What's new.

Every release of zerosoft, newest first. Written from the real commit history: no roadmap promises here.

Atom feed
  1. Voice: meeting transcripts no longer drop pieces when the provider is busy

    • VoiceA meeting sends a piece of audio every 20 seconds for your microphone and another for the computer audio. Transcription used a single model, which only accepts a few requests per minute, so many pieces failed and the transcript stayed empty. Now, when that model is full, the piece goes to a second transcription model (and to Groq when a key is set), and the full model is skipped for a minute. If every model is full, the recorder waits and sends the piece again instead of losing it. Voice notes in Inbox and Lines and Calls recordings use the same fallback.
  2. Voice: refreshing a meeting keeps your edits and checkboxes

    • VoiceWhile a meeting page refreshed itself, the next steps you had unchecked came back checked, and text typed right after an autosave could jump back. Now a refresh only replaces the note when it is a new one and nothing you wrote is waiting to be saved.
  3. Talk and Voice: analytics with the same layout as Lines

    • VoiceNew Analytics page with your own numbers: meetings, time in meetings, AI notes, dictations, words dictated and words per minute against the previous period, activity per day, when you record, where you dictate and a table of the meetings in the period. Filter by meetings or dictation.
  4. Desktop for Mac: Accessibility keeps working after an update

    • VoiceThe Mac app installed each new version over itself while it was open. From then until you restarted it, macOS did not recognize its Accessibility permission even with the switch on, so dictation copied the text instead of pasting it, and turning the switch on again or resetting it did not help. Version 0.3.11 downloads the update in the background and installs it when you choose Restart to update or quit the app. If the app is still replaced while it is open, for example by dragging a new copy into Applications, Settings says so and offers Restart zerosoft instead of Reset permission.
  5. Voice: the first dictation is as fast as the rest

    • VoiceThe first dictation after a while used to take much longer than the next ones, because the server had to start up while you waited. Now it stays ready, and the desktop app also wakes it up as soon as you start talking.
  6. Voice: dictation waits less before it pastes

    • VoiceA dictation now asks for your settings, dictionary and quotas while the audio is still arriving, and keeps them for a few seconds for the next dictation. Usage and history are saved after the text is back. The time spent between the audio arriving and the text coming back, outside transcription and cleanup, went from four to seven database round trips in a row down to one. Dictations are transcribed with GPT-4o mini Transcribe, which OpenAI reports as more accurate than Whisper, and fall back to Whisper if it fails. With a Groq key set, Groq goes first because it is the fastest. The response header now also shows how long the session check and the upload took.
  7. Voice: dictated lists, punctuation and emails come out formatted

    • VoiceSay "I need a list, colon, eggs, meat and milk" and Voice pastes a bulleted list with one item per line. "Number one… number two…" gives a numbered list. Spoken punctuation becomes the symbol: "comma", "colon", "open paren", "open quote", "new paragraph", and "ana at zerosoft dot ai" becomes an address. It works in English, Spanish and Portuguese, and short dictations from the desktop app get this formatting without waiting for the AI. In an email the greeting and the sign-off you said go on their own lines; in a chat the text stays as you said it. Voice never adds a greeting you did not say.
  8. Voice: dictation lays out lists and emails

    • VoiceWhen you dictate three or more items, they come out as a bulleted list; steps come out numbered. In an email, the greeting and the sign-off you say go on their own lines and the body in paragraphs. It only arranges what you said: it does not add words.
    • VoiceNew setting "How much to clean" (Voice › Settings, and in the desktop app after its next update): a little (punctuation and filler words), normal (also removes hesitations like "let me think" and builds lists) or a lot (also removes repetitions and detours).
    • VoicePhrases the transcription makes up during silence (subtitle credits, a newsreader greeting in another language) are removed.
    • VoiceDesktop app: if the computer audio is not coming through during a meeting, a notice says so and how to fix the permission, instead of recording only your microphone without telling you.
  9. Desktop app: tabs and windows

    • VoiceOpen each app or record in its own tab or window: Command + T opens a tab with search, Command + N a window, Command + W closes and Command + Shift + T brings back the last one. Command-click or middle-click a link to open it in a new tab, or right-click it. Drag tabs to reorder them or out to their own window. Tabs show the app and the record, with a dot when there is something new.
    • VoiceThe app reopens your windows and tabs where you left them, and puts tabs you have not used for a while to sleep to save memory (Settings › General › Tabs and windows).
  10. Voice: records the computer audio, with its own transcription allowance

    • VoiceMeetings record the audio of the meeting next to your microphone: the browser asks for a tab or screen with audio when you start, and the Mac app records the system audio and asks for the Screen & System Audio Recording permission. If nothing is shared, only the microphone is recorded and you are told why.
    • VoiceTranscription has its own monthly allowance of 600 minutes per workspace, so chatting with your notes no longer leaves a meeting without a transcript.
  11. Voice: a cleaner meeting page with Ask anything

