How to Master Dictation in Raycast  🔗 **Source:** [How to Master Dictation in Raycast](https://www.youtube.com/watch?v=RUh1KDgSOsc#transcript) > [!note]+ Document Notes > tags: [dictation, raycast, speech-to-text, productivity-tools, developer-tools, accessibility-tooling] > author: [MISSING] > source: Raycast (YouTube) > reference: https://www.youtube.com/watch?v=RUh1KDgSOsc > date created: 2026-06-18 09:45 > > # 🏷️ Title: How To Master Dictation in Raycast > <!-- deep-read:sections start --> ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [Raycast](https://raycast.com) - [Raycast for new beta (Mac)](https://raycast.com/new) - [Raycast manual — hotkeys page](https://manual.raycast.com) (linked in video description) ## 📝 Summary - **What happened.** A Raycast walkthrough video demonstrates the dictation feature in the new Raycast 2.0 beta (now also on Windows), moving from first-time setup (permissions, hotkey) to power-user tips around "styles," global custom instructions, custom vocabulary, microphone selection, dictate-to-note, and dictation history. - **Why it matters.** Dictation lets you speak into any text field in any app with a single hotkey, turning speech into a daily input method for messages, emails, prompts, and notes — a faster alternative to typing that becomes "second nature" with muscle memory. - **Key stakeholders.** Raycast users on macOS and Windows, knowledge workers who write a lot, multilingual users, and people building AI/LLM prompts by voice (Courtney's Claude Code workflow fits here — speak your prompts instead of typing them). - **Risks and tradeoffs.** Accuracy depends on accent handling, custom vocabulary, and microphone consistency; the feature is free during beta but moves to Raycast's Pro plan after 2.0 ships. Misrecognitions (e.g., "capture" → "capture") require manual vocab tuning. - **Bottom line.** Set a hotkey, lean on styles and global instructions for context-aware transcription, and dictation becomes a fast, hands-free way to write anywhere — including dictating prompts. ## 📖 Main Content - **What dictation is.** A Raycast 2.0 beta feature (macOS and Windows) that transcribes speech into any focused text field, triggered by a hotkey without opening Raycast or searching for a command. The presenter calls it "an absolute daily driver." - **Setup.** Search "dictate" in Raycast; on first run an onboarding screen grants permissions and sets a hotkey. A hotkey is strongly encouraged here (unlike most Raycast commands) so muscle memory makes triggering automatic. If onboarding was dismissed, set the hotkey by searching "dictate" and opening the action panel with Command/Control + K. Suggested keys: hyperspace (mimics Command+Space), Option/Alt+D, F5, the globe key on Mac, or the Windows key. - **Using it.** In any input field, fire the hotkey and speak. - **Settings page.** Reached via Option+Command+Comma (Mac) or Control+Alt+Comma (Windows). The presenter notes dictation is "settings-heavy" and worth exploring. - **Styles.** The most powerful feature: custom transcription instructions for specific contexts. Styles can fire automatically by being assigned to specific apps/websites, or be picked manually like any command. Styles can have their own hotkeys — e.g., one that formats output as markdown. - **Global custom instructions.** Apply to all dictations (including styles). Good for declaring a foreign accent or forcing lowercase output. - **Style tips/hacks.** Per-app name handling (e.g., a Slack style that always spells colleague "John" correctly vs. friend "John Whittington" outside work); auto-translation (dictate English, output Spanish); format output like IPA phonetic transcription or LaTeX math equations; create a blank named style with its own hotkey to get multiple hotkeys for the same base dictation command (useful when switching keyboards). - **Other settings.** Choose a priority microphone for consistency across AirPods/streaming mics; pick confirm/cancel shortcuts; add custom vocabulary (independent of style) for brand names or repeatedly misrecognized words; view dictation stats (time saved). - **Beyond text fields.** Dictate-to-note saves transcription into a Raycast note; dictating with no focused text field auto-copies to clipboard; dictation history keeps a list of everything dictated. - **Availability.