Wispr Flow Release Notes

On Windows, meeting uploads that used to get stuck now complete on their own, so a recorded meeting makes its way into Flow without you having to chase it. If an upload still needs a nudge, retry works without restarting Flow.

Meeting recordings recover from save and upload errors

If saving or uploading hits a problem as you stop or resume a meeting, the recording now recovers instead of ending up with no audio — the situation behind meetings that never got a summary.

Speaker detection in Teams gallery view and Google Meet

Notetaker now tells you who is talking when you are in Microsoft Teams gallery view and when you are presenting in Google Meet.

An idle Teams window no longer triggers meeting pop-ups

On Windows, leaving Microsoft Teams open in the background no longer sets off meeting detection cards when there is no meeting going on.

Mention people in your meeting notes

Type @ in My Thoughts or Summary to open a people picker. It lists the people in the meeting first, then your contacts, and gives you the option to add a new contact, so you can also mention someone who wasn't in the meeting. A mention stays in that meeting's notes: it doesn't notify the person, share anything with them or add them as an attendee. Hover a mention to see that person's contact card, and click it if you need to point the mention at a different contact.

Hover a name in notes to see who it is

When a name comes up in your notes, hover over it to see that person's contact card: their photo, their name, and their job title and company when available, or their email address if neither is known. A LinkedIn button appears when a link is on file. The card shows for names that are linked to a contact. If a name isn't linked yet, click it to assign one.

Flow warns you when your mic picks up no sound

If your microphone registers no sound when you start dictating, Flow now warns you right away, instead of leaving you to discover afterward that nothing was typed.

Flow offers to add a language you speak

Speak a language Flow isn't set to and you now get a notice offering to add it, so you don't have to go looking for the setting yourself.

The main Flow window reopens after you update

After you click Update, the main Flow window now opens again instead of staying closed.

Original source

Available on: Mac (macOS 13 or later) and Windows, where Notetaker is enabled. Notetaker is turned off for accounts with a signed BAA where HIPAA cloud-sync restrictions apply.

Notetaker transcribes meetings in 22 languages and detects the spoken language automatically, so there is nothing to select before a call.

Note: Settings → General → Languages applies to dictation only; adding or removing a language there does not change how Notetaker transcribes a meeting. For dictation, see Use Flow with multiple languages.

Supported languages

Meeting transcription supports Arabic, Bengali, Chinese, Czech, Danish, Dutch, English, French, German, Hindi, Hinglish, Italian, Japanese, Korean, Marathi, Polish, Portuguese, Russian, Spanish, Swedish, Turkish, and Urdu. Cantonese is handled as Chinese, and Hinglish — mixed Hindi and English speech — is detected automatically like any other supported language.

Meetings in other languages still record and transcribe, but transcripts may contain more errors. You cannot pin or override the language for a meeting, and saved account language preferences can influence meeting transcription.

Where post-meeting transcript refinement is enabled, language-related refinements happen only after the meeting, so the final transcript can differ from the live one.

Translate a transcript

Note: The translate button is dimmed when translation is unavailable, such as when you are offline.

1. Open the meeting in Notepad.
2. Select the Transcript tab.
3. Select the translate button, whose tooltip reads Translate transcript with a Beta pill.
4. Choose a language from the picker and wait for the change to finish; a second choice made during a change has no effect.

Changing the language during a live meeting does not interrupt capture: your speech keeps recording, and anything said while the change is in progress arrives translated. Translation applies to each complete line as shown. To return to the original, pick the meeting's own language.

If a translation cannot be applied

The unsupported-language warning

When a sustained stretch of recent speech is in an unsupported language, Notetaker plays an alert sound and shows a warning titled "This sounds like a language we don't support yet". Choose Learn more to open this article in your browser, or Got it to dismiss it.

• What triggers it: Only a sustained stretch of recent speech, not the whole meeting — a few stray words in another language during an English call will not trigger it, and very short utterances are ignored.
• Frequency: At most one warning per meeting. Once you have been warned about a language on a device, that device stays quiet for that language; a different unsupported language can trigger a warning in a later meeting.
• Audio sources: Your microphone and the meeting's audio are considered separately; either can trigger the warning.

