Meeting Minutes Settings

Meeting Minutes Settings is the configuration surface behind Settings > Meeting Minutes.

It controls how RealTimeX prepares meeting transcription and first-draft synthesis before you start capturing or importing meetings.

Important scope note

Current builds persist most settings from this page in the local app profile or browser storage.

That matters because:

runtime tuning values are local to the current installation or browser profile
custom templates are stored locally
these settings act as defaults layered into new meetings, not as a strict central policy engine

So if you switch machines or app profiles, you may need to reapply some preferences and custom templates.

The four tabs

The current page is split into:

Transcription
Storage
Synthesis
Templates

Transcription

The Transcription tab controls the engine and local runtime behavior used for meeting transcription.

Provider

Current builds expose:

Whisper
OpenAI (coming soon)
AssemblyAI (coming soon)

In practice, the current supported path is local Whisper.

Runtime status

When Whisper is selected, the page shows runtime diagnostics such as:

whether the native runtime is available
whether GPU acceleration was detected
whether the system is currently CPU-only
which backend devices were reported

This helps you tell the difference between:

normal local readiness
desktop-only limitations
missing native runtime availability

CPU threads

The page lets you choose how many CPU threads Whisper can use.

Current builds expose 1 through 4 threads.

Higher values can improve throughput on capable machines, but they also consume more local CPU.

Realtime processing

Realtime processing controls whether audio is processed continuously during recording.

Use it when you want to reduce the time between the end of the meeting and the availability of a transcript or first draft.

Available models

The bottom of the tab exposes local Whisper model management.

Use it to:

review which models are already available
download additional local models
prepare the machine before a meeting starts

Storage

The Storage tab is a review surface for the default Meeting Minutes working-data path.

Current behavior:

the path is shown as read-only
desktop builds can open the folder directly

This is where recordings, transcripts, and generated outputs are stored for that installation's working-data directory.

Synthesis

The Synthesis tab controls what happens when a new meeting generates its first draft.

Default template

Choose which template new meetings should use by default.

The selected template also determines the default output type for first-pass synthesis, such as:

minutes
summary
action items
decisions
follow-up

ACP synthesis agent

You can choose which ACP-backed agent should generate the first draft.

If you leave this at the system default, new meetings fall back to the default ACP agent path instead of one named agent.

Optional model override

When an ACP agent is selected, the page can also pass a specific model override to that agent for meeting synthesis.

Use this when:

the same ACP agent supports several models
meeting drafts should use a different model from the agent's usual default

Watcher cooldown

Watcher cooldown controls how long RealTimeX waits after a meeting goes quiet before auto-generating the first draft.

Shorter values trigger synthesis sooner. Longer values give the meeting more time to settle before automation runs.

Templates

The Templates tab manages reusable synthesis templates.

Built-in templates

Current built-in templates include:

Structured Minutes
Executive Brief
Action Item Tracker
Follow-up Email

Built-in templates are read-only.

Custom templates

To customize a built-in template:

Select it.
Click Duplicate.
Edit the copy.
Save the new custom template.

Custom templates define:

template key
template name
output type
instructions

The instructions are passed into synthesis and stored with new or regenerated drafts.

Good setup sequence

Choose the transcription provider, which is typically Whisper today.
Confirm the runtime status and available backends.
Download the local Whisper models you actually plan to use.
Review the storage path.
Choose the default synthesis template.
Pick the ACP synthesis agent and optional model override if needed.
Set a watcher cooldown that matches how quickly you want first drafts to appear.
Duplicate and customize templates for recurring meeting formats.

When to revisit this page

Return here when:

transcription feels too slow
a machine loses GPU acceleration
you want a different first-draft style
meeting outputs need a new standard format
you moved to a different local model setup