# Flow Keyboard -- Full Content for AI Ingestion > This file contains the complete content of the Flow Keyboard website inlined in a single document, so AI models can ingest everything in one pass without following links. --- ## Overview Flow Keyboard is an AI voice to text app publicly available on macOS, iOS, Windows, and Android — a smarter alternative to dictation and traditional voice typing. Unlike regular speech to text that transcribes word-for-word, Flow intelligently refines your words — removing filler words, fixing grammar, adding punctuation, and restructuring your thoughts into clear, polished writing. Speaking is 4x faster than typing. A paragraph that takes 30 minutes to write takes just 2 minutes to speak. Flow ensures the output is clean and ready to send. 85% of the time, you can send Flow's output without making a single edit (zero-edit rate). **Publisher:** Snowline LLC (service@snowline.ai) **Platforms:** macOS (download from flowkeyboard.com), iOS (App Store: https://apps.apple.com/app/id6758776986), Windows 10 or later on x64 PCs (download from https://flowkeyboard.com/download.html?platform=windows), and Android (website download: https://flowkeyboard.com/download.html?platform=android; Google Play availability is shown on that page). iPad can use the iOS app when the on-screen third-party keyboard is visible. Apple Vision Pro is not supported as a system keyboard because visionOS does not load third-party keyboard extensions. **Website:** https://flowkeyboard.com/ **Price:** Free plan for registered users: 5,000 words/week. 7-day free trial available. Plus: $3/month or $30/year (12,000 words/week). Pro: $12/month or $100/year (unlimited, subject to fair use). --- ## Current product capabilities For platform-specific versions and release dates, use the [public release notes](https://flowkeyboard.com/releases.html). For current installers and store availability, use the [download page](https://flowkeyboard.com/download.html). These are maintained separately by platform. - Everyday dictation reaches the input field sooner, reducing waiting. Processing time varies with audio, network conditions and the task; translation and voice editing still require their transformed result. - iOS offers swipe quick deletion in bounded batches before the cursor, not a guaranteed whole-field clear. Android also supports swipe clearing and undo after clearing in the typing keyboard. - On macOS 1.2.0 with default shortcuts, selected text plus hold-Fn invokes voice editing; Fn + Space remains hands-free dictation and Fn + Shift remains translation, even with text selected. - Desktop shortcuts include optional middle-click hands-free input and a configurable cancel key. Enable middle-click in Settings before using it. On macOS, the Dock icon can also be hidden. - On iOS and macOS, system preferred languages inform speech recognition. This differs from output variants (such as Simplified/Traditional Chinese or American/British English) and the target language selected for translation. - With VoiceOver enabled, iPhone Quick Access / Floating Window recording can reduce spoken-interface audio in the recording. It does not mute VoiceOver. Other media may become quieter or pause; resumption depends on the player. Android has an opt-in setting to pause other audio during recording; device and player behavior varies. - See [support and FAQ](https://flowkeyboard.com/support.html) and the [desktop shortcut guide](https://flowkeyboard.com/keyboard-shortcuts.html) for usage and limitations. Flow Rail remains the compact mobile voice interface; features are not identical across platforms. ## App and Task Guides — Summary and Links Traditional Chinese directory: https://flowkeyboard.com/zh-Hant/use-cases/ Japanese directory: https://flowkeyboard.com/ja/use-cases/ Korean directory: https://flowkeyboard.com/ko/use-cases/ Traditional Chinese guide: https://flowkeyboard.com/zh-Hant/use-cases/codex-voice-input.html Japanese guide: https://flowkeyboard.com/ja/use-cases/codex-voice-input.html Korean guide: https://flowkeyboard.com/ko/use-cases/codex-voice-input.html The localized guides cover Codex requests and feedback, personal dictionaries for work vocabulary, recording continuity, and an example of organizing a long message into paragraphs and lists. The 180 characters/minute reference is for Chinese dictation. Use-case directories: https://flowkeyboard.com/use-cases/ and https://flowkeyboard.com/zh/use-cases/ ### Voice Input in Codex English: https://flowkeyboard.com/use-cases/codex-voice-input.html 简中: https://flowkeyboard.com/zh/use-cases/codex-voice-input.html Flow combines fast, accurate, dependable voice input with a personal dictionary and text organization for Codex desktop prompts on macOS and Windows. For Chinese dictation, about 180 characters per minute is a reference speaking pace, not an English words-per-minute measurement or an end-to-end guarantee; actual input speed varies with speaking pace, content and processing time. Add product names, team abbreviations and specialist terms in Flow's Dictionary with +, or bulk import a word list, to improve recognition of your vocabulary. Flow supports recordings up to 10 minutes and hands-free input. Organized text makes it easier for the user to review intent and omissions before sending it to AI. The guides include an example about AI products that illustrates the difference between unstructured speech and an organized message. Flow 的核心价值是快速、准确、稳定。中文口述约每分钟 180 字是速度参考,实际输入速度随语速、内容和处理时间变化。将产品名、缩写和专业词汇加入个性化词典,可提升识别准确性;支持批量导入、最长 10 分钟录音和免提输入。准确识别减少纠错,结构化整理减少用户自己的回读和修改负担。页面完整复用 2026 年 7 月 Windows 正式发布帖第 3 页的真实输入输出,产品信息与观点反映录音当时的内容。 ### Desktop and iPhone workflow On Mac, hold Fn for Flow voice input and release to finish; Fn + Space starts hands-free mode. Switching windows or apps does not interrupt recording merely because the Codex input field is no longer in view. Windows supports Flow shortcuts and hands-free recording too. When using Codex remotely in ChatGPT on iPhone, start recording with Flow, then browse other content, leave the keyboard, switch apps or lock the phone while continuing to speak. Return to the intended conversation to review and send the resulting text. Flow also supports continued recording with an Android phone locked. Continued recording is not automatic message sending. Mac 上按 Fn 口述,Fn + Space 免提;切换窗口或 App 不必打断表达。