Summaries and speaker labels

• Summary language follows the audio: The TL;DR, key points, next steps, decisions, and action items are written in the language predominantly spoken. Your notes and calendar event details do not override it.
• Some elements stay in English: Top-level section titles ("Flow Summary", "Action Items") and speaker labels. An English "Next Steps" or "Decisions Made" heading is never added to a summary in another language; if no heading is available in the meeting's language, the bullets appear without a heading.
• Very short or silent meetings: These receive an English placeholder summary.

Still need help?

Reach out to support if a meeting in a supported language transcribes into a different language, or produces a transcript with widespread errors.

1. Click the ? icon in the Flow desktop app.
2. Select Talk to support.

Include the meeting language, your platform and device, what you tried, and a short example of what was said versus what Notetaker transcribed.

Original source

• Model training stays off for Growth and Enterprise plans: If your organization is on a Growth or Enterprise plan, model training is now off for every member. It’s tied to your plan rather than an individual setting, so members can’t turn it back on, and sensitive fields are removed from members’ history uploads on our servers. You don’t need to do anything. Find it: Settings → Privacy in the desktop app shows this choice as controlled by your organization.
• See who accesses shared notes and transcripts: Enterprise admins can now review note and transcript access by people other than the owner, including anonymous share-link views and reads through MCP tools. The Note access category shows the viewer—or “Anonymous” for public links—and the access source, helping you understand how meeting information is being used. Find it: Admin Portal → Audit Logs → Events → Note access.
• Audit Logs capture more membership and domain changes: Enterprise admins can now see role changes, invitations, revoked invitations, and domain additions and removals in Audit Logs. Member added and removed events are also recorded across more ways people join or leave, including SCIM provisioning, auto-add, invite links, join requests and disbanded organizations.

Original source

When Flow detects dictation in a language outside your selected languages, you’ll now see a “Speaking a different language?” card. Click Open language settings to add the languages you use and help Flow recognize your dictation correctly. Available on Mac and Windows.

Scratchpad sync works independently of dictation storage

Turning off dictation Cloud Storage no longer stops your Scratchpad notes from syncing across devices. Notes written while syncing was blocked will upload on your next sync. Your organization’s Scratchpad restrictions still apply.

Original source

Flow now runs on Canto, the first speech model we've built ourselves. It's now rolled out for everyone, with nothing to switch on.

Most speech models are trained and tested on clean audio recorded in quiet rooms with good microphones. That isn't where people use Flow. You're on a train, in an open office, walking between meetings, or talking quietly so you don't disturb anyone. Canto was built and measured against those conditions instead.

On our evaluation of real-world English Flow recordings, 9.8 hours from more than 2,300 speakers, Canto had the lowest word error rate of every model we tested, including models from Google, OpenAI, AssemblyAI and Deepgram.

On a separate set built to stress-test the hardest audio, with background conversation, music, traffic, wind and whispered speech, Canto was the most accurate of the real-time models we tested.

We're already training its successor at more than ten times the scale, with better recognition in difficult conditions and broader language coverage ahead.

Read the full research post here.

Original source

Speech recognition models have become remarkably good at transcribing clean audio recorded under controlled conditions. But real dictation rarely happens under those conditions. Millions of people use Wispr Flow to message friends, write emails, code, and work through ideas at their desks, between meetings, during commutes, and in busy offices. They speak through laptop microphones, earbuds, and headsets, often with other voices, music, or traffic in the background.

Today, the Wispr Advanced Interfaces Lab is introducing Canto, our latest speech model for real-time dictation. On an evaluation of real-world dictations, Canto achieved the lowest word error rate among all the models we tested. We compared Canto with models from Google, OpenAI, AssemblyAI, and Deepgram.

Canto is the first model in a broader research and development program at Wispr Advanced Interfaces Lab. In this post, we share how it performs, how we trained it to handle challenging real-world conditions, and the research already shaping what comes next.

“Today, the Wispr Advanced Interfaces Lab is introducing
Canto, our latest speech model for real-time dictation.”

