Dictation for developers
AI Dictation for Coding: What Actually Works
Most developers who say voice coding mean talking to an AI agent and writing the prose around code. Here is what dictation does well, where it breaks, and which app fits.
Facts and prices read from vendor pages on October 8, 2026.
Voice dictation helps developers most when they speak natural-language prompts, documentation, comments and explanations. Dictating literal source code is harder, because punctuation, identifiers and formatting have to be exact. Tools differ in how much of that second job they attempt. Pepys Dictate does the first job: it types clean text into any Mac app, and it does not write or run code.
Searches for voice coding, dictation for developers and voice typing for coding mean four different things, set out below. Most people who dictate every day mean the first two: speaking to a coding agent and writing the prose around code.
Pepys Dictate is our product. This page compares it with Superwhisper, Wispr Flow, Willow Voice and Aqua Voice using each vendor's own pages, read on October 8, 2026, and says where they do more than Pepys Dictate does. We read the Pepys Dictate app's source for this page and have not tested it in Cursor or VS Code.
- For the general question of which dictation app to buy, see Best dictation apps and software.
- For leaving Superwhisper, see Superwhisper alternatives.
- For the plain steps for dictating into a chat box, see Dictate prompts to an AI assistant on Mac.
What people mean by voice coding
The same search covers four different jobs. Dictation apps are built for the first two. The third needs a code-aware tool, and the fourth needs a voice-control system.
| Meaning | What it sounds like | What it needs | Pepys Dictate (our app) |
|---|---|---|---|
| A. Natural-language prompts for an AI agent | “Refactor this function so errors include the request ID.” | Fast, clean dictation into a chat box or terminal prompt, with product and file names spelled right. | Fits. Types into the focused text field. No agent integration. |
| B. Prose around code | A pull request summary, a README section, a ticket, a Slack update. | Good punctuation and paragraphs, and technical terms spelled your way. | Fits. This is what the app is built for. |
| C. Literal source code | “const user name equals open paren id comma name close paren.” | Exact symbols, casing, indentation, quoting and escaping. | Limited. In terminals and code editors only, it turns casing commands and a few symbol words into text. It is not a complete code dictation system. |
| D. Full voice programming | “Select this function, move it below the class, run the tests.” | Commands that control the editor, tied to its structure. | Not supported. Pepys Dictate produces text and does not control an editor. |
Scroll sideways to see every column.
Can you use dictation for coding?
Yes, for the parts of coding that are language: prompts to an AI agent, pull request text, documentation, comments and messages. Dictation apps such as Pepys Dictate, Superwhisper, Wispr Flow, Willow Voice and Aqua Voice type that text into any app.
Dictating literal source code is harder, because the symbols, casing and names have to be exact. Pepys Dictate covers it only in part.
Is voice dictation good for programming?
Voice dictation is good for the language around programming and weaker for the syntax. Prompts, documentation, comments and explanations suit dictation. Literal code, shell commands and regular expressions do not, and moving around an editor by voice needs a voice-control system such as Talon.
Is dictating prompts easier than dictating source code?
Yes. A prompt tolerates a wrong word, and a coding agent such as Cursor or Claude Code turns the sentences into syntax. Source code needs exact symbols, casing and names.
Dictation apps, Pepys Dictate included, are built for the first job and cover the second only in part.
What developers actually dictate
Dictation pays off where the output is sentences, not syntax.
AI coding prompts
Prompts for Cursor, Claude Code, ChatGPT or an agent. A good one carries context, constraints and examples, which is a few paragraphs. Speaking them is quicker than typing them, and the agent handles the syntax.
Documentation
README sections, docstrings and architecture notes. Explaining a design aloud and then trimming it is often faster than staring at an empty file.
Pull requests
Summaries, review feedback and issue descriptions. The text is prose, and a dictated first draft is easy to correct.
Communication
Slack replies, email and status updates. This is ordinary dictation, and any dictation app handles it.
Debugging
Say what you expected, what happened and what you already tried, then add the stack trace by hand. The spoken version of a bug report is usually fuller than the typed one.
Code comments
A comment above a function is a sentence. Names of identifiers inside the comment are the part to check.
Dictating prose is not the same as dictating code
Prose tolerates a wrong word. Code does not. A dictation tool that is good at sentences can still get a line of source wrong, so each of these is a place to look before you accept the text.
- Punctuation and brackets: parentheses, braces, semicolons, colons and quotes have to be the plain characters the language expects.
- Indentation and whitespace: spoken new lines are easy, spoken indentation is not.
- Capitalisation and naming styles: camelCase, snake_case and PascalCase each need an exact form.
- Exact names: one wrong letter in a variable or function name is a different identifier.
- File paths and shell commands: flags, slashes, quoting and escaping, where a wrong character runs a different command.
- Language syntax: operators, arrows and generics that people do not say the same way twice.
