Meeting data files — s1ft Guide

Every meeting that s1ft records gets a folder, and three plain JSON files in it carry the data: meeting.json, transcript.json and telemetry.json. This page lists every field of each file, with its type and its schema version, so you or an AI agent can read them without guessing. For the agent setup, see Connect AI agents.

Before you start

The files sit in the meeting folder. To find the folder, see Finding your recordings.
meeting.json is the last data file s1ft writes for a meeting. When it exists, the transcript is ready.
Every file is written in one step, so a reader sees the old file or the complete new one, never half a file.
Times in meeting.json are ISO 8601 date-times, for example 2026-07-16T10:00:00Z. Times in transcript.json and telemetry.json are seconds.
The files hold measurements and text. They hold no scores, no labels and no summary. A figure such as Talk ratio or Filler rate is not stored in any file. An agent can work it out from the segments in transcript.json.

Schema versions and the add-only rule

Each file starts with a schemaVersion number. Read it first, then read the fields.

File	Current `schemaVersion`
`meeting.json`	1
`transcript.json`	1
`telemetry.json`	2

A new schema version only adds fields. It never removes a field and never renames one. Code that reads version 1 keeps working when version 2 appears, as long as it ignores fields it does not know. The JSON Schemas below leave extra fields allowed for the same reason.

Meetings from before meeting.json

The command line reads meetings through meeting.json. A folder without that file is not a meeting for the CLI. Meetings recorded before meeting.json existed (before S1ft 1.3) are therefore not visible to the s1ft command, and s1ft does not change, convert or remove the files of those meetings. If a list comes back empty, see Where the agent looks for meetings.

meeting.json

meeting.json says which meeting the folder holds: its id, its name, when it ran and which calendar events overlapped it. Current version: 1.

Fields

Field	Type	Meaning
`schemaVersion`	integer	The version of this file. Always `1` today.
`id`	string (UUID)	The id of the meeting. The `s1ft` command takes this id.
`name`	string	The meeting name as s1ft shows it. If you rename a meeting, s1ft writes the file again with the new name.
`platform`	string	The bundle identifier of the app that was recorded, for example `us.zoom.xos`. A manual recording gives `manual`.
`start`	string (date-time)	When the recording started.
`end`	string (date-time)	When the recording ended.
`calendarEvents`	array of event	The calendar events that overlapped the recording. An empty array when no calendar account is turned on or nothing matched.

calendarEvents items

Each entry in calendarEvents has these fields.

Field	Type	Meaning
`title`	string	The title of the event.
`start`	string (date-time)	When the event starts.
`end`	string (date-time)	When the event ends.
`isAllDay`	boolean	`true` for an all-day event.
`attendees`	array of attendee	The people on the event.
`usedForName`	boolean	`true` only for the event that gave the meeting its name. After you rename a meeting, or when no event gave the name, every event has `false`.

attendees items

Field	Type	Meaning
`name`	string or `null`	The attendee name. `null` when the calendar has none.
`email`	string or `null`	The attendee email. `null` when the calendar has none, for example for a room.

Always check for null before you use name or email.

Example

{
  "calendarEvents": [
    {
      "attendees": [
        { "email": "[email protected]", "name": "Ana" },
        { "email": null, "name": "Room 4" }
      ],
      "end": "2026-07-16T10:30:00Z",
      "isAllDay": false,
      "start": "2026-07-16T10:00:00Z",
      "title": "Design Review",
      "usedForName": true
    }
  ],
  "end": "2026-07-16T10:30:00Z",
  "id": "AAAAAAAA-0170-4000-8000-000000000001",
  "name": "Design Review",
  "platform": "us.zoom.xos",
  "schemaVersion": 1,
  "start": "2026-07-16T10:00:00Z"
}

JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "title": "meeting.json (schemaVersion 1)",
  "type": "object",
  "required": ["schemaVersion", "id", "name", "platform", "start", "end", "calendarEvents"],
  "properties": {
    "schemaVersion": { "const": 1 },
    "id": { "type": "string", "format": "uuid" },
    "name": { "type": "string" },
    "platform": { "type": "string" },
    "start": { "type": "string", "format": "date-time" },
    "end": { "type": "string", "format": "date-time" },
    "calendarEvents": {
      "type": "array",
      "items": {
        "type": "object",
        "required": ["title", "start", "end", "isAllDay", "attendees", "usedForName"],
        "properties": {
          "title": { "type": "string" },
          "start": { "type": "string", "format": "date-time" },
          "end": { "type": "string", "format": "date-time" },
          "isAllDay": { "type": "boolean" },
          "attendees": {
            "type": "array",
            "items": {
              "type": "object",
              "properties": {
                "name": { "type": ["string", "null"] },
                "email": { "type": ["string", "null"] }
              }
            }
          },
          "usedForName": { "type": "boolean" }
        }
      }
    }
  }
}

transcript.json

transcript.json holds the words of the meeting, split by speaker. Current version: 1.

Fields

Field	Type	Meaning
`schemaVersion`	integer	The version of this file. Always `1` today.
`segments`	array of segment	The transcript, in order.

Each segment has exactly four fields.

Field	Type	Meaning
`speaker`	string	Who spoke. `Me` is you, from your microphone. Other speakers are `Participant 1`, `Participant 2` and so on, numbered by when they first speak. When s1ft finds only one other speaker, the label is `Participant`.
`start`	number	Start of the segment, in seconds.
`end`	number	End of the segment, in seconds.
`text`	string	The words of the segment.

Older meetings keep a transcript.md file instead. s1ft does not convert it.

Example

{
  "schemaVersion": 1,
  "segments": [
    { "end": 5.5, "speaker": "Me", "start": 0, "text": "Hello everyone" },
    { "end": 9, "speaker": "Participant 1", "start": 5.5, "text": "Hi" }
  ]
}

JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "title": "transcript.json (schemaVersion 1)",
  "type": "object",
  "required": ["schemaVersion", "segments"],
  "properties": {
    "schemaVersion": { "const": 1 },
    "segments": {
      "type": "array",
      "items": {
        "type": "object",
        "required": ["speaker", "start", "end", "text"],
        "properties": {
          "speaker": { "type": "string" },
          "start": { "type": "number" },
          "end": { "type": "number" },
          "text": { "type": "string" }
        }
      }
    }
  }
}

telemetry.json

telemetry.json holds raw measurements of your own microphone, frame by frame. It has no averages, no scores and no data about other speakers. Current version: 2.

Version 2 holds one span for each segment that transcript.json labels Me. Inside a span, the frames come one after the other: frame i starts at start + i * frameMs / 1000 seconds, where frameMs is the value in method.

Fields

Field	Type	Meaning
`schemaVersion`	integer	The version of this file. `2` for the current format.
`source`	string	Where the audio came from. Always `microphone`.
`method`	object	How the frames were measured. See below.
`sampleRateHz`	integer	The sample rate of the microphone audio, in Hz.
`spans`	array of span	One entry for each `Me` segment.

The method object:

Field	Type	Meaning
`frameMs`	integer	The length of one frame, in milliseconds.
`windowSamples`	integer	The number of audio samples that each measurement looks at.
`pitchAlgorithm`	string	The pitch method. `autocorrelation`.
`pitchRangeHz`	array of two integers	The lowest and the highest pitch the method looks for, in Hz.

Each span has four fields, and the three arrays have one value per frame.

Field	Type	Meaning
`start`	number	Where the span starts, in seconds.
`pitchHz`	array of number or `null`	The pitch of each frame, in Hz. `null` for a frame with no pitch, for example a silent frame.
`rmsDbfs`	array of number	The average loudness of each frame, in dBFS.
`peakDbfs`	array of number	The peak level of each frame, in dBFS.