** Free during beta; becomes part of the Pro plan once Raycast 2.0 leaves beta. Mac users upgrade at raycast.com/new; Windows users download from raycast.com. ### 🗒️ Notes #### ⏱️ [00:00:00](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=0) Dictation in Raycast 2.0, now on Windows - Dictation launched in the Raycast 2.0 macOS beta and is now also on Raycast for Windows. - "a bunch of tips to help you get the most out of this feature, which has been an absolute daily driver for me." - Get started: search "dictate" in Raycast. #### ⏱️ [00:00:34](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=34) Onboarding, permissions, and the hotkey - First run shows an onboarding screen for permissions and hotkey. - Hotkey is "highly encouraged" specifically for dictation so muscle memory triggers it in any text field in any app. #### ⏱️ [00:01:08](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=68) Setting the hotkey via the action panel - If onboarding was dismissed, search "dictate" and open the action panel with Command/Control + K. - Presenter uses hyperspace (reminds them of Command+Space); others use Option/Alt+D, F5, globe key, or Windows key. - Manual page link for hotkeys is in the description. #### ⏱️ [00:01:40](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=100) Just fire the hotkey and speak - "In any input field, just fire your hotkey and speak." - Transition to deeper, "settings-heavy" tips. #### ⏱️ [00:02:12](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=132) Settings page and Styles - Settings via Option+Command+Comma (Mac) / Control+Alt+Comma (Windows). - Styles set custom transcription instructions for specific contexts; fire automatically per app/website or pick manually. #### ⏱️ [00:02:44](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=164) Per-style hotkeys and global custom instructions - Styles can have their own hotkeys (e.g., a markdown-formatting dictate command). - Global custom instructions apply to all dictations — declare a foreign accent, force lowercase, etc. #### ⏱️ [00:03:17](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=197) Per-app name handling and languages - A Slack style ensures colleague "John" is never butchered; a different context handles friend "John Whittington." - Set a default dictation language, or leave on auto for multilingual use (Spanish/Portuguese). #### ⏱️ [00:04:23](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=263) Translation and format-output style hacks - A style can auto-translate: dictate English, output Spanish. - Styles can output hard-to-write formats: IPA phonetic transcription or LaTeX math. #### ⏱️ [00:05:27](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=327) Blank-style hack for multiple hotkeys - Create a blank style (name only) with its own hotkey to get multiple hotkeys for the same base dictation command. - Useful when switching between keyboards so muscle memory doesn't have to adapt. #### ⏱️ [00:06:00](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=360) Microphone, confirm/cancel, and custom vocabulary - Choose a priority microphone for a consistent experience across AirPods/streaming mics. - Pick confirm/cancel shortcuts while dictating. - Add custom vocabulary (independent of style) for brand names or repeatedly misrecognized words. #### ⏱️ [00:07:05](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=425) Vocabulary fix example and dictation stats - A dedicated command jumps straight to the vocabulary settings. - Example: accent made "capture" misrecognized; adding it as a vocab word fixed it. - Presenter enjoys watching dictation stats (time saved). #### ⏱️ [00:07:38](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=458) Dictate anywhere: notes, clipboard, history - Dictate-to-note saves transcription into a Raycast note. - No focused text field → transcription auto-copies to clipboard to paste later. - Dictation history keeps a list of everything dictated. #### ⏱️ [00:08:46](https://www.youtube.com/watch?v=RUh1KDgSOsc&t=526) Beta pricing and where to get it - Free during beta; part of the Pro plan when Raycast 2.0 ships. - Mac: raycast.com/new. Windows: raycast.com. ## 📚 References - [Raycast — official site](https://raycast.com) - [Raycast 2.0 new beta (Mac)](https://raycast.com/new) - [Raycast Manual](https://manual.raycast.com) — hotkey