Ariya Rastrow

CSO, Wispr Flow

Evaluating Canto in real-world conditions

To test Canto in real-world conditions, we created an evaluation set composed of 10 hours of English-language Wispr Flow dictations from more than 2,300 unique speakers, randomly sampled across applications and use cases. We were careful to enforce a strict separation between speakers represented in the train and test sets to avoid overfitting on speaker characteristics. Every sample came from a user who opted in to Wispr’s data-sharing setting, which allows their data to be used anonymously to evaluate and improve our models. More information about this setting is available in our data-controls documentation.

Canto achieved the lowest Word Error Rate (WER) of the models in our comparison. WER measures word substitutions, omissions, and insertions relative to a human-transcribed reference (lower is better).

Performance under the most challenging conditions

The model performed well on randomly sampled data, but we were also interested in studying its performance in the most challenging situations. We built a separate, 3-hour challenge evaluation set that featured those conditions most likely to cause dictation to fail.

The set includes audio affected by nearby speech, music, traffic, wind, low recording volume, and whispered or far-field speech. It also includes short dictations that give a model very little surrounding context and language to help resolve ambiguity. On the full challenge set, Canto ranked second behind Gemini 3.1 Pro, a much larger frontier-size multi-modal model that is not suitable for real-time low-latency applications. Among the real-time transcription models we evaluated, Canto achieved the lowest WER.

We further examined this dataset to understand how the models behave differently. Gemini 3.1 Pro achieved the lowest WER on noisy audio. Canto tied for the lowest WER on low-volume speech and short dictations. Short dictations produced the highest error rates across the comparison. A single mistake has a larger effect on WER when an utterance contains only a few words, and the models also have less linguistic context available to resolve ambiguity.

Comparing performance on public benchmarks

We also evaluated Canto on three public English datasets: LibriSpeech, FLEURS, and Common Voice. Canto tied for the lowest WER on LibriSpeech. It remained competitive on FLEURS and Common Voice, although it did not lead either evaluation.

These public datasets provide useful, reproducible comparisons, but they largely contain read speech. LibriSpeech is drawn from audiobooks, while FLEURS and Common Voice consist primarily of people reading prepared sentences. They capture a different distribution from everyday “in-the-wild” dictation, where people speak spontaneously, pause, revise their thoughts, and often provide very little linguistic context. The contrast helps explain why we evaluate Canto on both public datasets and real Wispr usage.

Post-training Canto with Supervised Fine-Tuning and Reinforcement Learning

Canto starts from a model pretrained on millions of hours of speech and text. We then train Canto in two stages. First, we show it audio paired with reference transcripts. The model learns to predict the words in each transcript, one step at a time. This stage, called supervised fine-tuning, teaches it how to perform the transcription task. Next, we train it to compare the quality of complete transcripts. For the same audio, the model generates several possible transcriptions. We score each one against a reference, then use those scores to make better transcriptions more likely in future training. This is known as reinforcement learning, or RL.

The distinction is in how the model receives feedback. Supervised fine-tuning provides the expected words at each step. Reinforcement learning evaluates the completed transcription (generated by the model). That lets us train around the outcomes we care about, such as reducing recognition errors under difficult recording conditions. Our approach uses Group Relative Policy Optimization, or GRPO. The central idea is simple: compare the candidate transcripts within each group and learn from their relative scores. For each audio example, we sample several candidate transcripts from the model being trained. We call these candidates rollouts. Each receives a reward: a numerical score measuring how well it satisfies the training objective.

GRPO compares each candidate’s reward with the rewards of the other candidates for the same audio. The resulting relative score, called an advantage, indicates whether that candidate performed better or worse than its group. These advantages guide the training update.

Sequence-level training has a long history in speech recognition. Researchers have previously optimized ASR systems using minimum word error rate training and policy-gradient methods. GRPO was introduced more recently in DeepSeekMath, and direct applications to autoregressive speech recognition have only begun to appear in recent work on GRPO for ASR and speech-domain adaptation.

“We built infrastructure that can generate and score speech rollouts at scale, allowing us to construct training environments around specific behaviors and failure modes.”