iPhone 上在 ChatGPT 中远程使用 Codex 时,Flow 录音不要求停留在当前对话或键盘,切换 App 或锁屏也可继续;Android 也支持锁屏继续录音。说完回到目标对话,检查、补充文字后发送。 ## The Problem Flow Solves Your head is full of ideas, but your fingers freeze on the keyboard. You know exactly what you want to say, but writing it feels impossible. Flow lets you skip that struggle. Just say what you are thinking -- no need for perfect phrasing, no worrying about grammar, you can even stumble over your words. When you are done, you get clear, polished text that captures exactly what you meant. Expression is no longer a barrier. Your ideas, finally delivered the way they deserve. --- ## How Flow Compares Flow Keyboard is fundamentally different from regular dictation tools. Here is how it compares: | Feature | Flow Keyboard | Apple Dictation | Typeless | Wispr Flow | |---------|:---:|:---:|:---:|:---:| | Filler word removal | Yes | No | Yes | Yes | | Grammar correction | Yes | No | Yes | Yes | | Text restructuring | Yes | No | Yes | Yes | | Translation mode | Yes (16 languages) | No | Yes | No | | Mixed-language input quality | Excellent | Basic | Good | Basic | | Optimized for AI coding tools | Yes (Claude Code, Codex, ChatGPT) | No | No | No | | Workplace-optimized output | Yes (professional tone) | No | No | No | | Preserves emotional tone in chat | Yes | No | No | No | | Custom dictionary accuracy | High (bulk import + sync) | No | Moderate | Moderate | | Privacy (no surrounding text collected) | Yes | Yes | No (collects context) | No (collects screen & context, uses data for training) | | Zero server-side data retention | Yes | No | Yes | Yes | | Price | 7-day free trial, from $2.50/mo | Free | $12-30/mo | $12-15/mo | | Platform | macOS, iOS, Windows, Android | macOS, iOS | Cross-platform (Electron) | Cross-platform (Electron) | **Key differences:** - Regular dictation gives you a raw transcript. Flow gives you polished, ready-to-send text. - Typeless collects surrounding text from your editing area for context. Flow does not collect any surrounding text -- only your voice input is processed. - Wispr Flow lacks translation mode and has basic Chinese support. Flow excels at Chinese, mixed-language input, and dialect support. - Choose your platform on the official download page: https://flowkeyboard.com/download.html. Installer size varies by version and platform and does not establish transcription speed or memory use. For detailed comparisons, see: - Flow vs Apple Dictation, Typeless & Wispr Flow: https://flowkeyboard.com/compare/apple-dictation-alternative.html - Flow vs Typeless: https://flowkeyboard.com/compare/typeless-alternative.html - Flow vs Wispr Flow: https://flowkeyboard.com/compare/wispr-flow-alternative.html --- ## Pricing - **7-day free trial** available -- try all features at no cost - **Free**: Registered users receive 5,000 words/week and can continue without a paid subscription after the Pro trial. Usage is calculated by Flow; remaining allowance is shown in the app. - **Plus**: $3/month or $30/year — 12,000 words/week - **Pro**: $12/month or $100/year — unlimited voice input, subject to fair use --- ## Features ### Mobile Keyboard Features On iOS and Android, Flow Rail provides voice input in a compact interface that takes up less screen space. Both mobile keyboards support Emoji input. Android also supports a split keyboard on supported wide screens and foldables, one-handed mode, and adjustment of keyboard height, width, and bottom spacing. Open the keyboard tools menu from the Flow logo to access layout and size controls. Split availability depends on the current window and input layout; it is not available in every keyboard configuration. These are platform-specific capabilities, not a promise of identical features on all four platforms. Public release details: https://flowkeyboard.com/releases.html. Current downloads: https://flowkeyboard.com/download.html. Android website and Google Play release versions may differ. ### Voice-to-Polished-Text Flow goes beyond simple dictation. Regular voice dictation transcribes your speech word-for-word, including all the "ums", "ahs", repetitions, and grammatical errors. Flow is fundamentally different: - **Removes filler words**: "um", "uh", "like", "you know" are automatically stripped - **Fixes grammar**: Sentence structure is corrected and improved - **Adds punctuation**: Proper commas, periods, and formatting are added - **Organizes logic**: Rambling thoughts are restructured into clear, organized text - **Preserves meaning**: Your intent and tone are kept intact The result reads like carefully written text, not a raw transcript. Typical response time is under 2 seconds for short recordings. ### How It Works Your audio is sent to our servers, where intelligent text refinement converts your speech into polished text -- removing filler words, fixing grammar, adding punctuation, and restructuring for clarity. Flow maintains zero server-side data retention for voice content and generated text. By default, Flow does not collect surrounding cursor context text from app input fields. An internet connection is required. ### Multilingual Support (30+ Languages) Flow supports a wide range of languages including: - English, Chinese (Mandarin, Cantonese, and other dialects), Japanese, Korean - French, German, Spanish, Portuguese, Italian, Russian - Thai, Vietnamese, Arabic, Indonesian, and 15+ more ### Mixed-Language Input Flow automatically detects when you switch between languages mid-sentence. You do not need to toggle any language setting -- just speak naturally, mixing languages as you normally would in conversation. For example, you might mix Chinese and English, Japanese and English, or any other language combination, and Flow handles it automatically. ### Translation Mode (16 Target Languages) Translation mode lets you speak in one language and have the output appear in another. Supported target languages include: - Chinese (Simplified), Chinese (Traditional), Cantonese, English, Japanese, Korean - French, Spanish, German, Russian, Portuguese, Italian - Thai, Vietnamese, Indonesian, Arabic This is especially useful when you need to write emails or messages in a language you can understand but find difficult to type in. You think and speak in your native language, and Flow delivers the result in the target language. ### Language Variant Preferences Flow lets you choose preferred language variants for dictation output. Supported preferences include Traditional Chinese, Simplified Chinese, American English, and British English. ### Localized App Interface The app interface is available in English, Simplified Chinese, Traditional Chinese, and Japanese, in addition to its 30+ supported dictation languages. ### Custom Dictionary Flow includes a custom dictionary for names, product names, brand names, acronyms, and specialized terms that are often misrecognized. The dictionary supports bulk import and syncs automatically across signed-in devices. ### System-Wide Input Flow Keyboard works as a system-wide input method on macOS and Windows, and as a keyboard app on iOS and Android. Use it in any application that accepts text input: - **Messaging**: Slack, iMessage, WeChat, Teams, Discord, Feishu, DingTalk - **Email**: Gmail, Outlook, Apple Mail - **Documents**: Notion, Google Docs, Word, Obsidian - **Code**: Cursor, VS Code, Claude Code, ChatGPT, Codex - **Any other text field** on macOS ### iPad Support Flow Keyboard can be used on iPad through the iOS app and third-party keyboard extension. Install it from the App Store, add Flow Keyboard in Settings > General > Keyboard > Keyboards, enable Allow Full Access, and switch to Flow from the globe key when the on-screen keyboard is visible. When an external hardware keyboard or Magic Keyboard is connected, iPadOS usually hides the on-screen keyboard. Physical key input is handled by iPadOS and does not pass through Flow's iOS keyboard extension, so Flow cannot transform text typed on the physical keyboard. To use Flow voice input on iPad, show the on-screen keyboard, switch to Flow, and tap the record button. If the current app does not let you show the on-screen keyboard, temporarily disconnect the hardware keyboard. For hardware-keyboard-first workflows, use Flow Keyboard for macOS. ### Apple Vision Pro Support Flow Keyboard is not supported as a system keyboard on Apple Vision Pro. The iPhone/iPad app may open on Apple Vision Pro as a compatible app if it is available there, but visionOS does not load third-party keyboard extensions, so Flow cannot be selected as the keyboard inside other Vision Pro apps. If you use Apple Vision Pro with Mac Virtual Display, use Flow Keyboard for macOS on the Mac. A future Vision Pro companion app could support dictating inside the app and copying text out, but it would not be a system-wide keyboard unless Apple adds third-party keyboard extension support to visionOS. ### Auto-Mute System Audio (macOS) On macOS, Flow can automatically mute system audio while you are recording, so music, videos, or notification sounds do not interfere with your voice input. The setting is optional and can be enabled in Settings. --- ## Keyboard Shortcuts on macOS and Windows Official guide: https://flowkeyboard.com/keyboard-shortcuts.html Localized guides: - Simplified Chinese: https://flowkeyboard.com/zh/keyboard-shortcuts.html - Traditional Chinese: https://flowkeyboard.com/zh-Hant/keyboard-shortcuts.html - Japanese: https://flowkeyboard.com/ja/keyboard-shortcuts.html - Korean: https://flowkeyboard.com/ko/keyboard-shortcuts.html Open **Flow Keyboard > Settings > Keyboard Shortcuts** to change a shortcut. Click the shortcut you want to change, then press the new key or key combination. ### macOS Default Shortcuts | Action | Primary | Built-in backup | |--------|---------|-----------------| | Voice Input | Fn | Control + Option | | Hands-free Mode | Fn + Space | Control + Option + Space | | Translation Mode | Fn + Shift | Control + Shift | Backup shortcuts are disabled by default. Turn on **Enable Backup Shortcuts** to use primary and backup shortcuts at the same time. Flow also supports custom modifier-only shortcuts and distinguishes the left and right Control, Option, and Command keys when a side-specific shortcut is recorded. Voice Edit uses the Voice Input shortcut after text is selected. Press Esc to cancel the current recording. Double-press the primary Voice Input shortcut to enter hands-free mode. With the default setup, double-press Fn. While hands-free recording is active, press the primary Voice Input shortcut once to finish. ### Windows Default Shortcuts | Action | Default shortcut | |--------|------------------| | Push-to-talk | Right Alt | | Hands-free Mode | Right Alt + Space | | Translation Mode | Right Alt + Right Shift | | Voice Edit | Right Alt + E | | Cancel | Esc | Windows hands-free mode can use **Right Ctrl** or **Right Alt** as a single-key shortcut. Right Ctrl is recommended because it preserves the default Right Alt push-to-talk shortcut. Press Right Ctrl once to start recording and again to finish. If Right Alt is assigned to hands-free mode while it is already used for push-to-talk, Flow automatically reassigns the conflicting shortcut. Windows shortcut rules: - Push-to-talk can use Left Alt or Right Alt as a single key. - Hands-free mode can use Right Alt or Right Ctrl as a single key. - Letters, numbers, Space, Enter, Shift, Win, Left Alt, and Left Ctrl cannot be single-key hands-free shortcuts. - Recording shortcuts containing the Windows key are not supported. - Ctrl + Space is not supported because it commonly conflicts with Windows and input-method behavior. ### How Recording Shortcuts Work - **Voice Input / Push-to-talk:** Hold the shortcut while speaking, then release it to finish and insert the text. - **Hands-free Mode:** Press once to start without holding a key, then press the hands-free shortcut again to finish. - **Translation Mode:** Press once to start hands-free recording and output the speech in the selected target language. - **Voice Edit:** Select text, use the Voice Edit shortcut, and say how the selected