    • VoiceA meeting is now a plain page: title, a chip for enhanced or your own notes, the day and attendees from your calendar (today’s event is picked on its own, attendees grouped by company) and the folder.
    • VoiceThe bar under the notes holds the live transcript and Ask anything (⌘J or /): recipes like List my todos or Write TL;DR, a recipe library you can share with the workspace, separate chats with history, a fast or thorough model, a text file attached to the question and dictation.
    • VoiceWrite follow up email drafts the email from the notes and sends it from your connected Gmail or Outlook, or you copy it.
  12. Every setting in one place, on the web and in the desktop app

    • VoiceDesktop 0.3.2: Permissions with Reset permission for Accessibility, microphone, camera and screen recording; the version and Check for updates in General and Updates; a notice in the app when an older version needs to be replaced; and a smaller recording indicator that you can drag anywhere.
  13. Voice: meetings start listening right away, and chat cites the transcript

    • VoiceA new meeting starts recording the microphone as soon as it opens, from New meeting, Take notes or ⌘K. If the browser blocks the microphone or there is none, the meeting says so and you can retry. A link someone sends you never turns the microphone on.
    • VoiceChat answers appear as they are written and cite the moment in the transcript: click a time to jump there. Suggested questions change while the meeting is on and after it ends.
    • VoiceAsk across your meetings with follow-up questions and dates like "last week". Answers link to each meeting and moment, and include meetings teammates shared with you, using only what the share lets you see.
  14. Voice: pick how much each dictation gets cleaned up

    • VoiceA new Cleanup setting in Voice: raw (word for word, only your dictionary replacements), clean, formal everywhere, or by app with the tones you already set. Dictations made from the desktop app now show as such in the history.
  15. zerosoft for Mac and Windows: dictation in any app and meeting notes with system audio

    • VoiceA desktop app for Mac and Windows with every zerosoft app in one window. Hold Control + Option (Mac) or Ctrl + Windows in any app, speak and let go: the clean text is pasted where the cursor is.
    • VoiceWhen a calendar meeting starts, or Zoom, Meet or Teams is in a call, the app offers to take notes. It records your microphone and the system audio, without a bot, and the notes land in Voice like any other meeting.
  16. Voice: your calendar with briefs and reminders, sharing by link, recipes, follow-up emails and Command Mode

    • VoiceMeetings now come from Cal bookings and the Google Calendar or Outlook you connected in Cal, with their attendees. Take notes from an event, get a reminder a few minutes before, and read a brief of what you discussed before with the same people.
    • VoiceShare a meeting by link with your workspace, with the people you pick or with anyone, read only and with the transcript only if you include it. Change or revoke the link anytime. Meetings shared with you show up in "Shared with me".
    • VoiceFolders with one level of subfolders, recipes you run with / on one meeting or across meetings, a follow-up email drafted by AI that opens in Gmail or Outlook, and sending the notes to Slack or Notion.
    • VoiceDictation: Command Mode rewrites the selected text with what you say, saved transforms rewrite it with one click, and fixing a dictation in the history teaches your dictionary. Admins get a team dictionary and a usage panel.
    • VoiceTranscripts can delete themselves after the days you pick (admins can set a maximum), the new MCP server lets Claude or ChatGPT read your meetings, and the voice.meeting.completed and voice.notes.ready webhooks reach only the webhooks created by the person who recorded.
  17. Voice: dictation in every app and meeting notes without a bot

    • VoiceHold Alt+Space in any zerosoft text field, speak and let go: the text lands where your cursor is, with punctuation, no filler words, your spoken corrections applied and lists formatted. A personal dictionary, snippets and a style per context (email, chat, note), plus a history to copy past dictations.
    • VoiceMeeting notes: record your microphone, or your microphone and a browser tab, with a live transcript split into Me and Them. Jot notes as you go and AI writes them up with a template (sales, 1:1, stand-up, interview, customer or your own), with a summary, decisions and next steps that become Track issues or CRM tasks.
    • VoiceAsk a meeting or all of them ("what did Ana say about pricing?"), search across transcripts, link a Cal booking so the notes land on the CRM contact timeline, and send the notes to Docs or Notes. API v1 endpoints with the voice:read and voice:write scopes, and the source of a desktop app for Mac and Windows.
  18. Voice, Rec and Talk join the suite, in the works

    • VoiceVoice, the app we are building to take on Wispr Flow and Granola (dictation and meeting notes), has a page with the plan and its parity at 0 %, and /app/voice says it is being built. You can propose ideas for it on the board.