For Wispr, the value of RL is not just a lower aggregate error rate. We built infrastructure that can generate and score speech rollouts at scale, allowing us to construct training environments around specific behaviors and failure modes. That system now supports experiments in contextual speech recognition, personalization, diarization, and difficult audio conditions. The research described below extends beyond the training used for the released Canto model and is already informing the next generation of our models.

Learning from corrections

The usage of names and vocabulary changes faster than speech models can be retrained. A word such as “Claude” might once have been an uncommon alternative to “cloud.” As its usage changes, a practical dictation model needs a way to learn the distinction.

For users who have opted into data sharing, corrections can provide a valuable signal for improving recognition. But not every edit points to a transcription error. People also change formatting, rewrite sentences, or simply change their minds.

A hypothetical example below illustrates this problem. The model transcribed “Claude” as “cloud,” which the person corrected. They also changed their mind and rewrote the end of the sentence. Only the first change represents a recognition error.

We use signals from the audio and the edit itself to identify corrections that are most likely to represent recognition errors. These include forced-alignment confidence, which measures how well candidate words match the audio, and the location and shape of the edit. We then graft only the likely correction into the original transcript, leaving the rest unchanged.

The resulting transcript becomes the reference used to score GRPO rollouts. A hypothesis receives a higher reward when it makes the intended correction. Because GRPO centers the advantages within each group, words shared by higher-reward and lower-reward hypotheses receive offsetting signals. This does not provide perfect token-level credit assignment, but it makes the strongest contrast more local to the decision we care about. Grafting turns a noisy document-level edit into a more focused sequence-level training signal while preserving the on-policy nature of GRPO.

Learning to use context selectively

At runtime, Canto is provided with specialized vocabulary from a person’s dictionary. This helps the model recover names and rare terms that may be difficult to identify from audio alone. Providing phrases to guide recognition is a longstanding area of contextual ASR research. The difficult part is deciding how strongly the model should trust that context.

Suppose “Barry” appears in someone’s dictionary and they dictate, “I want to make a berry salad.” In poor recording conditions, both words may be acoustically plausible. An overeager model might choose “Barry” because it appears in the dictionary, even though “berry” is what the person said.

To measure this behavior, we started from an SFT checkpoint and trained the model using RL while providing truthful context on 10% of training examples. We then tested the model with synthetic distractors that varied in phonetic similarity to the correct word. A distractor flip occurs when adding one of these false suggestions causes the model to adopt it.

After RL training, the model became more responsive to context, including when that context was wrong. Among the most phonetically similar terms, it adopted the distractor nearly five times as often as the initial SFT model. The training had made the context more useful, but the model sometimes trusted it too readily.

To separate context adoption from context discrimination, we trained three runs from the same SFT checkpoint. In always true, the supplied context contained only the correct term. In distractors + true, the correct term appeared alongside phonetically similar alternatives. In distractors only, selected examples contained plausible near-misses instead of the correct term, while the reward continued to favor the transcript supported by the audio.

At every checkpoint, we ran two probes. The vertical axis measures adoption of correct context, while the horizontal axis measures how often false context changes the transcript. The ideal trajectory moves upward without moving to the right.

The trajectories suggest that each mixture teaches a different relationship to context. With always true, the model can learn a simple shortcut: when a supplied term resembles the audio, use it. This improves adoption of correct terms, but also increases false-context following. With distractors + true, context becomes a selection problem. The correct answer is present, but it must be distinguished from plausible alternatives. The model is rewarded for choosing the candidate best supported by the audio, not merely for copying from the list. With distractors only, context becomes adversarial. Blindly adopting any supplied term now loses reward, so success requires resolving the conflict in favor of the audio. This pushed false-context following further left while preserving a similar level of correct-context adoption.

These experiments do not yet define a final context policy, but they show that the tradeoff itself is trainable. The goal is not simply to make the model follow context, but to also teach the model when context deserves to be followed.

What comes next

Canto is the first model in a larger family. We are already training its successor at more than ten times Canto’s scale, with the goal of improving recognition under difficult audio conditions, improving multi-speaker recognition (for our recently launched notetaker product), making better use of contextual vocabulary, and reaching the same standard across more languages. The post-training approach described here will also let us introduce more specialized rewards and harder training environments as the models grow.