text should change. Flow checks shortcuts for conflicts as they are recorded. On macOS, conflicts between Flow actions are blocked and common system-shortcut conflicts produce a warning. On Windows, unsafe or unsupported global shortcuts are rejected. Use **Reset to Defaults** if needed. --- ## Use Cases ### Quick Replies (Slack, iMessage, Teams, Discord) **You say:** "So I looked at, like, the proposal you sent, and I think overall it's pretty good, but there are a few spots that probably need some changes, um, like the data analysis part in section two feels off, and also the timeline in section three seems kinda tight, we can discuss it more" **Flow outputs:** "I looked at your proposal — overall it's good, but a few things need changes: 1. The data analysis logic in Section 2 feels off 2. The timeline in Section 3 seems tight. We can discuss further." ### Weekly Reports & Updates (Gmail, Outlook, Notion, Docs) **You say:** "Um so basically last week's project progress, the frontend team already finished the homepage redesign, and the backend API endpoints are mostly done but there are a few that haven't been integration tested yet, oh right the QA team found two bugs, one is the login page verification code sometimes doesn't come through, and the other is the order list pagination is broken, the frontend says they can fix both this week, and next week the plan is to finish the remaining API integration and start canary testing" **Flow outputs:** "Last week: 1. Frontend: homepage redesign complete 2. Backend: API endpoints mostly done, some pending integration 3. QA found 2 bugs: verification code + order pagination (Frontend estimates fix this week). Next week: 1. Complete remaining API integration 2. Begin canary testing" ### Translation Mode (16 target languages) **You say (in English):** "So the project with the team in Shanghai, our timeline is probably going to slip by about a week, mainly because QA found some pretty serious issues that need to be fixed first, can you let them know" **Flow outputs (in Chinese):** "关于和上海团队合作的项目——我们的时间线可能需要延后一周。主要是因为 QA 发现了几个比较严重的问题,需要先解决。麻烦帮我跟他们沟通一下。" --- ## Setup Instructions ### On macOS 1. Download Flow Keyboard from our official website at flowkeyboard.com 2. Install the application 3. Open System Settings 4. Go to Keyboard > Input Sources 5. Click the "+" button 6. Search for "Flow Keyboard" and add it 7. Use the input source menu in the menu bar to switch to Flow Keyboard ### On iOS 1. Download Flow Keyboard from the App Store: https://apps.apple.com/app/id6758776986 2. Go to Settings > General > Keyboard > Keyboards 3. Tap "Add New Keyboard" 4. Select "Flow Keyboard" 5. Tap "Flow Keyboard" again and enable "Allow Full Access" 6. When typing in any app, tap and hold the globe icon to switch to Flow Keyboard ### On iPad Use the same App Store app and setup path as iOS. Flow works when the on-screen keyboard is visible and Flow Keyboard is selected. With an external hardware keyboard or Magic Keyboard connected, show the on-screen keyboard first, then switch to Flow and tap the record button for voice input. If the current app does not let you show the on-screen keyboard, temporarily disconnect the hardware keyboard. --- ## Privacy & Security - **Zero server-side retention for voice content** -- we do not retain user voice recordings or generated text on our servers after processing - **Audio processed transiently** -- Audio is sent to our servers for processing, then immediately discarded - **No default surrounding-text context collection** -- by default, we do not collect nearby cursor context text from app input fields - **Device-side local history may exist** -- some recordings/history may be stored locally on your device for product features, not on our servers - **Contracted subprocessors** -- we may use cloud/AI processors configured for transient processing and no retention of user content after completion - **Zero server-side data retention** -- Flow does not store server-side recordings, transcriptions, or generated text - **No server transcriptions stored** -- all text is discarded after the response is returned on our servers - **Internet connection required** -- voice input requires an active internet connection - **No sale of personal information** -- we do not sell personal information or share it for cross-context behavioral advertising - **Regional privacy requests** -- GDPR/UK GDPR and CCPA/CPRA requests can be sent to service@snowline.ai ### Full Access (iOS) iOS requires keyboard extensions to have Full Access permission for microphone and network functionality. Flow uses Full Access only to capture voice input and send processing requests. Full Access does not bypass iOS protection for secure/password fields. Flow maintains zero server-side data retention for voice content and generated text. Full privacy policy: https://flowkeyboard.com/privacy.html --- ## Frequently Asked Questions Official support and FAQ: https://flowkeyboard.com/support.html Localized support pages: - Simplified Chinese: https://flowkeyboard.com/zh/support.html - Traditional Chinese: https://flowkeyboard.com/zh-Hant/support.html - Japanese: https://flowkeyboard.com/ja/support.html - Korean: https://flowkeyboard.com/ko/support.html ### What is Flow Keyboard? Flow Keyboard is a voice input method for macOS, iOS, Windows, and Android, that turns speech into polished, well-structured text. It supports 30+ languages, automatic mixed-language detection, and translation into 16 target languages. Speaking is 4x faster than typing, and Flow ensures the output is clean and ready to send -- not a raw transcript full of filler words and grammatical errors. ### How does voice input work? You speak, and Flow turns your words into polished text. Your audio is recognized across 30+ languages, then intelligent text refinement removes filler words, fixes grammar, adds punctuation, and restructures your thoughts into clear writing. The entire process typically takes under 2 seconds for short recordings. Flow maintains zero server-side data retention for voice content and generated text. ### How accurate is Flow Keyboard? Flow delivers