setup page referenced in the description ## 🗣️ Paraphrase - Dictation arrived in the Raycast 2.0 beta on macOS and is now on Windows too; this video collects tips for getting the most from it. - Start by searching "dictate"; the first-run onboarding sets permissions and a hotkey, which is strongly recommended so triggering becomes muscle memory. - Once set, fire the hotkey in any text field and speak. - The settings page hosts Styles — context-specific transcription instructions that can auto-fire per app or be chosen manually, each with optional hotkeys. - Global custom instructions apply everywhere (accent notes, lowercasing); styles can fix names per app, translate between languages, or output IPA/LaTeX. - A blank style with its own hotkey is a hack for multiple hotkeys on one command. - Other settings: priority microphone, confirm/cancel keys, and custom vocabulary for brand names or misheard words. - You can dictate without a text field (saves to a note or the clipboard), and dictation history keeps everything; the feature is free in beta and Pro afterward. ## 📰 Newsletter **Talk to your text field: mastering Raycast dictation** If you spend your day typing into messages, emails, and AI prompts, Raycast's new dictation feature (Raycast 2.0 beta, now on Windows too) is worth a serious look. The setup is one move — set a hotkey during onboarding and let muscle memory do the rest, so you can speak into any text field in any app without opening Raycast first. The real power lives in *Styles*: context-aware transcription instructions that auto-fire per app or run on their own hotkey. Want a markdown-formatting dictate command? A Slack style that never butchers a colleague's name? A style that translates your English into Spanish, or spits out LaTeX math and IPA phonetics? All doable. Round it out with global custom instructions (declare your accent, force lowercase), a priority microphone so AirPods-vs-desk-mic switching stays consistent, and a custom vocabulary list for brand names and the word your accent keeps tripping on. You can even dictate with no text field focused — it lands in a Raycast note or your clipboard — and dictation history keeps a running log. For anyone driving Claude Code or other LLM tools, the takeaway is simple: speak your prompts. It's free during the beta and joins the Pro plan when 2.0 ships. ## ✅ Action Items #### DevRel - Developer advocacy: demo "dictate your Claude Code prompts" as a hands-free LLM-prompting workflow. - Education: short how-to on Raycast Styles for context-aware dictation (per-app, translation, markdown output). - Developer experience: Not applicable to this source. - Developer marketing: clip the "blank style = multiple hotkeys" hack as a standalone tip. - Community management: Not applicable to this source. - Internal/external/cross-project: Not applicable to this source. #### OSPO - Not applicable to this source (proprietary productivity tool, no open-source governance angle). #### Community - Maintainers: Not applicable to this source. - Contributors: Not applicable to this source. - Sponsors: Not applicable to this source. - Ecosystem: Not applicable to this source. #### WordPress - Not applicable to this source (no WordPress content). #### The WPCC - Not applicable to this source. ### FAIR - Not applicable to this source. ## 🧪 Hypotheses - H1: Setting a dedicated dictation hotkey increases daily dictation frequency versus launching the command via search (muscle-memory effect). - H2: Adding misrecognized words to custom vocabulary measurably reduces repeat transcription errors for accented speech. - H3: Context-assigned Styles produce fewer per-app errors (e.g., proper-noun handling) than a single global configuration. - H4: Dictating LLM prompts increases prompt length/throughput compared to typing them. ### 📊 Methodology - H1: A/B usage logging via dictation stats — compare dictation counts for hotkey-users vs. search-launch users over two weeks. - H2: Before/after error-rate count on a fixed phrase set for one accented speaker after adding target words to vocabulary. - H3: Controlled transcription of proper-noun-heavy text with a Style assigned vs. unassigned; count name errors. - H4: Timed task — compose N prompts by voice vs. keyboard; compare words-per-minute and edit count. ### ❓ Questions - Does dictation