One focus is a unified architecture for transcription and diarization. Meeting-transcription systems often recognize words and identify speakers in separate stages, allowing errors in one stage to compound errors in the other. A joint model can reason about what was said, who said it, and when the speaker changed at the same time. Speaker structure can also provide useful evidence in noisy settings, helping distinguish the conversation being transcribed from unrelated speech in the background.

Over time, our speech models will need to do more than recognize words. They will need to use context selectively, understand the application in which someone is speaking, and connect what was said with what the person is trying to accomplish. Canto gives us the speech model, training approach, and evaluation framework from which to pursue that broader interface.

These remain open research problems. If you work on speech recognition, reinforcement learning, diarization, multilingual modeling, or multimodal interfaces, the Wispr Advanced Interfaces Lab is hiring.

Original source

Capture the details, then work from them

Notetaker is now available on Windows, alongside Mac. Here’s what it does and what we’ve added since the Mac launch.

Notetaker detects meetings across the platforms you already use and offers to take notes. It uses your personal dictionary to help capture names, acronyms, and technical terms correctly so you can:

• Catch up during a call: Ask “What did I miss?” for a recap of the discussion so far.
• Find answers across meetings: Ask “What follow-ups do I owe people from my calls this week?” or “Based on the calls I’ve led this month, how could I ask better questions?” Answers include citations back to the meetings.
• Draft from your notes: Turn conversations into follow-up emails, project updates, or briefs directly in Notetaker.

Find it: Open the Notetaker tab in Wispr Flow.

Connect your meetings to your other work

Notetaker just added one-click setup for Gemini, alongside connections to Claude and ChatGPT. You can also connect to any other MCP-compatible AI tool.

Once connected, your AI tool can grab your meeting notes and transcripts along with any other files and tools you’ve connected there. You’ll be able to dictate complex prompts:

• For sales: “Compare the objections from my discovery calls this quarter with our objection-handling doc. Where are we not answering well?”
• For founders: “Draft this month’s investor update from the board call, my team syncs, and the metrics in this sheet.”

And your AI tool can pull the meeting context directly from Notetaker, no copy + paste required.

The MCP connection only grants access to meeting transcripts, not your dictations.

Learn more about using AI tools/MCP with Wispr Notetaker.

More updates since the Mac launch

• Meeting transcription in 21 languages: Including Spanish, French, German, Chinese, Japanese, Hindi, and Arabic. Summaries and other features can also work in other languages, but the experience may be unreliable.
  See all supported languages.
• Outlook calendar: Connect Outlook to see upcoming meetings in Notetaker and bring in calendar context. Google Calendar is also supported.
• Import from Otter: Bring your previous transcripts into Notetaker with speaker names preserved, so you can search old and new meetings together. Otter transcript imports require a paid Otter plan. Granola imports are also supported.

Controls for your organization

Admins can manage Notetaker access, sharing, and privacy restrictions across their organization.

Notetaker captures audio from your computer, so no bot joins the meeting. Always inform participants before recording. Organizations can also require users to confirm they’ve obtained consent before recording starts.

Find it: Organization settings in the admin portal.

See Notetaker admin controls.

Original source

Available on: Mac and Windows desktop app. Requires a signed-in Wispr Flow account.

Bring your meeting history from Granola or Otter.ai into Notetaker. Imported meetings sit alongside the ones you record, keeping their original dates, summaries, notes, and a badge showing where they came from.

Before you start

• Wispr Flow open and signed in on Mac or Windows.
• An export file from Granola or Otter. Both platforms email the export and can take a few hours, so request it first.

PLATFORM STATUS FILE YOU DROP IN Granola Available CSV export, or the zip containing it Otter.ai Available Bulk export zip, or its unzipped folder Fathom Coming soon Not yet available

How to import

1. Go to Settings → Data and click Import. On Mac, a Granola "Import your meeting notes" banner in the Notetaker hub opens the same flow.
2. Choose your platform on the "Where would you like to upload from?" step.
3. Request your export using the steps below for your platform.
4. Reopen Settings → Data when the export email arrives and drag the file onto the drop zone, or click to browse.
5. Wait for "Upload complete!" and click Done. Your meetings are now in Notetaker.