high-quality output through intelligent text refinement. It supports 30+ languages with strong accuracy across accents and speaking styles. Beyond recognition, Flow removes filler words, fixes grammar, adds proper punctuation, and restructures text for clarity. The result is polished text that reads as if you had carefully written it. ### How is Flow different from Apple Dictation? Apple Dictation is built into macOS and is designed to enter spoken text. Apple's current macOS documentation says it can dictate text of any length without a timeout. Flow Keyboard focuses on AI-refined output: it removes filler words, fixes grammar, restructures text, automatically detects mixed-language speech, and provides a dedicated translation mode. Flow supports recordings up to 10 minutes. Apple Dictation reference: https://support.apple.com/guide/mac-help/mh40584/mac ### Which languages are supported? Flow supports 30+ languages for voice recognition including English, Chinese (Mandarin, Cantonese, and other dialects), Japanese, Korean, French, German, Spanish, Portuguese, Italian, Russian, Thai, Vietnamese, Arabic, Indonesian, and more. Mixed-language input is automatically detected. Translation mode supports 16 target languages. ### How does mixed-language input work? Flow automatically detects when you switch between languages mid-sentence. You do not need to toggle any language setting -- just speak naturally, mixing languages as you normally would in conversation. This works for any language combination, not just specific pairs. ### What can I use Flow Keyboard for? Flow Keyboard works in apps that accept text input on supported platforms. Common use cases include: quick replies in Slack, iMessage, Teams, Feishu, WeChat, and Discord; composing emails in Gmail or Outlook; writing documents in Notion, Google Docs, or Obsidian; coding prompts in Claude Code, Cursor, or ChatGPT; and any situation where you want to speak instead of type. ### Does Flow work with Slack, WeChat, VS Code? Yes. Flow Keyboard works system-wide on macOS and Windows, so it works in many apps that accept text input. This includes Slack, WeChat, Feishu, DingTalk, VS Code, Cursor, Claude Code, Gmail, Outlook, Notion, iMessage, and other text fields on supported platforms. ### What is translation mode? Translation mode lets you speak in one language and have the output appear in another. Flow supports 16 target languages including Chinese (Simplified and Traditional), Cantonese, English, Japanese, Korean, French, Spanish, German, Russian, Portuguese, Italian, Thai, Vietnamese, Indonesian, and Arabic. This is especially useful when you need to write in a language you can understand but find difficult to type in. ### How fast is Flow? What is the latency? Flow typically delivers results in under 2 seconds for short recordings (a few sentences). Longer recordings may take slightly more time proportional to their length. Latency depends on your internet connection speed and recording length. Flow requires an active internet connection. ### How do I enable Flow Keyboard on my device? On macOS: Download and install Flow Keyboard from https://flowkeyboard.com/download.html, grant Accessibility and Microphone permissions when prompted, then hold Fn to record and release to finish. On Windows: Open https://flowkeyboard.com/download.html?platform=windows, download and install the app, sign in, and configure your microphone and shortcut. On iOS: Download from the App Store (https://apps.apple.com/app/id6758776986), then go to Settings > General > Keyboard > Keyboards > Add New Keyboard > Flow Keyboard, and enable Full Access. On Android: Download from https://flowkeyboard.com/download.html?platform=android, open Flow, follow the guide to enable and select Flow Keyboard, and allow microphone access for voice input. Google Play availability is shown on the download page. ### Is my voice data private? Yes. Flow Keyboard maintains zero server-side data retention for voice content and generated text. We do not store recordings, transcriptions, or generated text on our servers. Some history items may be stored locally on your device for product features. ### What platforms are supported? Flow Keyboard is available on macOS, iOS, Windows 10 or later on x64 PCs, and Android. It provides system-wide voice input on macOS and Windows, and a keyboard on iOS and Android. Downloads and platform requirements: https://flowkeyboard.com/download.html. Android APK and Google Play availability: https://flowkeyboard.com/download.html?platform=android. iOS download: https://apps.apple.com/app/id6758776986. ### Can I use Flow Keyboard on iPad? Yes. Install Flow Keyboard from the App Store, add it in Settings > General > Keyboard > Keyboards, enable Allow Full Access, then switch to Flow from the globe key when the on-screen keyboard is visible. If an external hardware keyboard or Magic Keyboard is connected, iPadOS usually hides the on-screen keyboard and physical key presses are handled by iPadOS, not by Flow's iOS keyboard extension. To use Flow voice input on iPad, show the on-screen keyboard, switch to Flow, and tap the record button. If the current app does not let you show the on-screen keyboard, temporarily disconnect the hardware keyboard. If you mainly work with a physical keyboard, Flow Keyboard for macOS is the better fit. ### Can I use Flow Keyboard on Apple Vision Pro? Not as a system keyboard. The iPhone/iPad app may open on Apple Vision Pro as a compatible app if it is available there, but visionOS does not load third-party keyboard extensions, so Flow cannot be selected as the keyboard inside other Vision Pro apps. If you use Apple Vision Pro with Mac Virtual Display, use Flow Keyboard for macOS on the Mac. A future Vision Pro companion app could support dictating inside the app and copying text out, but it would not be a system-wide keyboard unless Apple adds support for third-party keyboard extensions on visionOS. ### Is Flow Keyboard free? Flow Keyboard offers a 7-day free trial so you can try all features at no cost. After the trial, choose Plus ($3/month or $30/year) for 12,000 words/week, or Pro ($12/month or $100/year) for