Wispr Flow + Granola alternative zerosoft Voice, open source.
Voice does the core of what Wispr Flow + Granola does. It runs on your server for $0 under AGPLv3, or on zerosoft Cloud with a free plan and usage-based pricing, with no charge per seat. By our count it covers 73% of the 48 Wispr Flow + Granola capabilities we measure.
What you get
- Remove filler words and fix grammar and punctuationAI cleanup of fillers, punctuation and capitals, with a rules fallback when AI is off
- Personal dictionaryWords to spell right (they also bias the transcription) and replacements; fixing a dictation in the history adds the corrected names and terms by itself, with undo. Fixes made in other apps after pasting are not seen
- Live transcriptSearchable transcript split into Me (mic) and Them (tab), arriving in parts of about 20 seconds
- Calendar sync and meeting remindersThe Cal bookings you host or co-host plus the Google Calendar and Outlook events connected in Cal, with attendees and the call link; take notes from an event and get a reminder a few minutes before (in the page and as a browser notification) while zerosoft is open in a tab. Admins only see their own meetings, not the whole workspace
- AI notes from your notes and the transcriptAI combines your notes with the transcript and the template in headings and nested bullets; your own lines stay in ink and the AI ones in gray. The note is edited in place like a document (Enter adds a bullet, Tab nests it) and saves itself
- Note templatesSales, 1:1, stand-up, interview, customer and general, plus your own
- Chat with one meetingAn "Ask anything" bar under the notes (⌘J or /) opens the chat during and after the meeting: recipes first, answers that stream in and cite the transcript time, "Say more", separate chats with their history, a fast or thorough model, a text file attached to the question and the question dictated by voice. Checked end to end with a test model; the quality of citations with real models is not measured yet
- Share notes by linkA read-only link with access for the workspace, for the people you pick or for anyone; the transcript only if you include it; change or revoke it anytime. No collaborator role yet
What is still missing
Capabilities Wispr Flow + Granola has that Voice lacks or only covers in part. Vote on the public board for the ones you need.
- Dictate into any text field with a shortcut partialHold Alt+Space (or the field button) in any zerosoft field and the clean text is inserted at the cursor. In any other app, the native desktop app (Tauri, Mac and Windows): hold Option + Space on Mac or Ctrl + Alt + Space on Windows (configurable), or tap it to start and stop, and the text is pasted where the cursor is, with the clipboard put back afterwards. A small HUD shows the level and a timer. Shortcuts made only of modifiers (Fn, Control + Option) are not available. Tested with the parts in Rust unit tests; the real paste into other apps and real microphone audio are checked by hand, not automatically
- More than 100 languages partial16 languages to pick in settings, plus automatic detection with whatever the configured provider supports (Whisper, Deepgram or AssemblyAI)
- Mac and Windows desktop apps partialA native app for Mac and Windows built with Tauri (source in desktop/): the system webview instead of a bundled Chromium, dictation in any app done in Rust (global shortcut, microphone, HUD, paste), a guided first run for the microphone and Accessibility, settings with language, cleanup, dictionary and history (the same ones as Voice on the web), menu bar menu, native notifications and signed updates. Not signed by Apple yet. Meeting notes with system audio are not in this version
- Record meetings without a bot partialNo bot: the browser asks to share a tab or the screen with its audio (Meet, Zoom or Teams on the web, or a native call app with the system audio where the browser offers it) and records it next to your mic; the meeting audio is asked by default. The Mac desktop app records the system audio natively with ScreenCaptureKit (macOS 13 or later, Screen & System Audio Recording permission). Tested with fake capture only, not yet with a real call; Windows desktop still goes through the browser share
- Chat across meetings partialQuestions across your meetings and the ones teammates shared with you (only what each share lets you see), with follow-ups, dates like "last week" and names of attendees. Answers stream in and link to each meeting and moment. Meetings are picked by words, names and dates, up to six of yours and three shared; there is no semantic search, so a question worded differently from what was said can miss a meeting
- SSO and SCIM missingWispr Flow Growth (SSO) and Enterprise (SCIM); Granola Enterprise
- iOS and Android missingWispr Flow keyboard on iPhone and Android; Granola apps for iOS, Android and Apple Watch
- Dictation for code editors missingWispr Flow: knows visible variable and function names in Cursor, Windsurf and VS Code, and @-tags files in Cursor and Windsurf chat
- Spell names from screen context missing
- Whisper mode missingWispr Flow: dictate at low volume in shared spaces
- Apply spoken self-corrections partialSpoken corrections ("Monday, no, I mean Tuesday", "scratch that") are applied by the AI prompt and by a rules fallback; the rules cover clear cases only and the AI behaviour has not been measured against real recordings yet
- Keep dictations without a connection partialIn the desktop app, a dictation that cannot reach the server stays on the computer and is sent again from the menu bar or the history. Transcription still needs the server: there is no on-device model
- Sync across devices partialDictionary, styles and history live on the server, so every browser and the desktop app (with your session) use the same ones. No phone app yet
- Speaker names partialMe and Them by audio source, separate speakers with Deepgram or AssemblyAI, and names set by hand; no names from the call
- In-person meetings and phone calls partialMicrophone mode records in-person meetings from the browser; no phone calls and no mobile app
- Follow-up email partialAI drafts the follow-up from the notes, addressed to the attendees, right from the bar under the notes; you edit it and send it from your connected Gmail or Outlook, copy it, or open it in Gmail or Outlook. Sending was checked end to end with a test Outlook mailbox only
- Calendar event and attendees on the note partialThe date chip on a note picks today's event from the connected calendar (the one happening now is chosen on its own for a new note), adds attendees from the workspace or the CRM, and groups them by company. Checked with a test Google Calendar only
- People and companies from meetings partialA meeting links to a CRM contact (by hand or from the Cal guest email); no automatic directory of people and companies
- Spaces and folders partialYour own folders with one level of subfolders and a "Shared with me" view with what the team shared; no custom spaces with their own members
- Zapier partialNo Zapier app; the voice.meeting.completed and voice.notes.ready webhooks work with Zapier "Catch Hook", and Flow covers the same automations
- API and webhooks partialAPI v1 with voice:read and voice:write (dictate, sessions, audio chunks, dictionary). The voice.meeting.completed and voice.notes.ready webhooks only reach webhooks created by the person who recorded; creating webhooks is for admins, so a member’s meetings do not reach any webhook yet
- Recording notice for attendees partialThe recorder reminds you to tell attendees; nothing is posted in the call
Price
Wispr Flow Pro costs $12 per user per month billed yearly ($15 monthly) and Granola Business costs $14 per user per month. Together that is $26: a team of ten pays $3,120 a year. List prices checked on October 25, 2026. Voice on your server costs $0 plus whatever your transcription provider charges.
$0 on your server (AGPLv3). On zerosoft Cloud: free plan, then usage. Seats are not billed.
Moving from Wispr Flow + Granola
What is missing compared to Wispr Flow + Granola?
iPhone and Android apps, signed downloads of the desktop app, whisper mode, dictation that recognizes code and names on screen, and SSO with SCIM. The parity table above lists every capability we measure.