Note: Once the import starts, closing the modal or the window leaves it running in the background.

Exporting from Granola

Exporting from Otter

What you'll see after importing

• Source badge: each imported meeting carries a Granola or Otter logo on its avatar.
• Original dates: meetings sort into your history on the day they happened, not the day you imported them.
• Summaries and notes: both come across. Otter summaries appear under a "Summary (Otter AI)" heading.
• Speaker names: Otter transcripts keep them; Granola imports arrive without them.
• Meetings first, transcripts after: meetings appear when the import finishes, then transcripts sync in the background. A transcript that is not there yet is still on its way.

FAQs

Can I import from Granola and Otter at the same time?

Does importing bring over my meeting audio?

Will importing twice create duplicates?

I quit Flow in the middle of an import. What now?

Some of my meetings were skipped. Why?

My selected file was cleared before I imported it.

Error messages explained

"That isn't a Granola export" or "That doesn't look like an Otter export"

"That file's columns don't match a Granola export"

Size and volume errors

"No transcripts could be read from this export" or "No meetings could be read from this file"

"That file's text couldn't be read" or "couldn't be opened"

"Sign in to Wispr Flow to import your meetings"

"Import paused. It will resume automatically."

Limitations and notes

• Importing works in the desktop app only, not on iOS or Android.
• The Granola import banner appears on Mac only. On Windows, start from Settings → Data.
• One import runs at a time, across both platforms.
• Imported meetings have no audio, no calendar link, and no attendee list.
• Granola exports transcripts as unlabeled text, so they arrive without speaker names.
• An existing transcript in Flow is never overwritten by an import.
• Transcripts sync to the cloud only if Dictation cloud storage is on; with it off, imported meetings stay on your device.
• Imported meetings are not live recordings and are not re-processed by Flow.

Still need help?

Reach out to our support team if:

• You see an error not listed above and the import won't proceed.
• Imported meetings are missing summaries or speaker names you expected.
• Transcripts stay pending far longer than your import size would explain.

Include your operating system, which platform you imported from, the size of your export, and the exact error message.

Original source

Available on: Mac. Requires Notetaker access.

Ask Claude or ChatGPT about last week's meeting without pasting the transcript. Connecting Wispr Flow gives any MCP-compatible client read-only access to your Notetaker meeting transcripts, notes, and calendar events. Your dictations are never exposed.

How to connect

Claude or ChatGPT

All other apps

Note: Where you paste an MCP server URL differs by client — check your client's documentation.

What connected clients can access

• Meetings: Search and read summaries, notes, transcripts, and the attendee list.
• Notes: Search and read your notes, and resolve a shared-note link you already have access to.
• Calendar: Search and read events, resolve a pasted calendar or conference link to the matching event, and move from an event to the recording it produced.

FAQs

Do I need Wispr Flow running locally for MCP to work?
Do I need to install anything in Claude Desktop?
Which MCP clients are supported?
How do I disconnect a client?
Why did my MCP client say a transcript is deleted?

Limitations and notes

• MCP settings are available on Mac. Windows availability is separately controlled and may not be available in all cases.
• Enterprise admins can enable or disable MCP for the whole organization. When it's off, the MCP page and Connectors MCP banner are hidden, or connectors show "Connected, but unavailable because your organization turned off this connector."
• The connection is read-only: connected clients cannot create, edit, or delete meetings or notes.
• Only Notetaker meeting transcripts, notes, and calendar events are exposed — not your dictations.
• Each new client connection requires browser sign-in to authorize, which opens in your default browser.
• Transcripts deleted under a data retention policy are not accessible through MCP.

Original source

You can now bring your Otter.ai meeting transcripts straight into Wispr Flow. Just drop in your Otter export and your meetings appear alongside everything else, with speaker names preserved instead of collapsed into a single anonymous voice. Imports run in the background, so you can keep working while your history fills in.

