Files
Where things are
| Path | Content |
|---|---|
~/Plaud/ |
The archive (the folder is a setting) |
/Applications/My Plaud.app |
The app |
~/Library/Application Support/my-plaud/settings.json |
Settings, written by the Settings pane |
~/Library/Application Support/my-plaud/tokens.json |
The Plaud login tokens. Never share this file |
~/Library/Application Support/my-plaud/elevenlabs.key, anthropic.key, openai.key, openrouter.key |
The API keys |
~/Library/Application Support/my-plaud/models/ |
The speech models of the engine on this Mac, about 1.7 GB |
~/Library/Application Support/my-plaud/state.json |
The last sync’s summary, pauses, the balance estimate, the recordings you deleted |
The tokens and the keys are readable by your user only.
The archive
The archive is organised by local start date and time, one folder per recording:
~/Plaud/2026/11/2026-11-19_1455 Website relaunch kickoff/
audio.ogg the original audio (audio.mp3 for imported files), or the audio recorded on this Mac
video.mov the video of a recording made on this Mac with one
transcript.md the transcript with speakers and times (cleaned, with a cleanup model)
summary.md the summary, with a summary model
speaker-1.ogg a voice sample of "Speaker 1", up to 59 seconds
speaker-2.ogg one per speaker without a name
Four hidden files sit beside them:
| File | Content |
|---|---|
.transcript.json |
The raw transcript as the speech-to-text provider returned it |
.transcript.clean.json |
The cleaned text of each turn |
.transcript.edits.json |
Your corrections and the names you gave “Speaker N” |
.manifest.json |
The recording’s ID, checksums and the name you gave it. Do not edit |
Folder names
- A folder is named by the recording’s start,
2026-11-19_1455. A second recording that started in the same minute gets-2. - With a summary model and “Add the name to the folder” on, the name follows the date and time.
/and:become spaces, and the name is cut to 80 characters. - The part before the first space is the folder’s identity. Anything you type after it in Finder is left alone.
- Renaming a recording in Plaud changes nothing locally.
What is not downloaded
Plaud’s own transcripts and AI notes are not downloaded; they stay in Plaud.
What a sync leaves alone
Once a recording’s audio is stored, later syncs leave it alone unless the file goes missing or Plaud reports a different length; then it is downloaded again.
The log
There are no log files. The log is in the macOS unified log:
log show --predicate 'subsystem == "net.mejta.my-plaud"' --last 1h
Copy log command in Diagnostics copies this command. The log holds recording IDs, HTTP statuses, error codes and counts; never recording names, transcript text, summaries, keys or the account’s email.