accuracy degrade noticeably with cheaper microphones, justifying the priority-mic setting? - How well do auto-translation Styles handle technical jargon and code terms? - What's the latency of transcription, and does it interrupt flow when dictating long prompts? - Will Pro pricing change adoption among casual users after the beta ends? ### 🔮 Future Research - Comparative accuracy of Raycast dictation vs. macOS/Windows native dictation and other tools (e.g., Whisper-based apps). - Effectiveness of Styles for domain-specific transcription (legal, medical, code). - Long-term ergonomic/RSI benefits of substituting voice for typing in developer workflows. ### 💡 Implications - **Practice:** Voice becomes a primary input for messages, emails, notes, and LLM prompts; Styles let one tool serve many contexts. - **Policy:** Speaking prompts/content aloud raises privacy considerations in shared/open offices. - **Security:** Dictated content routed through a transcription service implies data-handling questions (where does audio go?). - **Education:** Lowers the typing barrier; useful for learners who think faster than they type or have motor constraints. - **Funding models:** Free-in-beta → Pro is a typical productivity-tool conversion lever. ### 🧠 Learning Styles Assessment - Reading level estimate: spoken tutorial transcript; ~Flesch–Kincaid Grade 7–9 (short, conversational sentences, light jargon: "styles," "global custom instructions," "action panel"). Flesch Reading Ease ~60–70. #### 👀 Visual learners - Present: strong — it's a screen-demo video showing each setting and hotkey in action. - Improve: add a labeled settings-map diagram and a before/after vocabulary-fix screenshot. #### 🔊 Aural learners - Present: strong — narrated video with spoken walkthrough. - Improve: a 2-minute audio recap of the key settings. #### 🗣️ Verbal learners - Present: clear step-by-step verbal instructions with concrete examples (John/John Whittington, "capture"). - Improve: a glossary callout (style, global custom instructions, custom vocabulary, action panel). #### 🧩 Kinesthetic learners - Present: strong — it's a do-it-yourself procedural walkthrough you follow live. - Improve: a printable setup checklist (hotkey → permissions → first style → vocabulary). #### 🧠 Logical learners - Present: logical progression from setup → core use → power tips. - Improve: a decision tree for "global instruction vs. style vs. vocabulary." #### 🧑‍🤝‍🧑 Social learners - Present: light — presenter invites viewers to drop dictation stats in the comments. - Improve: a community thread for sharing favorite Styles. #### 🧘‍♀️ Solitary learners - Present: strong — self-paced tinkering is explicitly encouraged. - Improve: a self-check list to confirm each setting is configured. #### 🌿 Naturalistic learners - Present: none. - Improve: tie hands-free input to reduced repetitive strain and sustainable work habits. ### 📊 Accessibility Assessment #### Screen reader semantics - Present: YouTube page provides standard headings and structure around the video. - Improve: ensure a properly structured transcript and descriptive chapter headings. #### Keyboard navigation - Present: the feature itself is hotkey-driven (keyboard-first), and the YouTube player is keyboard-operable. - Improve: verify player controls have visible focus and labels. #### Contrast and readability - Present: standard YouTube UI generally meets contrast norms. - Improve: confirm 4.5:1 on any on-screen captions burned into the demo. #### Text resizing and reflow - Present: responsive YouTube layout. - Improve: confirm no clipping of on-screen text at 200% zoom. #### Media and captions - Present: a transcript exists; auto-captions likely available. - Improve: human-verified captions and a downloadable transcript (the feature itself aids accessibility as a typing alternative). #### Forms and interactions - Present: subscribe/comment forms referenced. - Improve: ensure form labels and accessible controls. #### Standards reference - Maps to WCAG 2.2 AA: 1.2.2 Captions, 1.2.1 Transcript, 1.4.3 Contrast, 2.1.1 Keyboard, 1.4.10 Reflow, 2.4.6 Headings/labels. Note: dictation as a voice-input alternative supports 2.5.x input modalities for users who can't type easily. ### 🌍 DEIB Assessment #### Cultural breadth - Present: explicit multilingual support — auto language