Find it: Settings > Notetaker > Import, or the import picker where you'll see the Otter tile alongside Granola.

[Help Center: Import your Otter transcripts into Wispr Flow Notetaker]

For Teams & Enterprise

• Assign teams at invite time: Admins can now pick which team a new member joins right from the invite form, for single invites or bulk batches, so new hires land in the right place from day one.
• Org-wide and department-wide dictionary words: You can now choose to share dictionary entries to either your whole organization or your department.

[Help Center: Manage Users and Billing]

Quality-of-Life Improvements

• Updates wait until you're done: Flow no longer restarts itself mid-task. It detects when you're actively working and holds pending updates until you step away.

Original source

Admins can now turn Notetaker on or off for their entire organization from the admin portal with a single toggle, enforced across every member's account. There's a confirmation step before disabling, so it never happens by accident.

New self-serve Growth plans:

Teams can now buy a new Growth plan straight from the admin portal, choosing dictation-only or bundling in unlimited Notetaker, with monthly or annual billing shown per currency. Upgrade your whole team in a couple of clicks with most of the features of our Enterprise plan, no sales call required.

Original source

Signing in with Google on Android now happens in a secure in-app browser tab that returns you straight to Wispr Flow. No more landing on a random browser tab, an old page, or the Wispr Flow website after picking your account. If your sign-in gets interrupted, a resume prompt lets you pick up right where you left off.

Find it: automatic, the sign-in screen the first time you open Flow or sign back in.

Help Center: Login Issues with Wispr Flow

Reports That Never Get Lost

Issue reports, feedback, and transcription reports now keep sending even if you leave the screen. You'll get a confirmation wherever you are in the app, and the form clearly shows when it's busy: a spinner, a dimmed Send button, and a short note explaining why it's temporarily locked.

Find it: Dashboard menu > Report an issue or Share feedback, and Home > any transcript > Report.

Original source

For Teams & Enterprise

• Daily teammates added digest: Admins now get a daily email summarizing every teammate who joined the workspace in the last 24 hours, whether they came in through domain auto-add, SCIM, or an invite link, so seat growth (and the bill) never comes as a surprise. Admins who don't want it can turn it off from the admin portal.
• Team trials for every joiner: Anyone joining a team for the first time now receives a 14-day trial, even if the team is already on a paid plan. New team members will not be billed until the first 2 weeks are complete.

Help Center: FAQs for Flow Pro, Team, and Enterprise plans

Shortcuts & Controls

• A clear "Done" button in Shortcuts: The Shortcuts dialog now has a primary Done button, so you always know when you're finished. Your shortcuts still save automatically the moment you set them, Done just closes the dialog.
• Hands-free made simpler: Double-tapping your Push-to-talk hotkey has always toggled Hands-free mode, and now the Shortcuts editor shows you a live hint built from your own keys so you don't have to guess. The dedicated Hands-free shortcut is also fully optional now: delete it and it stays gone.

Help Center: Use Flow hands-free

A Smoother Windows Experience

• A quieter startup: Flow now starts quietly in the background when it launches at login, instead of popping the hub window open. Your tray icon and Flow bar are ready immediately, and a restart you trigger yourself still brings the hub back.

Quality-of-Life Improvements

• Straight back to Flow after connecting apps: Connecting Notion, Google Calendar, or any other integration now brings you right back into Flow once the connection succeeds, with a redesigned confirmation page that names exactly what you just connected.
• A fresh look at sign-in: The desktop login screen has a redesign, with new full-motion slides showing what Flow can do while you sign in.
• Confirm before exiting the Notetaker tutorial: Closing the Notetaker guided tour now asks you to confirm instead of quitting instantly, so an accidental click or stray Esc doesn't wipe out your progress.

Flow now lets you pick virtual and routing audio devices, like Krisp, NVIDIA Broadcast, BlackHole, or VDI redirects, as your dictation mic. If you run your voice through a noise-suppression tool or work in a virtual desktop session, you can finally point Flow at exactly the mic you want.

Virtual inputs live behind a "Show other devices" expander, so nothing changes if you don't use them, and Flow won't auto-select a loopback device on you. Automatic mic selection stays safely on your real, physical microphones.