Example

{
  "method": {
    "frameMs": 100,
    "pitchAlgorithm": "autocorrelation",
    "pitchRangeHz": [50, 1000],
    "windowSamples": 2048
  },
  "sampleRateHz": 48000,
  "schemaVersion": 2,
  "source": "microphone",
  "spans": [
    {
      "peakDbfs": [-6, -6],
      "pitchHz": [220, null],
      "rmsDbfs": [-9, -60],
      "start": 1
    }
  ]
}

JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "title": "telemetry.json (schemaVersion 2)",
  "type": "object",
  "required": ["schemaVersion", "source", "method", "sampleRateHz", "spans"],
  "properties": {
    "schemaVersion": { "const": 2 },
    "source": { "type": "string" },
    "method": {
      "type": "object",
      "required": ["frameMs", "windowSamples", "pitchAlgorithm", "pitchRangeHz"],
      "properties": {
        "frameMs": { "type": "integer" },
        "windowSamples": { "type": "integer" },
        "pitchAlgorithm": { "type": "string" },
        "pitchRangeHz": { "type": "array", "items": { "type": "integer" } }
      }
    },
    "sampleRateHz": { "type": "integer" },
    "spans": {
      "type": "array",
      "items": {
        "type": "object",
        "required": ["start", "pitchHz", "rmsDbfs", "peakDbfs"],
        "properties": {
          "start": { "type": "number" },
          "pitchHz": { "type": "array", "items": { "type": ["number", "null"] } },
          "rmsDbfs": { "type": "array", "items": { "type": "number" } },
          "peakDbfs": { "type": "array", "items": { "type": "number" } }
        }
      }
    }
  }
}

Meetings recorded before version 2 hold a version 1 telemetry.json with a different shape. The s1ft command reads version 2 only.

Numbers from the frames: s1ft voice stats

Raw frames are long. The command s1ft voice stats <id> turns the frames of telemetry.json into a short set of numbers for each span. It reads the file and saves nothing. Add --json to get JSON on stdout.

--silence-threshold-dbfs, --min-silence-ms and --percentiles change the thresholds. The defaults are -50, 300 and 10,50,90.

Output fields

Field	Type	Meaning
`frameMs`	integer	The frame length, copied from `method.frameMs`.
`options`	object	The thresholds in use: `silenceThresholdDbfs` (number), `minSilenceMs` (integer) and `percentiles` (array of integers).
`spans`	array of span stats	One entry for each span of `telemetry.json`, in the same order.

Each entry in spans has these fields.

Field	Type	Meaning
`start`	number	Where the span starts, in seconds.
`durationSec`	number	The length of the span, in seconds.
`frameCount`	integer	The number of frames.
`voicedFrameCount`	integer	The number of frames that have a pitch.
`pitchMinHz`	number	The lowest pitch. Absent when no frame has a pitch.
`pitchMaxHz`	number	The highest pitch. Absent when no frame has a pitch.
`pitchRangeHz`	number	Highest minus lowest pitch. Absent when no frame has a pitch.
`pitchPercentilesHz`	object	Pitch values by percentile, with keys such as `p10`, `p50` and `p90`. Empty when no frame has a pitch.
`rmsMeanDbfs`	number	The mean of `rmsDbfs`. Absent when the span has no frames.
`peakMaxDbfs`	number	The highest value of `peakDbfs`. Absent when the span has no frames.
`silenceCount`	integer	The number of silences. A silence is a run of frames below the silence threshold that lasts at least `minSilenceMs`.
`silenceTotalSec`	number	The total length of those silences, in seconds.

cli.json

The s1ft command keeps one small settings file, ~/Library/Application Support/s1ft/cli.json. It has one key.

Field	Type	Meaning
`meetingsDir`	string	The meetings folder that the command uses. `s1ft config dir <path>` sets it. The folder must exist.

--dir beats this key, and this key beats the output folder set in the s1ft app. See Where the agent looks for meetings for the full order.

What’s next

Connect AI agents — Install the s1ft skills and run them after each meeting.
Finding your recordings — Where each meeting folder lives and what else is in it.
Review your delivery with an AI agent — Use your voice numbers for a review of how you speak.