detection, Spanish/Portuguese examples, English→Spanish translation styles, and accent-aware global instructions. - Improve: show more non-English/non-Latin-script examples (e.g., Hindi, Arabic, CJK). #### Inclusive language - Present: plain, friendly language; jargon mostly demonstrated in context. - Improve: define "style," "global custom instructions," and "action panel" inline for newcomers. #### Neurodiversity support - Present: chunked, step-by-step structure; consistent demo pattern. - Improve: add on-screen section markers/timestamps and a summary checklist. #### Representation - Present: limited — single presenter voice plus brief dialogue cameos. - Improve: feature diverse voices/accents demonstrating dictation accuracy across speakers. #### Psychological safety - Present: encouraging, low-pressure "tinker and explore" tone; normalizes accent-driven misrecognitions as fixable. - Improve: none needed; the framing is supportive. ### ✅ Logical Validity #### Strong reasoning - Causal (hotkey → muscle memory → frequent use): cleanly reasoned — a dedicated hotkey removes the search step, making dictation reflexive. - Practical demonstration: each claim ("styles can translate," "vocabulary fixes misrecognitions") is shown live, giving inductive support by demonstration. - Cost/benefit framing (the "capture" vocab fix): concrete before/after example supports the recommendation. #### Fallacies - **Anecdotal generalization (mild)** — "an absolute daily driver for me" and personal hotkey/accent preferences are individual experience presented as broadly applicable. Fix: note that ideal settings vary by user and workflow. - **Survivorship/best-case framing (mild)** — demos show the happy path (settings work as intended); accuracy limits and latency aren't stressed. Fix: include failure cases and how to recover. #### Tone of voice - Friendly, enthusiastic, instructional, peer-to-peer ("from newbie to dictation pro"); light humor ("don't tell our engineers… I'm giving you this tip"). Mild product promotion balanced by candid tips and acknowledged quirks. ### 🛡️ DARVO Analysis - Not applicable to this source. This is a procedural product tutorial with no accusation, dispute, or defense to deny, attack, or reframe. No Deny / Attack / Reverse-victim-offender dynamics are present. ## 🎯 Conclusion - **Bottom line.** Raycast dictation turns speech into a one-hotkey input method for any text field; the leverage comes from Styles (context-aware, per-app, translation, format-output), global custom instructions, and a custom vocabulary that tames accents and brand names. Free in beta, Pro after 2.0. - **Next steps for readers.** (1) Search "dictate," grant permissions, and set a memorable hotkey. (2) Add a global custom instruction for your accent or casing. (3) Create one Style for your most common context (e.g., markdown output for prompts, or a per-app name fix). (4) Add repeatedly misrecognized words to custom vocabulary. (5) Try dictating your next Claude Code prompt instead of typing it. ## 🔄 Refactor Appendix — How To Master Dictation in Raycast > Density: 2–5 Literature atoms (H3), 0–3 Permanent candidates (H4). > Defaults applied (short procedural source → 3 Literature, 0–1 Permanent). Resonance filter on. > > Tag pool: #bmf · #addendum · #xref · #articleseed · #storystem · #share · #aphorisms ### 📚 Literature Atoms (H3) #### [H1] Set a dictation hotkey so triggering becomes muscle memory > [!quote] ID:rw_pending > "it's highly encouraged to have a hotkey, so you can really leverage your muscle memory and make it second nature to just press it to dictate into any text field in any app" **Context & use**: - [00:00:34] Onboarding step; unlike most Raycast commands, dictation specifically benefits from a hotkey. - Why it matters: removes the open-and-search step, so voice input becomes reflexive — the precondition for daily use. **Tags**: #xref #share **Highlight**: [view highlight](readwise://pending) <!-- atomicity: 1 idea ✓ --> #### [H2] Styles set context-aware transcription instructions per app > [!quote] ID:rw_pending > "they basically allow you to set custom instructions for transcribing your dictations that will be used in specific contexts" **Context & use**: - [00:02:12] Styles auto-fire per app/website or run manually, each with optional hotkeys; can translate, format markdown, output LaTeX/IPA, or fix per-app names. - Why it matters: one dictation tool covers many contexts without manual reconfiguration. **Tags**: #xref #articleseed **Highlight**: [view highlight](readwise://pending) <!