Find it: Settings > Microphone, or the mic switcher on the Flow bar.

Help Center: Using Wispr Flow Discreetly (Microphone Guide)

Original source

Available on: Mac, Windows, iOS, Android

Dictating in an open office, a coffee shop, or on a packed train? Flow understands whispers as accurately as normal speech — the closer your mic sits to your mouth, the quieter you can speak.

When to use it

Use discreet dictation when you want to:

• Dictate in a shared office without disturbing coworkers
• Use Flow in public spaces like coffee shops or airports
• Keep voice notes private in open environments
• Capture a thought on the go without taking your phone out of your pocket

How it works in Flow

Overview

Mac and Windows offer in-app microphone selection and a Microphone Quiz. iOS lets you dictate from the in-app mic button, the lock screen widget, a Control Center toggle, or a Siri command. On Android, the Flow Bubble appears above your keyboard in any text field.

Key behaviors by platform

Why earbuds and AirPods fall short

Earbuds and AirPods work fine at normal speaking volume but struggle with whispers. If you mostly work around others, an external microphone gives the best results.

• The mic is too far from your mouth. Earbud microphones sit on the cable or in the housing — typically 6 to 8 inches from your lips, where a whisper barely registers above ambient noise.
• Bluetooth adds its own problems. Wireless earbuds like AirPods and Galaxy Buds compress audio and add delay that degrades whisper accuracy; AirPods may be slow to activate the mic, causing missed words at the start of a session.
• Wired earbuds are only slightly better. No compression issue, but the same mic-distance problem.

Any microphone within a couple of inches of your mouth dramatically improves whisper dictation.

Microphone options

• Clip-on lav mics — best for on the go
• Attachable boom mics — upgrade your headphones
• Podcast mics — best desk audio quality
• Gooseneck mics — hands-free desk setup

Best practices

• Position the mic close. The closer it is to your mouth, the quieter you can speak.
• Enunciate clearly. Many people mumble when they whisper.
• Select your mic manually (Mac and Windows). Choose your mic in Flow's mic picker instead of Auto-detect. If the volume bars move when you talk, your setup is working.
• Set your system default (Mac and Windows). Make sure your OS Sound settings and Flow agree on which mic to use — the most common source of Auto-detect issues.
• Restart Flow after switching devices. Quit and relaunch Flow after plugging in a headset or swapping from Bluetooth to wired.
• Disable virtual microphones. Apps like NVIDIA Broadcast, VoiceMeeter, and Teams Audio can intercept your input and cause unexpected behavior.
• Stay consistent. Use the same mic regularly so results stay predictable.
• Use Siri on iOS. Say "Dictate a Flow note" to dictate hands-free without taking your phone out of your pocket.

FAQs

I already have a mic — will it work?
Can I use my laptop's built-in mic for whispering?
Can I dictate with my Mac lid closed?
What about AirPods?
What if I already have headphones I like?
Flow isn't appearing under Microphone in iOS Settings — how do I fix this?
How do I dictate a Scratch Pad entry discreetly on iOS without opening the app?
How do I use Flow discreetly on Android?

Limitations and notes

• Discreet dictation is available on Mac, Windows, iOS, and Android.
• The built-in Mac microphone is unavailable in clamshell mode — an external mic is required.
• The lock screen widget, Control Center toggle, and Siri shortcuts are iOS only.
• On iOS, Flow has no in-app microphone selection — it always uses the system default input, so mic choice depends on which device is connected or selected at the OS level. If you deny microphone access, enable it later in the iOS system Settings for Wispr Flow.
• On Android, if you deny microphone access, go to Settings > Apps > Wispr Flow > Permissions and enable Microphone manually.
• On Android, the Flow Bubble requires accessibility service and "display over other apps" permissions. Some manufacturers require additional setup — see Setup Wispr Flow on Android.
• Scratch Pad is available on Mac, Windows, and iOS. Android is not yet supported.
• Dictation sounds are always on and cannot be disabled on Android. On desktop they can be turned off in Settings → System → Sound; on iOS, in Settings → Audio → "Interaction Sounds."

Original source