Changelog

v0.7.13

3 October 2026, Latest release

Speaker labels for imported recordings

New

• Who said what in imports. A recording you import, from a dropped audio file or a pasted link, now gets speaker labels for up to eight voices, from the same on-device speaker model as your live meetings. An import has no microphone side, so everyone in it, you included, appears as Speaker 1, Speaker 2 and so on. You can rename them as you would in a meeting.
• A Label speakers switch for links. When you import from a link, Label speakers sits beside the language choice. Kimiko remembers your choice and uses it for dropped files too. It appears while speaker diarization is on in Settings → Transcription.
• Re-transcribe and Re-diarize for imports. With Keep session audio set to 7 or 30 days, Kimiko now keeps an import’s audio encrypted on your computer, as it does for recordings, so Re-transcribe and Re-diarize work on imports too. With it off, nothing is kept.

Improved

• One speaker model everywhere. Re-diarize now uses the same Nemotron 3 speaker model as live meetings and imports, on Mac and Windows alike, so it separates up to eight voices and labels a meeting the way a live call would.
• If a recording seems to have more than eight speakers, Kimiko tells you that labels past the eighth were merged.

v0.7.12

2 October 2026

A tidier License page

Improved

• Clearer license settings. Each row in Settings → License shows one short line, with the details behind its ⓘ. When a license has lapsed or ended, its key row reminds you to Deactivate before you enter a new key.
• Press Tab to reach the ⓘ details in Settings, not only by hovering.

Fixed

• The trial countdown leaves out an allowance you’ve used up. It shows “3 meetings left” instead of “0 days and 3 meetings left”, which could look as if the trial ends today.

v0.7.11

1 October 2026

Manage your license on kimiko.app

New

• Subscribe and manage billing on kimiko.app. Subscribe in Settings → License or on the recording-paused screen opens the monthly or annual plan in your browser. Manage billing opens your Kimiko account, where you can update your card or cancel.
• Manage computers and find your key. Manage computers and Find my key open your Kimiko account, where you can find your key and remove computers you no longer use.
• Easier activation. Already subscribed? Enter your key takes you straight to the key field. If your key is already active on 3 computers, choose Manage computers to free a spot, then Try again.

Improved

• Faster license updates. While Kimiko is open and online, it checks your license about every 15 minutes instead of once a day, and again when you switch back to it if the last check was more than 5 minutes ago. A computer you remove from your account now notices within about 15 minutes. If a check can’t connect, your current license status stays as it was.
• Clearer license wording: an ended subscription reads This license isn’t active any more, and a computer removed from your license says so.

Fixed

• The recording-paused screen now names the real reason recording stopped, such as a removed computer or a lapsed license, instead of saying your trial has ended.
• A computer removed from your license offers Re-activate instead of Subscribe, so you don’t buy a second subscription. You can re-activate straight from the recording-paused screen.

Manage your license and computers

v0.7.10

27 September 2026

Tidier task proposals

Improved

• Proposed tasks take one line. In a meeting’s Tasks pane, a proposed task’s title fits on one line, like a kept task. Hover for the full title, or open it to review. The pane’s header no longer names the AI provider and model behind the proposals.

Fixed

• The Review proposed task and New task windows keep their buttons inside the window, without extra scroll bars.
• A long proposed-task title opens at its beginning in the review window instead of showing only its end.
• Proposed tasks no longer show an empty extra line when the AI leaves a detail blank.

v0.7.8

26 September 2026

An updated speech engine

Improved

• Updated the on-device engine behind the Parakeet, Nemotron 3.5 Streaming, Qwen3 ASR, Voxtral, and Cohere Transcribe speech models.

On a Mac, the first recording or dictation after an update can take noticeably longer to start while Kimiko prepares the speech engine. Later ones start normally.

v0.7.7

26 September 2026

Graphics memory freed after meetings on Windows

Fixed

• Fixed a Windows issue that could keep part of your graphics memory in use after a meeting ended, until you quit Kimiko. It was seen on NVIDIA graphics with the Nemotron 3.5 Streaming speech model. Kimiko now frees that memory when you stop recording.

v0.7.5

25 September 2026

Speaker labels for larger meetings

Improved

• Up to eight remote speakers. Live speaker labeling now uses NVIDIA’s Nemotron 3 Diarization, raising the limit from four to eight. Labeling runs on your device, and your microphone’s lines still appear as You.
• Long meeting, note, and project titles wrap and adapt to the window width instead of being cut off.