unlimited usage, subject to fair use. ### How do subscriptions work across devices? Subscribe in the app. Available payment methods are shown in the app at checkout. After subscribing, sign in with the same Flow account to use your subscription on supported platforms. ### Why does Flow need Full Access? On iOS, keyboard extensions require Full Access to use microphone and network features. Flow needs this for voice capture and request/response processing. Full Access does not grant access to secure/password fields protected by iOS. Flow maintains zero server-side data retention for voice content and generated text. ### Why does my iPhone show the yellow microphone indicator or Dynamic Island microphone alert? This is Apple's system privacy indicator. Flow's iOS keyboard uses the main app to provide voice input, so after you stop one voice input session, Flow keeps the voice input channel available until the auto-close timer expires. The default timer is 5 minutes. The yellow indicator means iOS still considers the microphone channel available; it does not mean Flow is continuously recognizing, transcribing, or uploading audio. Audio is used for speech-to-text only when you tap the record button. To reduce how long the indicator appears, open the Flow Keyboard app and change Auto Close on the Home screen to 5 seconds or 1 minute. Shorter timers make the indicator disappear sooner, but may require more frequent jumps back to the app when starting voice input. ### How is Flow different from regular voice dictation? Regular voice dictation transcribes your speech word-for-word -- every "um", "uh", repetition, and grammatical mistake goes straight into the text. Flow is fundamentally different: it removes filler words, fixes grammar, adds punctuation, restructures logic, and preserves your meaning. The result reads like carefully written text, not a raw transcript. It is the difference between having a stenographer and having an editor. ### How do I turn off Live Activities on iPhone? Go to your iPhone's Settings → Flow Keyboard → Live Activities and turn off the toggle. This will also remove it from CarPlay and Apple Watch. ### How do I add words to the dictionary? Adding frequently used words that are often misrecognized improves accuracy. There are three ways: 1. Open the Flow Keyboard app, go to Dictionary, and use bulk import for multiple terms. 2. Tap "+", and enter a word manually. 3. While using Flow Keyboard in any app, select a word in the output -- an Add button will appear at the bottom-right of the keyboard for quick addition. This is especially useful for people's names, product names, brand names, and specialized terms. ### Does my dictionary sync across devices? Yes -- your custom dictionary syncs across signed-in devices automatically. Requires Flow Keyboard 0.2.17 or later on iOS and 0.8.1 or later on macOS. Sync runs as soon as you open the Flow Keyboard app and usually finishes within a few seconds. To verify, open the Flow Keyboard app and go to the Dictionary page. ### The keyboard only shows English keys. How do I type in other languages? Flow Keyboard includes a built-in Simplified Chinese Pinyin keyboard for typing Chinese directly (iOS 0.2.23 and later). For other languages: 1. Tap the globe button at the bottom-left to switch to your system keyboard (e.g., Kana for Japanese). 2. Or use voice editing -- after speech-to-text, tap the magic wand button at the bottom-right and speak your changes to modify the content by voice. ### Voice input confuses Chinese 他 (he) and 她 (she). How do I fix it? In Chinese, 他 (he) and 她 (she) are pronounced identically (both "tā"), so voice recognition alone cannot distinguish them. Three workarounds: 1. Use the magic wand after dictation -- tap the magic wand button and speak a correction like "change 他 to the female 她." 2. Provide more context while speaking -- include gender-related context so Flow can infer the correct character automatically. 3. Use the Pinyin keyboard to fix it directly -- tap the keyboard button to switch to Pinyin input and select the correct character (iOS 0.2.23 or later). The richer the context, the more accurate the result. This capability is continuously being improved. ### How do I report a problem or give feedback? The best way: open Flow Keyboard Settings and tap the Feedback button. This opens an email with your app version and OS version pre-filled in the subject line, helping locate the issue quickly. If a specific voice input result has a problem (too slow, poor quality), find it in your History, tap the Feedback button, and describe the issue. After submitting feedback, sending an additional email with details helps the team investigate more effectively. You can also reach the team via social media channels or the user community -- email service@snowline.ai to join. --- ## Insights Flow Insights is a multilingual editorial section for deeper essays about creation, expression, voice input, AI workflows, and human-computer interaction. ### I Spoke an Idea Into My Phone. It Became a GitHub PR. URL: https://flowkeyboard.com/insights/thought-to-action-workflow.html Markdown: https://flowkeyboard.com/insights/thought-to-action-workflow.md This essay follows a real workflow from a spoken idea captured with Flow Keyboard, through iterative discussion and research in ChatGPT, to a GitHub Issue, Codex implementation, and a reviewable GitHub pull request. It argues that the largest gain comes from preserving context between stages instead of repeatedly rebuilding it. ### 我只是在手机上说了一个想法,它最后变成了 GitHub PR URL: https://flowkeyboard.com/zh/insights/thought-to-action-workflow.html Markdown: https://flowkeyboard.com/zh/insights/thought-to-action-workflow.md This essay documents a real workflow in which an initially unstructured idea was spoken into Flow Keyboard on a phone, discussed and researched across multiple ChatGPT turns, converted into a GitHub Issue, executed by Codex, and reviewed as a GitHub pull request. Its central argument is that the main productivity gain comes from preserving context across each stage instead of repeatedly reconstructing the same background. It links to Flow's dedicated Codex voice-input use case for readers who want a task-oriented implementation guide. ### 我對著手機說出一個想法,它後來成了一個 GitHub PR。 