voxtype meeting

Overview

Meeting mode provides continuous transcription for calls and meetings, automatically chunking audio into manageable segments with speaker attribution and export capabilities. Prerequisites:

Meeting mode must be enabled in config.toml:

[meeting]
enabled = true
chunk_duration_secs = 30
retain_audio = false
max_duration_mins = 180

Key features:

Continuous recording with chunked transcription (30-second chunks by default)
Speaker attribution via microphone/loopback audio sources
Export to multiple formats: text, markdown, JSON, SRT, VTT
AI-powered summarization via Ollama
Meeting history with SQLite storage

Subcommands

start

Start a new meeting transcription.

voxtype meeting start [--title "Meeting Title"]

--title, -t

string

Optional meeting title. If not provided, defaults to “Meeting YYYY-MM-DD HH:MM”.

Example:

voxtype meeting start --title "Team Standup"

The meeting will be assigned a unique UUID and begin recording immediately.

stop

Stop the current meeting and finalize the transcript.

voxtype meeting stop

Saves the transcript to storage and marks the meeting as completed.

pause

Pause recording for the current meeting.

voxtype meeting pause

Recording can be resumed later with voxtype meeting resume.

resume

Resume a paused meeting.

voxtype meeting resume

Continues recording from where it was paused.

status

Show the current meeting status.

voxtype meeting status

Displays whether a meeting is active, paused, or idle, along with duration and chunk count.

list

List past meetings from storage.

voxtype meeting list [--limit N]

--limit, -l

number

default:"10"

Maximum number of meetings to display. Set to 0 to show all meetings.

Example:

voxtype meeting list --limit 20

Displays meetings ordered by start time (most recent first).

export

Export a meeting transcript to various formats.

voxtype meeting export <meeting_id> [--format FORMAT] [--output FILE] [--timestamps] [--speakers] [--metadata]

meeting_id

string

required

Meeting UUID or “latest” for the most recent meeting.

--format, -f

string

default:"markdown"

Output format. Options: text, markdown, json, srt, vtt.

--output, -o

path

Output file path. If not specified, writes to stdout.

--timestamps

boolean

Include timestamps in the output (format-dependent).

--speakers

boolean

Include speaker labels in the output.

--metadata

boolean

Include meeting metadata header (title, date, duration, model).

Export formats:

Format	Extension	Description
text	.txt	Plain text transcript
markdown	.md	Markdown with formatting
json	.json	Structured JSON with all data
srt	.srt	SubRip subtitle format
vtt	.vtt	WebVTT subtitle format

Example:

# Export latest meeting as markdown to file
voxtype meeting export latest --format markdown --output meeting.md --speakers --metadata

# Export as JSON with all details
voxtype meeting export <meeting-id> --format json --output transcript.json --timestamps --speakers

# Export as SRT subtitles
voxtype meeting export latest --format srt --output subtitles.srt --timestamps

show

Display details for a specific meeting.

voxtype meeting show <meeting_id>

meeting_id

string

required

Meeting UUID or “latest” for the most recent meeting.

Shows meeting metadata including title, duration, status, chunk count, and storage path.

delete

Delete a meeting and its associated files.

voxtype meeting delete <meeting_id> [--force]

meeting_id

string

required

Meeting UUID to delete.

--force, -f

boolean

Skip confirmation prompt.

Example:

# With confirmation prompt
voxtype meeting delete abc123def456

# Skip confirmation
voxtype meeting delete abc123def456 --force

label

Assign human-readable names to auto-detected speaker IDs.

voxtype meeting label <meeting_id> <speaker_id> <label>

meeting_id

string

required

Meeting UUID or “latest” for the most recent meeting.

speaker_id

string

required

Speaker ID to label (e.g., “SPEAKER_00” or just “0”).

label

string

required

Human-readable label to assign (e.g., “Alice”, “Bob”).

Example:

# Label speaker 0 as Alice
voxtype meeting label latest 0 Alice

# Label speaker 1 as Bob using full ID format
voxtype meeting label abc123def456 SPEAKER_01 Bob

Labels are applied to existing transcript segments and saved to the database.

summarize

Generate an AI summary of a meeting using Ollama or a remote API.

voxtype meeting summarize <meeting_id> [--format FORMAT] [--output FILE]

meeting_id

string

required

Meeting UUID or “latest” for the most recent meeting.

--format, -f

string

default:"markdown"

Output format. Options: text, json, markdown.

--output, -o

path

Output file path. If not specified, writes to stdout.

Example:

# Generate markdown summary
voxtype meeting summarize latest --format markdown --output summary.md

# Generate JSON with action items
voxtype meeting summarize abc123def456 --format json --output summary.json

The summary includes:

Brief meeting overview
Key discussion points
Action items with assignees
Decisions made

Complete Workflow Example

Here’s a complete workflow for recording, labeling, and exporting a meeting:

# 1. Start a meeting
voxtype meeting start --title "Sprint Planning"

# 2. Recording happens automatically as you speak...
#    Meeting transcribes in 30-second chunks

# 3. Pause if needed
voxtype meeting pause

# 4. Resume when ready
voxtype meeting resume

# 5. Stop when finished
voxtype meeting stop

# 6. List recent meetings
voxtype meeting list --limit 5

# 7. Label speakers (if using ML diarization)
voxtype meeting label latest 0 Alice
voxtype meeting label latest 1 Bob

# 8. Export transcript
voxtype meeting export latest --format markdown --speakers --metadata --output sprint-planning.md

# 9. Generate AI summary
voxtype meeting summarize latest --output summary.md

Configuration

Meeting mode is configured in config.toml:

[meeting]
# Enable meeting mode
enabled = true

# Duration of each audio chunk in seconds
chunk_duration_secs = 30

# Retain raw audio files (default: false)
retain_audio = false

# Maximum meeting duration in minutes (0 = unlimited)
max_duration_mins = 180

# Storage path (default: ~/.local/share/voxtype/meetings)
# storage_path = "auto"

Storage structure:

~/.local/share/voxtype/meetings/
├── index.db                     # SQLite database with meeting metadata
└── 2024-01-15-team-standup/     # Meeting directory
    ├── metadata.json            # Meeting metadata
    ├── transcript.json          # Full transcript
    └── audio/                   # Raw audio chunks (if retain_audio = true)

Speaker Attribution

Meeting mode supports two speaker attribution methods:

Phase 2: Dual Audio Sources (Current)

Uses microphone and loopback audio to distinguish between local speaker (“You”) and remote participants (“Remote”). Configuration:

[audio]
# Microphone for your voice
device = "default"

# Loopback device for remote audio (virtual device)
loopback_device = "alsa_output.pci-0000_00_1f.3.analog-stereo.monitor"

Phase 3: ML Diarization (Future)

Automatic speaker detection using machine learning models. Speaker IDs like SPEAKER_00 can be labeled with the label command.

Export Format Details

Text Format

Plain text with optional timestamps and speaker labels.

[00:15] Alice: Let's start with the budget review.
[00:32] Bob: The Q4 numbers look solid.

Markdown Format

Formatted transcript with headers and speaker sections.

# Team Standup - 2024-01-15

**Duration:** 45 minutes
**Model:** whisper

## Transcript

**Alice**: Let's start with the budget review.

**Bob**: The Q4 numbers look solid.

JSON Format

Complete structured data with all metadata and segments.

{
  "metadata": {
    "id": "abc123...",
    "title": "Team Standup",
    "started_at": "2024-01-15T10:00:00Z",
    "duration_secs": 2700
  },
  "transcript": {
    "segments": [
      {
        "id": 0,
        "start_ms": 15000,
        "end_ms": 18000,
        "text": "Let's start with the budget review.",
        "source": "microphone",
        "speaker_label": "Alice"
      }
    ]
  }
}

SRT Format (SubRip)

Standard subtitle format for video editors.

1
00:00:15,000 --> 00:00:18,000
Alice: Let's start with the budget review.

2
00:00:32,000 --> 00:00:37,000
Bob: The Q4 numbers look solid.

VTT Format (WebVTT)

Web-compatible subtitle format.

WEBVTT

00:00:15.000 --> 00:00:18.000
<v Alice>Let's start with the budget review.

00:00:32.000 --> 00:00:37.000
<v Bob>The Q4 numbers look solid.

Troubleshooting

Meeting won’t start

Ensure meeting mode is enabled in config.toml:

[meeting]
enabled = true

No speaker attribution

Phase 2 dual audio requires a loopback device configured. Check your audio configuration:

voxtype config

Chunks not processing

Check chunk duration and VAD threshold:

[meeting]
chunk_duration_secs = 30  # Try shorter chunks

[vad]
threshold = 0.01  # Lower = more sensitive

Storage path issues

Verify the storage directory is writable:

ls -la ~/.local/share/voxtype/meetings/

Core Commands

Configuration

Meeting Mode

voxtype meeting

Overview

Subcommands

start

stop

pause

resume

status

list

export

show

delete

label

summarize

Complete Workflow Example

Configuration

Speaker Attribution

Phase 2: Dual Audio Sources (Current)

Phase 3: ML Diarization (Future)

Export Format Details

Text Format

Markdown Format

JSON Format

SRT Format (SubRip)

VTT Format (WebVTT)

Troubleshooting

Meeting won’t start

No speaker attribution

Chunks not processing

Storage path issues

See Also

Build docs developers (and LLMs) love

Core Commands

Configuration

Meeting Mode

​Overview

​Subcommands

​start

​stop

​pause

​resume

​status

​list

​export

​show

​delete

​label

​summarize

​Complete Workflow Example

​Configuration

​Speaker Attribution

​Phase 2: Dual Audio Sources (Current)

​Phase 3: ML Diarization (Future)

​Export Format Details

​Text Format

​Markdown Format

​JSON Format

​SRT Format (SubRip)

​VTT Format (WebVTT)

​Troubleshooting

​Meeting won’t start

​No speaker attribution

​Chunks not processing

​Storage path issues

​See Also

Build docs developers (and LLMs) love

Overview

Subcommands

start

stop

pause

resume

status

list

export

show

delete

label

summarize

Complete Workflow Example

Configuration

Speaker Attribution

Phase 2: Dual Audio Sources (Current)

Phase 3: ML Diarization (Future)

Export Format Details

Text Format

Markdown Format

JSON Format

SRT Format (SubRip)

VTT Format (WebVTT)

Troubleshooting

Meeting won’t start

No speaker attribution

Chunks not processing

Storage path issues

See Also