When speaker labeling is enabled, Kimiko downloads the replacement model (about 401 MB) and removes the old one after the update. If the new model is not ready, recording can continue without live speaker labels. Local-only mode blocks the download; let it finish before turning that mode back on.

Fixed

• Activating a license now ends the trial on that installation. Deactivating it leaves Kimiko in view-only mode instead of resuming unused trial time or starting a new trial. This also applies to installations activated before this update. Existing notes, search, chat, and exports remain available.

v0.7.4

24 September 2026

A tidier toolbar in smaller windows

Fixed

• The note formatting toolbar no longer overlaps the side pane in a narrow window. Formatting controls fold into a compact menu, and the surrounding controls take up less room.

v0.7.3

22 September 2026

Give the conversation more room

New

• Expand the meeting pane. Give the transcript, live summaries, or tasks the full content area with Expand, then choose Restore to return to your note. You can also use ⌘⇧M on Mac or Ctrl+Shift+M on Windows while the pane is open.
• Chat stays available over the expanded pane. When recording ends, Kimiko restores the split view so you can see the meeting note being written.

Fixed

• Pinned chat keeps unsaved refinements when the layout temporarily tucks it away, and no longer takes typing focus when it returns.
• The empty-projects hint stays still while the sidebar expands, instead of repeatedly wrapping onto new lines.

v0.7.2

21 September 2026

More reliable audio imports

Improved

• Updated security components used by cloud connections and the app’s windows.
• Strengthened local-only safeguards by removing an unused network lookup from AI response validation.

Fixed

• Fixed MP3 imports that could skip a short silent section or add extra audio padding, changing the recording’s duration and shifting transcript timestamps.

v0.7.1

20 September 2026

Find the details in your transcripts

Fixed

• Fixed a search-indexing issue that could leave chat answering from a meeting’s summary while missing details in its transcript. This affected recorded meetings and imported recordings.
• Kimiko now repairs affected search entries in the background after you open the updated app. Large libraries may take a little time to catch up.
• Fixed a timing issue that could attach search information to the wrong passage while a meeting was being reindexed.

Ask questions about your notes

v0.7.0

19 September 2026

More control over imported recordings

New

• Choose the language for a link import. The import window starts with your transcription language setting and offers the languages supported by your selected speech model.
• Import files straight into a project. Project pages now offer both Import audio file… and Import from link…, with the finished note filed into that project.

Improved

• Cancel a link import while it is downloading speech models or transcribing, up until the note is saved.
• The import window names your cloud speech provider when the recording will be sent there for transcription.

Fixed

• File and link imports now respect your transcription language setting. Link imports prefer the recording’s original audio track over an automatically dubbed version.
• Incomplete downloads now report a failure instead of creating a note with parts of the recording missing.
• Turning on local-only mode stops an active link lookup or download. Closing the import window also stops its pending link lookup.
• Fixed long titles overflowing the import window, stuck progress rows, and retries losing their destination project.

v0.6.8

19 September 2026

Turn a recording link into notes

New

• Import from link. Choose Import from link… on the Notes or project page and paste a supported recording link to create a transcript and summary. The note keeps a link back to the source, including in Markdown exports.
• Follow the download and transcription stages from the notes list. Kimiko asks before downloading the helper needed by some sites, and removes the temporary media after processing.
• Added support for importing MP3 audio.

Link imports need an internet connection and use your selected local or cloud speech model. Site restrictions still apply; recordings that require a login or passcode, live streams, playlists, and DRM-protected media are not supported. Only import recordings you are allowed to use.

Improved

• Long audio files use less memory during preparation, for both file and link imports.

v0.6.7

16 September 2026

More control over dictation memory

Choose how long local dictation models stay ready between uses.

New

• Keep models loaded. A new setting in Settings → Dictation controls the idle window for both the speech model and built-in AI formatting. Choose from unloading after each dictation through to never unloading.

