Comparatif
Vous cherchez une alternative à Wispr Flow ?
Voice dictation into any app on Mac, Windows, iPhone and Android, with a meeting Notetaker on Mac and Windows.
Chaque prix et chaque fonctionnalité ci-dessous a été vérifié en date du 2026-09-22 auprès de la source citée à côté. Les choses évoluent — signalez-nous tout élément obsolète.
Quand Wispr Flow est le meilleur choix
- Dictation that types straight into any app on four platforms. Ours runs in the browser tab and installable app, where you copy or share the result; our Mac client that types at the cursor is not distributed.
- Their own speech model with published English accuracy results. We use open Whisper models on your device and have published no accuracy numbers.
- Speaker labels carry real names rather than one unidentified speaker — we have no diarization at all.
- A mature product with context-aware formatting, voice commands, transforms, team dictionaries and enterprise admin that we have not built.
Nos différences
- All transcription runs on your machine and the audio is never uploaded. Wispr's documentation states that transcription happens in the cloud and that Notetaker requires cloud sync, so this is a difference of architecture rather than of settings.
- Every reader gets the whole meeting in their own language — transcript, summary, headings, decisions and action items — with the original kept beside each line. In Chrome and Edge on a computer that translation runs on the reader's own device, free.
- Dictate in any of ten languages and have it written out in another of the ten, with what you actually said kept alongside.
- No meeting cap on the free tier, because on-device transcription costs us nothing per meeting to serve.
- Action items push to Slack, Discord, Teams or any webhook URL, and the MCP server can write actions back. Wispr's MCP server is read-only.
- A disclosure step before every capture on every plan, with the consent record saved beside the note and exported with it.
Côte à côte
| Wispr Flow | Nous | |
|---|---|---|
| Platforms | Dictation on Mac, Windows, iPhone (as a keyboard) and Android (as a floating bubble). Notetaker records on Mac and, since 15 September 2026, Windows; iPhone can view upcoming meetings but not record.docs.wisprflow.ai (What is Flow?; Notetaker on Windows) and wisprflow.ai/whats-new · vérifié le 2026-09-22 | The web app and installable PWA, on desktop and phone browsers. A macOS client and an iOS/Android app exist in source but are not distributed. |
| Languages | Dictation: 100+ languages, detected once at the start of each dictation; mixing languages within one dictation is a documented weak case. Notetaker: 21 languages, auto-detected, with summary section titles and speaker labels kept in English.docs.wisprflow.ai, Use Flow with multiple languages; Notetaker supported languages · vérifié le 2026-09-22 | English, Spanish, Chinese, Hindi, French, German, Japanese, Portuguese, Italian and Arabic, for capture, dictation and reading, with the original always kept. Recognition of dense code-switching within a sentence is not measured. |
| Speech model | Canto, Wispr's own speech model, has served all users since 17 September 2026. Wispr publishes English accuracy results for it; no other language is reported.wisprflow.ai/canto and wisprflow.ai/whats-new · vérifié le 2026-09-22 | — |
| Speaker attribution | Speakers are labelled with real names drawn from the calendar, participant tiles in supported meeting apps and spoken cues, and can be reassigned by hand.docs.wisprflow.ai, meeting notes and the editor in Notetaker · vérifié le 2026-09-22 | — |
| Capture | No bot joins the call. Microphone and system audio are captured on the computer and uploaded for processing; a recording can run up to three hours.docs.wisprflow.ai, recording a meeting with Notetaker · vérifié le 2026-09-22 | — |
| Price | Free tier with weekly word limits; Pro $15/month or $12/month billed annually; Growth $23 per seat (dictation) or $33 per seat (with unlimited Notetaker) monthly for teams; Enterprise on request.wisprflow.ai/pricing and docs.wisprflow.ai, plans · vérifié le 2026-09-22 | An unlimited free tier, then Pro $15/month ($12/month billed annually) for the parts that need a server. No Team or Enterprise plan. The pricing page says whether checkout is open. |
| Processing | Transcription runs in Wispr's cloud. Notetaker requires cloud sync; there is no local-only mode.wisprflow.ai/data-controls and docs.wisprflow.ai, privacy mode and cloud sync · vérifié le 2026-09-22 | All transcription runs on your own device. Microphone and shared-tab audio are saved in recoverable on-device chunks and transcribed by Whisper running in your browser; the audio is never uploaded, to us or to anyone else. There is no cloud transcription option and no ASR vendor. Summaries, cross-meeting Ask and translation do call a hosted model, and the interface says so at the moment they run. |
| API & MCP | A read-only remote MCP server exposes meetings, notes, summaries, transcripts and Scratchpad, with one-click setup for Claude, ChatGPT and Gemini. Export is copy-to-Markdown.docs.wisprflow.ai, connect an MCP client · vérifié le 2026-09-22 | A local MCP server for coding agents that can read notes and write actions back, plus webhook connectors for Slack, Discord, Teams and any generic URL. |
| Consent | Flow can type a recording notice into the chat of Zoom, Teams, Slack huddles and Google Meet. Organisations can require a consent confirmation, stored as an audit record, and set their own notice text.docs.wisprflow.ai, consent for Notetaker · vérifié le 2026-09-22 | A disclosure step before capture, and the recorded consent state — who agreed, when, for which meeting — travels with the note. |
| Usage limits | Notetaker has weekly limits on the free and Pro tiers; no number is published.wisprflow.ai/pricing · vérifié le 2026-09-22 | No meeting cap on any tier, including free. Because transcription happens on your machine rather than on a server we pay for, volume costs us nothing to serve, so there is no cap to publish. A single recording is capped at 180 minutes, which is a technical limit of one browser capture session and applies on both plans. |
Ce que nous ne faisons pas
- We launched in September 2026. The web app and installable PWA are live and you can use them now; every product we compare against has years of production history and we have days. There is no independent track record to check yet.
- No diarization. No engine we use separates speakers, we do not store voiceprints as a matter of policy, and speaker names are typed by the user. Wispr pulls real names off the calendar invite; on this one they are simply better and it is the most visible thing we lack.
- No system-wide dictation you can install today. Dictate works inside the browser tab and installed app, and you copy or share the result; the macOS client that types where your cursor is exists in source but is not distributed. Wispr Flow types into any app on four platforms.
- No native app is distributed. The product is a web app and an installable PWA; the macOS scaffold compiles but ships to nobody. On-device transcription therefore needs a browser with WebGPU or WASM rather than a Mac, which is a wider net than it sounds, but it is not a native app.
- On-device transcription costs a one-time model download before the first meeting, and runs slower on a weak machine than a datacentre GPU would. That is the trade we are making for never uploading your audio; we state it rather than hiding it.
- Newer and smaller than every product on this page, with no track record to point at.
Essayez-le sur une réunion
Enregistrez votre prochain appel avec les deux outils et comparez les transcriptions. C’est la seule comparaison qui tranche vraiment.
Ouvrir l’application