-- atomicity: 1 idea ✓ --> #### [H3] Add misheard words to custom vocabulary to fix accent errors > [!quote] ID:rw_pending > "I added 'capture' as a vocab word, and now it doesn't happen anymore." **Context & use**: - [00:07:05] Custom vocabulary is independent of style; good for brand names and repeatedly misrecognized words. - Why it matters: a low-effort, durable fix for the most common dictation frustration. **Tags**: #xref #aphorisms **Highlight**: [view highlight](readwise://pending) <!-- atomicity: 1 idea ✓ --> ### 🧠 Permanent Candidates (H4) #### [P1] Voice input becomes a habit only when friction drops below typing **Claim** (one sentence, Courtney's voice): A speech-input tool gets adopted as a daily driver only once a single dedicated trigger removes the launch friction, making speaking faster and more reflexive than reaching for the keyboard. **Evidence pointer**: [[#H1]] **Scope & conditions**: Holds for input methods competing with an entrenched default (typing); breaks where accuracy or latency is poor enough that error-correction friction exceeds typing friction. **Links**: - [[Muscle memory and tool adoption]] — because extends - [[Dictating LLM prompts]] — because applies-in the Claude Code prompting workflow - [[Friction reduction]] — because evidence-for <!-- atomicity: 1 idea ✓ --> ### ✏️ Fleeting Queue - [F1] Try dictating a real Claude Code prompt via a Raycast Style that outputs markdown — https://www.youtube.com/watch?v=RUh1KDgSOsc #fleeting #inbox - [F2] Open question: where does Raycast send dictation audio, and what are the privacy implications for client work? #fleeting #inbox ### → Things 3 (max 2 actions) - [ ] Set a Raycast dictation hotkey and add a global custom instruction for accent/casing — #things - [ ] Create one markdown-output Style and dictate the next Claude Code prompt with it — #things <!-- deep-read:sections end --> 📄 **Full text:** ![[Full Document Contents/Articles/How to Master Dictation in Raycast.md|How to Master Dictation in Raycast by Raycast]] ## Highlights - onboarding screen will show up, so you can grant the necessary permissions. And here, you can set a hotkey. Now, this is not true for every Raycast command, but for dictation specifically, it's highly encouraged to have a hotkey, so you can really leverage your muscle memory and make it second nature to just [music] press it to dictate into any text field in any app without having to open Raycast and search for it. If you've already dismissed this screen for some reason, you can also set your hotkey by ([View Highlight](https://read.readwise.io/read/01kvdbe93ws1yd0yjen76rezqq)) - Note: H1 — a dedicated hotkey drops launch friction so dictation becomes reflexive (relevant to dictating Claude Code prompts). - Tags: raycast, dictation, atom/literature/h1 - search for dictate in Raycast and press option command comma or control alt comma on Windows, you'll find its settings page. One of the most powerful features in Raycast dictation is styles. They basically allow you to set custom instructions for transcribing your dictations that will be used in specific contexts. The cool thing is you can have them fire automatically by assigning a style to specific apps and websites, but you can also pick a style manually on ([View Highlight](https://read.readwise.io/read/01kvdbearqe7gds6kqv1qxaz6w)) - Note: H2 — "styles" make dictation context-aware per app (custom transcription instructions that fire automatically). - Tags: atom/literature/h2 - mistake consistently. If that happens, then just add that word to the custom vocabulary. We actually have a command in Raycast that takes you directly to this part of dictation settings, so you can quickly edit the list. Like, for some reason, because of my accent, every time I said "capture", it would describe it as "capture". So, I added "capture" as a vocab word, and now it doesn't happen anymore. Let's add more padding to the "capture ([View Highlight](https://read.readwise.io/read/01kvdbedk94p2g6s75wpme07mn)) - Note: H2 — custom vocabulary is the low-effort durable fix for accent misrecognitions. - Tags: atom/literature/h2