Anarlog 1.4.18 records meeting audio from whatever speaker your call uses, tells apart people sharing one mic, learns your voice on its own, connects your Claude Pro or Max plan, and adds a reasoning effort setting for your own models.
Recording and audio
Capture system audio from whatever output your meeting app is playing
through, so a call routed to a headset or second display no longer records
silence; the Speakers setting is gone because it is no longer needed
Skip echo cancellation when your default output is a headphone, which keeps
your own speech clean and saves CPU
Record from a wired or built-in microphone when the default input is a
Bluetooth headset, avoiding the muffled, gated call-mode mic, and warn when a
Bluetooth mic is dropping audio (macOS and Windows)
Start listening immediately instead of waiting on a sync round, and open a
scheduled meeting link the moment the meeting is due even while listening is
still connecting
See Share in the note header once a meeting has ended instead of a
redundant Record button
Speakers and transcripts
Keep the speaker names you assign during a meeting in the live transcript,
including after listening restarts on the same note
Separate the people sharing one microphone in an in-person meeting instead
of labeling everyone as You
Name the last unidentified remote participant automatically when every other
speaker is already known and the count matches
Merge over-split speakers in Pro batch transcripts down to the number of
people actually in the meeting
Learn your own voiceprint from headphone-isolated recordings without a
manual speaker assignment
Keep live speaker separation on for ad-hoc meetings that have no calendar
attendees
Search the transcript by keyword, and use the speaker picker without it being
clipped
Intelligence
Connect your Claude Pro or Max plan from the Anthropic card on the
Intelligence page and use Sonnet, Opus, and Haiku without an Anthropic API
key
Pick a Reasoning effort (Default, Low, Medium, or High) for your own
language models on the Intelligence page; it resets to Default when you
change provider or model
Add Smallest AI as a transcription provider, with Pulse for live and
after-recording transcription and Pulse Pro for after-recording
Refresh every transcription provider's model list: retired AssemblyAI,
AquaVoice, and Soniox models move to their current replacements, deprecated
OpenAI transcription models are labeled, and the discontinued Fireworks
provider is marked as such so existing selections fall back cleanly
Download the built-in Soniqo Parakeet models even when a leftover Hugging
Face token is set on your machine, instead of stopping on a token prompt the
public models never needed
Recover Pro summaries after signing in again instead of failing on an
expired token, and wait for sync to settle before saving
Keep chat open when a request fails instead of dropping to an error screen
Polish
Use smooth-cornered controls throughout Settings and the note header,
including selects, search fields, and the calendar navigation
Show Join & record, Record, and Share in the note header as
outlined buttons that follow your theme
Align the session header controls with the sidebar toggle and compact the
view tabs
Keep Settings controls visible in a narrow window, list AI above
Workspace in the Settings sidebar, and left-align the signed-out account
panel
Show a single sign-in action on the onboarding account step instead of two
buttons that did the same thing
