Talk Recording

Press C in tap dev to record your screen and microphone. Press it again to stop. Recording writes a chapter list of slide timings beside the video, so a video description is most of the way written by the time you stop talking.

Recording every talk automatically

tap dev is for writing a talk, so it never records on its own; C is always a manual choice. For giving the talk itself, tap present can record every run from launch to quit without a key press, following the projector across HDMI swaps. See the Presenting guide.

macOS Only

Recording uses screencapture, the capture binary already built into macOS, so there is nothing to install. Tap does not bundle ffmpeg and does not ask you to.

On Linux and Windows, C does nothing and no recording controls appear. If a deck's frontmatter has a recording block, tap dev prints a startup warning that recording is macOS only and the block is ignored; the deck itself still runs normally.

Screen Recording Permission

Recording needs the Screen Recording permission, and it is granted to the terminal application that runs tap, not to Tap itself. Open System Settings > Privacy & Security > Screen Recording, enable your terminal (Terminal, iTerm, Ghostty, whichever you use), and restart the terminal for the grant to take effect.

tap dev checks this when it starts the terminal interface on macOS, by taking a tiny test screenshot, and says nothing when it succeeds. If the permission is missing, or was revoked, the check fails immediately and the terminal prints a warning naming the exact fix above. This check runs whether or not the deck asks for recording, so a missing grant surfaces while you are still setting up rather than when you press C on stage. The same grant covers audio capture too, so macOS never raises a separate microphone prompt on top of it.

Choosing a Display

With one display attached, pressing C starts recording immediately; there is nothing to choose. With more than one display attached, such as your laptop plus a projector, C opens a picker first:

  Record

  Display
  > Built-in Retina Display   3456 x 2234   main
    DELL U2718Q               3840 x 2160

  Audio
    MacBook Pro Microphone (system default input)
    Change it in System Settings or the menu bar.

  enter start    t test 5s    esc cancel

Use the arrow keys or j/k to move between displays, then enter to start recording the highlighted one, or esc to cancel.

Press t to run a five second test capture of the current selection without leaving the picker. It opens in QuickTime Player when it finishes, so you can watch the right screen came through and listen for your microphone before you are on stage. Test again after switching displays or plugging in the projector; the picker stays open so you can.

The Microphone

The picker shows the name of the system's current default input, read only. Tap cannot offer a dropdown to change it: screencapture selects a microphone by its CoreAudio device UID, and no pure-Go binary can read those UIDs back out of macOS. Building that lookup would mean adding cgo to a project that cross-compiles cleanly today, for a picker that would only ever show a UID nobody recognizes anyway.

To record from a different microphone, change the system default input first, in System Settings > Sound or from the microphone icon in the menu bar, then open the picker. It will show the new default.

If you already know your microphone's CoreAudio UID, recording.audio accepts it directly:

recording:
  audio: BuiltInMicrophoneDevice

recording.audio also accepts default (the same as leaving it out) and none, which records a silent video, useful for a demo where the audio would just be typing.

Where Recordings Land

By default, recordings are written to a recordings/ directory next to the deck, alongside a matching chapter file with the same name and a .txt extension. Nothing creates that directory until you actually press C: tap dev does not touch the filesystem just because a deck's frontmatter has a recording block. Set recording.output to write somewhere else:

recording:
  output: /Users/me/Movies/talks

A relative path is resolved against the deck's own directory.

The first time a recording finishes inside a git repository with no existing .gitignore entry covering the output directory, tap dev asks whether to add one:

  Recording saved

  A recording is large, and this deck is in a git repository.

  Add recordings/ to .gitignore? (y/n)

The entry offered is derived from wherever recording.output actually points, so it is correct even when recordings do not land in the default recordings/ directory. Answering y appends the entry once; the prompt does not appear again once it is there.

After tap dev exits, the terminal prints the last recording's path and its chapter file:

  Last recording: recordings/quarterly-results-2026-09-20-1432.mov
  Chapters: recordings/quarterly-results-2026-09-20-1432.txt

The Chapter List

While recording, Tap writes one line to the chapter file every time the slide changes, in the format YouTube's chapter parser expects: a timestamp, then the slide's title.

• The first line is always 0:00, whatever slide is on screen when recording starts.
• Moving to a new slide adds an entry.
• Moving back to a slide you have already shown adds another entry: the chapter list follows what actually happened, not just forward progress.
• Landing on the same slide again immediately, such as an accidental double key press, adds nothing.
• Revealing a fragment or a step on the same slide adds nothing; only a change of slide creates a chapter.

A slide's title comes from its first heading. A slide with no heading, such as an image-only or component-only slide, is labeled Slide N.

A worked example, ready to paste into a YouTube description:

0:00 Introduction
0:42 The problem
2:15 What this adds
5:03 Live demo
9:47 Questions
1:04:12 Live demo

Long Recordings

Tap warns in the TUI if a recording is still running after 90 minutes, and stops it automatically at 3 hours, so a recording nobody remembered to stop does not fill the disk overnight. Both are configurable as durations such as 90m or 3h:

recording:
  warnAfter: 45m
  stopAfter: 2h

Set stopAfter: off to disable the automatic stop entirely. There is no way to disable the warning; it only ever writes a line to the event log.

Stopping and Quitting

Pressing q while a recording is running asks first, since screencapture cannot resume a file it has already closed:

  Recording in progress

  my-talk-2026-09-20-1432.mov  14:32

  Stop recording and quit? (y/n)

y stops the recording, finalizes the file, and quits. n or esc returns to the TUI with the recording untouched.

Ctrl+C does not ask. It still finalizes the file: tap dev stops the recorder as part of its normal shutdown, so an interrupted recording is still a playable video, just without the chance to say no.

Frontmatter Reference

Every key under recording is optional; an omitted key keeps its default.

---
recording:
  # Directory recordings are written to, relative to the deck unless given
  # as an absolute path. Default: recordings
  output: recordings
  # "default" for the system input, "none" for a silent recording, or a
  # CoreAudio device UID. Default: default
  audio: default
  # How long a recording runs before the TUI warns about it. Default: 90m
  warnAfter: 90m
  # How long a recording runs before it stops itself. "off" disables the
  # cap. Default: 3h
  stopAfter: 3h
  # Preselects a display in the picker by its screencapture index (1 is
  # the main display). 0 means no preference. Default: 0
  display: 0
  # Draws mouse clicks in the recording. Default: false
  showClicks: false
  # Writes the sidecar chapter list; set to false to turn it off.
  # Default: true
  chapters: true
---