Improved

• A five-minute default. Dictation models no longer stay loaded for hours by default. A model also being used for chat can stay ready longer.
• Earlier preparation for formatting. When built-in AI formatting is enabled and selected, Kimiko starts loading it as dictation begins to reduce the wait afterward.

Adjust dictation memory

v0.6.6

16 September 2026

More room for meetings on laptops

Improved

• Dictation makes room for the meeting. Kimiko unloads the waiting dictation model before loading the meeting model.
• More flexible memory use on Windows laptops. On computers with only shared-memory graphics, built-in AI loads model data in a way that lets Windows reuse some of that memory more easily.
• Simpler processing indicators. The transcription indicator shows CPU or GPU, with speaker-processing details available in its tooltip.

Understand memory use on laptops

v0.6.5

15 September 2026

Less graphics work on Windows

Improved

• Speech detection uses the main processor. Windows no longer sends this small, frequent job to the graphics processor.
• Speaker labels adapt to shared-memory computers. On Windows computers with only built-in shared-memory graphics, speaker labeling also uses the main processor. Transcription and local AI can still use graphics acceleration.
• A clearer chat panel. Floating chat no longer dims the document behind it, and the dock control uses a consistent icon.

Choose a setup for your computer

v0.6.4

13 September 2026

Lower resource use after meetings

Improved

• Background AI frees memory sooner. Models used for summaries, task extraction, and search preparation can unload shortly after finishing. Models you are actively using in chat stay ready for follow-up questions.

Fixed

• Fixed a Windows issue that could keep the graphics processor busy after a meeting ended, even after the speech model had been unloaded.

Read about local AI

v0.6.2

10 September 2026

Clearer help when AI tasks fail

Improved

• Summary and note-polishing failures appear in the notification stack with provider details and a Retry action. Failed jobs remain visible after restarting Kimiko.

Fixed

• Improved request handling for OpenCode Zen and Go, including missing-session errors and supported response formats. Model availability still depends on your provider account.

Check your cloud provider setup

v0.6.1

9 September 2026

Speaker names stay in sync

Improved

• Rename a linked speaker from the summary or transcript, and the linked names update together in that meeting.
• Complete saved speaker names, including single-word names, can be linked in summaries. Editing a linked name treats it as one name, helping keep later changes consistent.

Work with speaker labels

v0.6.0

9 September 2026

Project notes and tasks, together

New

• Projects open with their notes and a tasks side panel. Browse the whole project’s tasks or focus on the tasks from one note.
• Create a note or start recording from a project, and the new item is filed into that project.

Improved

• Large speech and AI model downloads use available bandwidth more efficiently, with more resilient retries and resume handling.
• Drag individual list items in notes without moving the whole list.
• Tidier task rows leave more room for the task itself.

Keep track of tasks

v0.5.9

7 September 2026

More room for the conversation

The meeting side pane makes better use of smaller windows, with tidier controls and task rows.

Improved

• More flexible meeting controls. The transcription model label and pane controls adapt as you resize the window. Compact controls keep their names available on hover.
• Easier-to-scan tasks. Denser task rows leave more room for the work itself, with less space spent on secondary details.
• Quieter license reminders. License notices now join the notification stack and stay quiet while you record or dictate.
• A familiar Windows installer. The installer now uses Kimiko’s official color artwork.

Fixed

• Task text truncates correctly in the affected compact rows.

Explore meeting notes

v0.5.8

6 September 2026

Smoother dictation on Windows

This release focuses on starting dictation and getting your words into the right place.

Improved

• A new Windows default shortcut. Dictation uses Ctrl + Win by default. Mac keeps its Ctrl shortcut; check Settings for your current binding.
• A quicker start. Dictation starts preparing the microphone and showing its indicator earlier in the shortcut flow.
• An optional ready sound. Enable a short tone when the microphone is ready. It is off by default.

Fixed

• The dictation indicator no longer takes focus away from the Windows app you are typing in.
• Improved insertion into Kimiko’s own notes and Windows 11 Notepad.
• Improved handling of paste shortcuts, line breaks, and held modifier keys during insertion.

Set up voice dictation

Get the latest version

Open Kimiko’s update controls in Settings, or download the latest Mac or Windows installer.