosts and guests
• Call recordings — Distinguish callers

Granite Speech speaker-turn transcripts are documented separately as Speaker Attributed ASR. Use diarization when you need timestamped speaker segments.

​

Getting Started

​

Download Diarization Pipeline Models

izwi pull diar_streaming_sortformer_4spk-v2.1
izwi pull Parakeet-TDT-0.6B-v3
izwi pull Qwen3-ForcedAligner-0.6B

​

Start the Server

izwi serve

​

Using the Web UI

1. Navigate to Transcription in the sidebar
2. Switch to Diarization mode
3. Upload an audio file with multiple speakers
4. Click Analyze
5. View the speaker-segmented transcript

​

Output

• Speaker labels — Speaker 1, Speaker 2, etc.
• Timestamps — When each speaker talks
• Transcript — What each speaker said

[00:00 - 00:05] Speaker 1: Welcome to the meeting.
[00:05 - 00:12] Speaker 2: Thanks for having me.
[00:12 - 00:20] Speaker 1: Let's start with the agenda.

​

Using the API

​

Endpoint

POST /v1/speech-to-text/jobs?job_kind=diarization

​

Request (multipart/form-data)

​

Example (curl)

curl -X POST "http://localhost:8080/v1/speech-to-text/jobs?job_kind=diarization" \
  -F "file=@meeting.wav" \
  -F "model=diar_streaming_sortformer_4spk-v2.1" \
  -F "asr_model=Parakeet-TDT-0.6B-v3" \
  -F "aligner_model=Qwen3-ForcedAligner-0.6B" \
  -F "min_speakers=2" \
  -F "max_speakers=2"

​

Response

{
  "id": "diarization_...",
  "kind": "diarization",
  "processing_status": "pending"
}

​

Configuration

​

Number of Speakers

# Via API
curl -X POST "http://localhost:8080/v1/speech-to-text/jobs?job_kind=diarization" \
  -F "file=@meeting.wav" \
  -F "min_speakers=3" \
  -F "max_speakers=3"

​

Speaker Labels

​

Tips for Best Results

1. Quality audio — Clear recordings with minimal background noise
2. Distinct voices — Works best when speakers have different voice characteristics
3. Minimal overlap — Speakers talking over each other reduces accuracy
4. Specify speaker count — If known, helps the algorithm
5. Longer segments — Short utterances are harder to attribute

​

Limitations

• Similar voices — May confuse speakers with very similar voices
• Overlapping speech — Simultaneous talking is challenging
• Background noise — Reduces speaker detection accuracy
• Very short clips — Need enough audio to identify speaker patterns

​

Use Cases

​

Meeting Minutes

1. Record your meeting
2. Upload to Diarization
3. Export the speaker-labeled transcript
4. Edit speaker names as needed

​

Interview Transcription

1. Record the interview
2. Process with diarization
3. Get clean Q&A format output

​

Podcast Production

1. Upload raw podcast audio
2. See who spoke when
3. Use timestamps for editing

​

See Also

• Transcription — Single-speaker transcription
• Speaker Attributed ASR — Granite speaker-turn transcripts
• Voice Mode — Real-time conversations
• CLI Reference — Command documentation