MemoFlow 0.5.0

Recordings that survive the real world, and a home screen that shows what you've captured.

Dashboard

MemoFlow now opens on a summary of what it has captured — words dictated and transcribed, meetings recorded, action items extracted, when you dictate, and activity over time — instead of an empty pane.

Recording keeps going when your mic doesn't

• Unplug a headset mid-meeting and the recording moves to another input instead of ending
• Set a preferred order of microphones, so it falls back the way you want rather than to whatever macOS picks
• A warning when your microphone is clipping, because clipped audio silently transcribes to nothing
• A short tone when recording starts and stops — often the only confirmation you get when dictating into another app

Dictation modes you define

• Per-app and per-site modes with their own engine, prompt, and style rules
• Say a mode's trigger phrase to switch on the fly
• Auto-send: have Return pressed for you after insertion, per mode
• Word replacements, so names and jargon come out spelled your way every time

Better answers in Ask

• Answers render as Markdown — headings, lists, bold
• A question about a standup on a date, or across a range, is answered day by day
• OpenRouter joins Gemini as a cloud provider: one key, any model

Transcription and reliability fixes

• The Hinglish model's confidence gates are genuinely disabled now — they were gated behind a flag that was never set
• The start tone no longer lands at the head of the dictation buffer, where it made the model return nothing
• nan no longer appears in transcripts
• Retention policy: delete old meeting audio on a schedule while keeping transcripts and summaries
• Diagnostics report, for bug reports

Housekeeping

• What's New appears once after an update, and any time from Help ▸ What's New
• Settings fills the window instead of sitting in a narrow column

Requires macOS 26.

MemoFlow 0.4.4

What's new

• Dictated lists format themselves — say "bullet point…" or enumerate ("first… second… next…") and your dictation is inserted as clean bullet points; "new line" and "new paragraph" work too. Only your words, structured.
• Custom tags — add your own tags from the + menu (New Tag…), filter by them in the sidebar, and ask about them in chat. Chips now remove with an × on hover.
• Instant standup extraction — manually tagging a meeting as Standup extracts your update immediately, as tidy bullet items in the Standups journal.
• Design polish — more breathing room in the meeting header and sidebar; standup cards render bullet points.

Requires macOS 26. Existing installs update in-app via MemoFlow → Check for Updates.

MemoFlow 0.4.3

What's new

• Meeting tags — meetings are auto-categorized (Standup, 1:1, Planning, Review, Interview, Client Call, Brainstorm, All-Hands) with chips on the meeting header and list, a sidebar tag filter, and manual add/remove. Existing meetings are tagged in the background on-device.
• Standup journal — meetings tagged Standup get your own update (done / next / blockers) extracted from your mic track into a new Standups view, organized by day.
• Smarter Ask — the global chat now looks things up itself (transcripts, meetings by date or tag, standup updates, action items), so questions like "what was my update last Monday?" get precise answers with citations.
• Streaming answers — chat responses type out live when using Gemini.
• Long-meeting summaries — Gemini summarization now processes long transcripts in chunks, faster and more reliably.

Requires macOS 26. Existing installs update in-app via MemoFlow → Check for Updates.

MemoFlow 0.4.2

Context-aware dictation: output matches the app you're dictating into (code stays code, emails read professionally, chats stay casual), on-screen names and identifiers are spelled exactly, and in Cursor/Windsurf/Antigravity/VS Code chats spoken file names become @-tags with real project paths. Personal dictionary teaches Whisper your names and terms. Optional Gemini cloud intelligence for summaries and chat (your own API key, off by default — audio and transcription always stay on-device).

MemoFlow 0.4.1

Much faster Hinglish model downloads: 8 parallel resumable streams from Cloudflare R2 (GitHub as automatic fallback). Interrupted downloads now resume where they stopped.

MemoFlow 0.4.0

In-app updates: MemoFlow now updates itself. Check for Updates lives in the app menu and Settings; automatic daily checks are on by default. From this version on, new releases arrive as a one-click update inside the app.

MemoFlow 0.3.0

Engine-choice onboarding with real Hinglish examples and a shortcuts step, factory reset, guided one-time Neural Engine optimization with progress, Hinglish AI as the default engine (Apple fallback until the model is downloaded), and a redesigned onboarding + Settings. Notarized.

MemoFlow 0.2.0

Notarized production build. Private, on-device AI meeting notes and Hinglish dictation for macOS 26+. The Hinglish model downloads in-app on first use.

Hindi-Hinglish Apex (WhisperKit CoreML)

Oriserve/Whisper-Hindi2Hinglish-Apex converted to WhisperKit CoreML. Downloaded in-app by MemoFlow.