Wispr Flow

Wispr Flow is an AI dictation layer designed to replace typing across applications rather than merely transcribe recordings. On supported desktop and mobile platforms, users speak naturally and Flow turns the speech into edited text: filler words are removed, punctuation and formatting are added, corrections spoken mid-sentence can be interpreted, and tone/layout can adapt to the text field and application context.¹²

The product is now genuinely cross-platform. Flow operates on Mac and Windows, uses a dedicated keyboard on iPhone, and has an Android app that overlays the user's existing keyboard. TechCrunch reported the Android launch in February 2026 and noted support for more than 100 languages, cross-app dictation and contextual formatting.³ Current Apple App Store material likewise describes a voice keyboard that works in apps such as Slack, Messages, email, WhatsApp, ChatGPT and Docs, with custom dictionary support, whispering and quick-capture notes that sync with desktop.²

Context awareness is a major reason Flow can produce polished output—and one of the product's most important trust considerations. Wispr documents that context can include the active app, text around the cursor, selected text, visible on-screen text, variable or file names, conversation history and, in some configurations, screenshots.⁴ That can materially improve names, formatting and intent, especially in coding or domain-specific writing, but it means Flow is closer to a context-aware input agent than a conventional microphone-to-text service.

Wispr provides controls around this behavior. Password fields are generally excluded, known banking/financial applications are skipped, Context Awareness can be disabled, and Enterprise administrators can disable it organization-wide.⁴ Wispr also offers Privacy Mode and cloud-storage controls so users can restrict retention and model-training use independently.⁵

The dictation product has continued to evolve technically. In August 2026, TechCrunch reported that Wispr introduced a proprietary speech-recognition model called Canto after several users complained about a recent quality dip. Wispr says Canto reduces error rates from around 30% to below 10%; that is a vendor claim reported by TechCrunch, not an independently reproduced benchmark.⁶ Reuters separately confirmed the Canto launch while reporting Wispr's latest financing and enterprise adoption.⁷

Independent hands-on evidence is generally positive. Computerworld's Android test described transcription as near-instant and unusually accurate, including formatting, long pauses, whispers and multilingual switching.⁸ WIRED tested Flow several times and found the output good and the setup polished, while questioning whether the benefit justifies a recurring subscription when free local alternatives exist.⁹ Computerworld also reported a Thumbtack pilot that expanded from a handful of users to more than 200 IT and engineering staff, with one executive describing Flow as part of his daily AI workflow.¹⁰

Wispr has also launched a meeting Notetaker, but that is a separate adjacent surface and is not treated as evidence of the canonical Flow dictation product's quality here. Features or privacy behavior that apply only to Notetaker are not silently transferred to dictation.¹¹

Flow currently has Free, Pro, Growth and Enterprise tiers. Wispr's current plan documentation lists Flow Pro at $15 per month or $144 per year and the Student plan at $7.50 per month or $72 per year. Growth is now self-serve from $23 per user/month or $18 per user/month billed annually; Enterprise remains sales-assisted with custom contract pricing.¹²¹³ Annual Pro pricing is roughly 20% lower than paying monthly for twelve months.

The Free tier is usable but capped. Current official documentation lists a desktop soft limit of 2,000 words per week and a 5,000-word hard limit; on iOS the soft/hard limits are 1,000 and 1,500 words respectively.¹² Pro removes the dictation cap and adds features such as Command Mode, priority support and early access. Growth adds organizational controls including SAML SSO, HIPAA/BAA options and usage reporting; Enterprise adds capabilities such as SCIM, audit logs, MDM support, invoicing/PO workflows, SLAs and dedicated account support.¹²¹³

Trial duration depends on route: the default Pro trial is two weeks, referrals can receive a month, students can receive three months and team trials are generally two weeks.¹² Apple's U.S. App Store independently lists Pro at $15 monthly and $143.99 annually, closely matching first-party pricing.²

Feature parity is not complete across platforms. Official plan documentation notes that Command Mode is unavailable on iOS, while Android still lacks some features including Dictionary, Snippets, Styles, Privacy Mode and manual language selection.¹² Buyers should therefore evaluate the device mix they actually use rather than treating the subscription as a perfectly identical experience everywhere.

The value question is unusually competitive. WIRED argues that Flow's $15 monthly/$144 annual subscription is meaningful because speech recognition and LLM post-processing are widely available in free or local tools.⁹ Google's free, offline AI dictation products and operating-system-level voice features increase that pressure. Flow therefore has to justify its price through superior cross-app consistency, context awareness, multilingual handling, workflow polish and reliability—not simply through the existence of AI transcription.