URL: https://flowkeyboard.com/zh-Hant/insights/thought-to-action-workflow.html Markdown: https://flowkeyboard.com/zh-Hant/insights/thought-to-action-workflow.md 繁體中文版以同一個真實流程說明:先用 Flow Keyboard 口述尚未成形的想法,再透過 ChatGPT 討論與研究,將共識整理成 GitHub Issue,交由 Codex 執行,最後在 GitHub PR 中完成審查。重點在於讓背景脈絡持續流轉,而不是每一步重新交代。 ### スマホにアイデアを話した。それがGitHub PRになった。 URL: https://flowkeyboard.com/ja/insights/thought-to-action-workflow.html Markdown: https://flowkeyboard.com/ja/insights/thought-to-action-workflow.md Flow Keyboardで口述した未整理のアイデアを、ChatGPTでの対話と調査、GitHub Issue、Codexによる実装、GitHub PRでのレビューへつなげた実例を紹介する。各段階で前提を説明し直すのではなく、同じ文脈を成果物として引き継ぐことが生産性を高める、という考え方が中心にある。 ### 휴대폰에 아이디어를 말했더니 GitHub PR이 되었다. URL: https://flowkeyboard.com/ko/insights/thought-to-action-workflow.html Markdown: https://flowkeyboard.com/ko/insights/thought-to-action-workflow.md Flow Keyboard로 말한 정리되지 않은 아이디어가 ChatGPT 대화와 조사, GitHub Issue, Codex 구현, GitHub PR 검토로 이어진 실제 과정을 소개한다. 단계마다 배경을 다시 설명하는 대신 같은 맥락을 결과물로 이어 가는 것이 핵심이다. ### 1987 年,雷军想了一整夜,最后决定先做一件小事 URL: https://flowkeyboard.com/zh/insights/lei-jun-first-step.html This essay starts from Lei Jun's account of walking around Wuhan University's track for an entire night in 1987 after deciding that he wanted to build a great company. It follows how he reduced that distant ambition to a concrete goal, sought advice from senior students, and used action to bring back information he could not obtain by thinking alone. The article connects this story to research on task ambiguity, short-term mood regulation, cognitive offloading, external representations, and implementation intentions. Flow Keyboard is positioned as a low-friction way to move an idea out of working memory: users can speak freely, turn a long and messy explanation into structured text, discuss it with ChatGPT, and continue with GitHub Issues or Codex when the task involves software development. ### 张朝阳克服拖延的方法:从“拖到最后一天”到“当天晚上就想做完” URL: https://flowkeyboard.com/zh/insights/zhang-chaoyang-procrastination-warmup.html This essay starts from Zhang Chaoyang's 2022 conversation with positive-psychology researcher Peng Kaiping about procrastination and mental warm-up. It connects task ambiguity, short-term mood regulation, attention residue, spoken expression, and AI-assisted questioning to explain why difficult work often needs an entry point before it needs more willpower. Flow Keyboard is positioned as the bridge between thinking aloud and continued action: users can speak through an unfamiliar task, preserve messy raw material as structured text, then use an AI conversation to identify missing information and reduce the task to one concrete next step. ### 为什么好想法,总在走路时出现? URL: https://flowkeyboard.com/zh/insights/walking-ideas-voice-input.html This essay explores why ideas often appear while walking, why speech is much older than writing and keyboards, and why voice input should be understood as a creative entry point rather than only a faster typing tool. It references research on walking and divergent thinking, mind wandering, nature and attention, oral tradition, creator rituals, and the historical evolution from speech to writing to keyboards. The essay positions Flow Keyboard as "voice input + automatic editing": users can speak naturally, including pauses, filler words, repetition, and messy phrasing, while Flow removes filler words, adds punctuation, organizes structure, merges repetition, and turns spoken thoughts into text that can be edited further. ### MacBook 最被低估的硬件,是麦克风 URL: https://flowkeyboard.com/zh/insights/macbook-mic.html This essay argues that the MacBook's three-microphone array and Fn key are underused input hardware. It cites Apple's MacBook Pro and MacBook Air technical specifications for three-mic arrays, directional beamforming, high signal-to-noise ratio on MacBook Pro, and call-oriented Voice Isolation / Wide Spectrum microphone modes. The essay explains why laptop voice input is difficult, why typing compresses thoughts, and why AI prompts benefit from richer spoken context. Flow Keyboard is positioned as the software layer that turns natural speech into usable text by removing filler words, adding punctuation, organizing paragraphs, and clarifying structure. ### 给 AI 的信息太少,它只能靠猜 URL: https://flowkeyboard.com/zh/insights/voice-ai-prompt.html This essay explains why short AI prompts often produce weak results: the model has to guess missing context such as users, scenarios, data structure, constraints, edge cases, and acceptance criteria. It argues that typing makes users compress their thinking, while speaking lets them provide fuller context more naturally. Flow Keyboard is positioned as a prompt-writing layer: users speak for a few minutes, Flow removes filler words, adds punctuation, organizes paragraphs, clarifies logic, and turns the spoken context into a prompt that can be pasted into Codex, ChatGPT, Claude, Cursor, or other AI tools. ### 频繁记录自己的人,最后都发现了什么 URL: https://flowkeyboard.com/zh/insights/frequent-self-recording.html This essay argues that frequent self-recording preserves a concrete version of the past self, making repeated worries, memory distortion, and personal change visible over time. It uses Schacter's work on memory bias to explain why remembered experience can diverge from contemporaneous records. Flow Keyboard is positioned as a way to lower the threshold of journaling by turning short spoken reflections into clean written records with filler words removed, punctuation added, and paragraphs organized. ### 费曼学习法,大部分人卡在第二步 URL: https://flowkeyboard.com/zh/insights/feynman-learning.html This essay explains why the Feynman learning technique often fails in practice: people feel they understand a lesson until they have to explain it aloud. It uses Feynman's childhood bird story, his Brazil teaching experience, retrieval practice research, the protégé effect, and expecting-to-teach research to argue that speaking is the lower-threshold output mode. Flow Keyboard lets learners explain a concept aloud, removes filler words, adds punctuation, organizes