Wispr Flow now has a materially better evidence base than many young AI utilities. The U.S. Apple App Store shows 4.8/5 from about 14,000 ratings, making it a substantial user signal for the iPhone surface.² Reviews include strong praise for context-aware formatting and accessibility use cases, alongside complaints about mobile/iPad workflow friction and support responsiveness. This is large enough to be retained as a meaningful quantitative user signal, while still being treated as an iOS-specific surface rather than proof of identical quality on desktop or Android.

Other rating surfaces are much smaller. TrustRadius currently shows 10/10 from three reviews and confirms a $15 Pro plan, free version and free trial.¹⁴ The sample is too small for direct numeric weighting. PeerSpot shows growing category mindshare, but some comparison-page content appears inconsistent with Wispr Flow's actual dictation product—describing generic communication/analytics or marketplace behavior—so that material is excluded from product-quality conclusions despite being a current page.¹⁵

Editorial testing adds useful texture. WIRED found results good and the interface polished but questioned subscription value.⁹ Computerworld's Android test was much more enthusiastic, calling the transcription near-flawless in its test and highlighting long pauses, whispered input and multilingual switching.⁸ Tom's Guide published a positive workflow account of using Flow with Claude, but that article is now old enough to be treated as stale evidence for current product performance.¹⁶

Recent evidence also includes a caution. TechCrunch reported in August 2026 that users had complained about a dip in dictation quality in the preceding weeks; Wispr says Canto is intended to correct this and materially reduce errors.⁶ The complaint signal is important because it is current and directly concerns the core product. Until independent evidence confirms Canto's impact, the vendor's error-rate improvement should remain a claim rather than a settled performance fact.

Wispr's business maturity is increasing quickly. Reuters reported a $280 million Series B at a $2 billion valuation in August 2026, bringing total funding to $361 million, and said the product was used by more than 10,000 enterprises.⁷ Funding and enterprise count do not prove dictation quality, but they reduce the concern that Flow is an unsupported hobby-scale product. Computerworld's Thumbtack deployment provides one concrete enterprise-use example.¹⁰

Privacy requires more scrutiny than a normal keyboard because Flow can inspect surrounding context. Wispr says it never sells user data, allows users to opt out of model training and encrypts data in transit and at rest; its public security material states SOC 2 Type II and ISO 27001 compliance and HIPAA support.¹⁷ Dictation is cloud-processed rather than fully local.⁵ Wispr says third-party AI providers used for Flow Dictation operate under zero-data-retention agreements, and Privacy Mode can prevent dictation data from being retained or used for training.⁵

Those controls are meaningful, but the scope of context collection should be understood explicitly. The product can use visible text, text near the cursor, application/session information and other screen context; password fields and known financial apps receive special treatment, but no heuristic exclusion system should be assumed infallible.⁴ Organizations handling privileged, regulated or highly confidential material should test Privacy Mode, cloud-storage settings and enterprise controls before standardizing deployment.

Wispr Flow is strongest for people who produce a large volume of text across many applications. Executives, salespeople, recruiters, consultants, lawyers, writers and developers can dictate email, documents, messages and AI prompts without moving content into a separate transcription tool.²¹⁰

Developers and heavy AI users are a particularly good fit because context-aware dictation can handle technical vocabulary and Flow can be used directly inside coding assistants and chat interfaces. Computerworld's Thumbtack example shows engineers using it for prototyping, code exploration, technical strategy and analytics through AI tools.¹⁰

Accessibility is another compelling use case. App Store reviewers describe Flow as materially easier than typing when tremors or other physical limitations make keyboard input difficult.² Whisper support can also make voice input more socially practical in shared spaces.

Flow is less attractive for users who want fully offline/local transcription, who are highly uncomfortable with context-aware cloud processing, or whose typing workflow is already efficient enough that the subscription does not save meaningful time.⁹ Organizations with Android-heavy fleets should also verify feature gaps before rollout, because current Android functionality still trails desktop/iOS in some controls.¹²

Teams considering broad deployment should run a real-work pilot that measures correction rate, time saved, privacy-mode usability and failure rate on their own accents, domain vocabulary and noisy environments. The independent evidence is encouraging, but recent quality complaints demonstrate why current workflow testing matters more than old benchmark-style claims.