the spoken explanation, and keeps the result as a clean written note, making knowledge gaps visible without turning learning into a heavy writing session. ### 灵感来了接不住,怎么办? URL: https://flowkeyboard.com/zh/insights/inspiration-capture.html This essay argues that inspiration is often lost during premature editing rather than simple forgetting. It uses Peter Elbow's freewriting practice to separate generation from editing, then explains why speech is a better first capture mode: speaking lets the idea unfold continuously before the writer starts judging wording. Flow Keyboard is positioned as a capture layer that turns rough spoken ideas into clean notes with filler words removed, punctuation added, and paragraphs organized, so the user can edit later. ### 如何像马伯庸一样,一年写下 17 万字日记? URL: https://flowkeyboard.com/zh/insights/maboyong-diary.html This essay explains why Ma Boyong's diary method resonated: he lowered the threshold of journaling by using short voice notes, recording one concrete thing at a time, and avoiding literary polish. It frames diary persistence as a mechanism problem rather than only a willpower problem. Flow Keyboard is positioned as a way to make "spoken recording" practical by turning natural speech into cleaned-up text with filler words removed, punctuation added, paragraphs organized, repetition merged, and messy phrasing clarified. --- ## Terms of Use https://flowkeyboard.com/terms.html - Flow's 7-day Pro trial requires no payment method and does not automatically convert to a paid subscription. Without an affirmative subscription, the account returns to the Free tier when the trial ends. - Monthly subscriptions are billed monthly. Annual subscriptions are billed in advance for the full one-year period and renew annually unless canceled. A displayed monthly-equivalent annual price is informational only. - Paid subscriptions auto-renew for the same billing period unless canceled before the next renewal date. Cancellation has no fee, stops future renewal, and normally preserves paid access through the current billing period. - Pro Unlimited covers normal intended use, including normal personal and professional use, subject to fair use. It does not authorize automated excessive use, sharing account credentials among multiple people, resale, bypassing technical protections, or other abusive/non-subscriber use. Internal safeguards are not presented as an ordinary customer quota. - Refund handling depends on purchase channel and applicable law. Apple determines App Store refund eligibility and processes requests through Apple Billing and Subscriptions: https://support.apple.com/billing - Google Play purchases follow Google Play's subscription and refund processes: https://support.google.com/googleplay/answer/7018481 and https://support.google.com/googleplay/answer/15574908 - Cancellation of a subscription purchased directly from Snowline does not entitle the subscriber to a refund or credit for unused time, except where applicable law requires otherwise. This does not affect prorated billing adjustments or credits that may apply when changing plans under the relevant purchase channel's rules. Non-waivable consumer rights remain intact. - These clarifications apply prospectively. Earlier purchases continue to be evaluated under purchase-time disclosures, applicable law, and the relevant purchase channel's support process. --- ## Contact Email: service@snowline.ai Response time: typically within 24 hours ## Referral rewards Official rules and program updates: https://flowkeyboard.com/referral.html (localized rules are linked there). Eligible invitees receive 30 days of Flow Pro after experience completion and review. Inviters receive up to 30 days per eligible invitation, with a cumulative 90-day base-reward cap including historical base rewards, excluding past milestone bonuses; only remaining capacity is awarded. The inviter cap does not reduce an eligible invitee reward. Invitation-count milestones are discontinued. From the rules' published effective date, all unissued rewards follow the updated rules; legitimately issued rewards are unaffected, but rewards obtained through violations may be revoked or recovered. Fraud may lead to reward refusal, participation restrictions, or, for severe or repeated violations, account suspension or termination with an appeal contact in the rules. Progress completion is not proof of reward approval or delivery. Consult the official page for effective dates, scope and appeals. ## 朱自清也发现:有些旅行,回来就不好写了 从朱自清《伦敦杂记》的往事出发,谈旅行期间的文字记录、个人化回忆线索,以及如何用 Flow 输入法把旅途中的经历说下来。 HTML: https://flowkeyboard.com/zh/insights/travel-writing-zhu-ziqing.html Markdown: https://flowkeyboard.com/zh/insights/travel-writing-zhu-ziqing.md ## 朱自清也發現:有些旅行,回來就不好寫了 繁體中文版:從《倫敦雜記》談旅行中的文字記錄、個人化回憶線索與語音筆記。 HTML: https://flowkeyboard.com/zh-Hant/insights/travel-writing-zhu-ziqing.html Markdown: https://flowkeyboard.com/zh-Hant/insights/travel-writing-zhu-ziqing.md ## Accessibility Support Program (Flow Accessibility Program) People who have meaningful difficulty with conventional text entry because of accessibility-related needs can apply for 12 months of Flow Plus at no cost, with a renewal application possible when the period ends. No disability certificate, medical record, diagnosis, income proof, payment method, or promotional testimonial is required. Applicants provide their Flow account email and a brief self-declaration by email to service@snowline.ai; applications are reviewed individually. Eligibility is assessed for whether an alternative input method may be helpful, not for a particular diagnosis or income bracket. The normal Plus usage quota applies. Applying does not cancel an existing paid Apple, Google Play, or Stripe subscription. Flow cannot guarantee perfect speech recognition for every speaking pattern or full screen-reader compatibility across platforms. - English: https://flowkeyboard.com/accessibility.html - 简体中文: https://flowkeyboard.com/zh/accessibility.html - 繁體中文: https://flowkeyboard.com/zh-Hant/accessibility.html - 日本語: https://flowkeyboard.com/ja/accessibility.html - 한국어: https://flowkeyboard.com/ko/accessibility.html