What Pepys Dictate does about it
- Since release 1.0.23, in terminals and code editors only: say "camel case user name" and get userName. Snake case, Pascal case, kebab case and screaming snake case work the same way, and "end case" stops the name early.
- In the same apps, a few symbol words become symbols: open and close paren, bracket and brace, equals, double equals, triple equals, not equals, fat arrow, thin arrow and underscore.
- In those apps the formatting model is skipped, so only spoken punctuation, spoken layout commands and your own dictionary entries are applied. Commands and identifiers are not reshaped by a model.
- A Dictionary of names and terms, and “when I say, write” entries, for product names, packages and abbreviations. Ten entries are free, and the one-time Pro licence makes the list unlimited.
- A second shortcut, Paste raw, pastes exactly what the speech model heard, with no formatting and no developer syntax.
- An optional setting, Use words on my screen, is off by default. While you dictate, it reads the text of the focused field and the window title through macOS Accessibility (never a password field) and uses uncommon words in them as a temporary Dictionary for that one dictation. It keeps nothing and sends nothing. We have not tested which editors expose their text, or measured any accuracy gain.
What Pepys Dictate does not do
- It does not write or run code. It types the words you say.
- It does not index your codebase, read files or know which function you mean, and it has no editor integration, so it cannot tag files.
- English spoken punctuation covers comma, period, question mark, exclamation mark, semicolon and quotation marks. In ordinary apps spoken quotes come out as typographic curly quotes, which most languages do not accept; in terminals and code editors they come out straight (from 1.0.24).
- There is no spoken colon, slash, hyphen or indentation command, and no way to say a whole code block.
- Short commands still come out as sentences. A leading capital letter and a trailing full stop are normal, so a command such as git status needs fixing before you run it.
- File names and paths with extensions are a known weak spot. Check any name such as config.ts before you rely on it.
- In a code editor or terminal, the symbol words apply to everything you dictate there, including a prompt typed into that editor's chat panel, so “equals” becomes an equals sign. Use Paste raw when you want your words untouched.
- We have not tuned the speech model for programming vocabulary or measured it on it. A first benchmark is planned and no result is published here.
Where voice dictation fits a developer's day
A rough guide to the fit between voice dictation in general and each task. It describes the task, not one product, and it is no ranking.
| Developer task | Voice dictation fit | Why |
|---|---|---|
| Cursor prompt | Strong | Long natural language, and the agent turns it into code. A small speech error rarely changes the result. |
| Claude Code prompt | Strong | The same reasoning. Claude Code also has its own built-in dictation. |
| Explaining a bug to an agent | Strong | Spoken context is fuller than typed context. Add error text by hand. |
| PR description | Strong | Prose with a few technical terms. Easy to read through and fix. |
| README and documentation | Strong | Explaining a design aloud is quick, and the draft is easy to trim. |
| Ticket or issue text | Strong | Plain prose with names and steps. |
| Slack or email update | Strong | Ordinary dictation, and where any app does well. |
| Commit message draft | Useful | Short, so you read it all. Conventions with symbols, such as a type and scope prefix, need typing. |
| Code comment or docstring | Useful | A sentence about code. Check the identifiers inside it. |
| Prompt that names files, functions or flags | Useful | A dictionary entry helps with names you use often, but check every name. |
| Literal Python or TypeScript fragments | Limited | Casing commands and symbol words help for a name or a short expression. A whole function is slow and easy to get wrong. |
| Shell command | Limited | Flags, paths and quoting have to be exact, and a wrong character runs a different command. Dictated text arrives as a sentence, so read and fix it before you press Return. |
| Complex regex | Poor | Dense symbols with no natural spoken form. |
| Moving, selecting or running code by voice | Poor | This is voice control, not dictation. A command system such as Talon is built for it. |
Scroll sideways to see every column. Strong, Useful, Limited and Poor describe the task, not a product.
Pepys Dictate for developers who mostly talk to AI
The workflow is think, speak, prompt, then let the agent write the code. You describe the change in plain sentences, and you do not speak syntax.
“Refactor this function so errors include the request ID, then add tests for the timeout path.”
Pepys Dictate types that sentence where your cursor is. Your coding agent, not Pepys Dictate, reads it and decides what code to write or run.
- Click into the prompt field so the cursor is where the text should go.
- Hold the shortcut (Option+Space by default), speak, and let go. Or tap once to start and again to stop, or use the hands-free shortcut.
- Read the text. Fix a name or two. Send it yourself.
An Advanced setting can press Return after each paste. It is off by default, and it is better left off wherever a stray word could run something.
For prose you have already dictated, such as a pull request description, Edit by voice lets you select the text, hold a second shortcut and say what to change, for example make it shorter or fix the grammar. It is a Pro feature, also open during the trial, and it runs on your Mac. It edits prose, it is not code-aware, and we have not tested it in an editor.
Does Pepys Dictate work for coding?
For prompts, documentation and comments, yes: Pepys Dictate types clean text into any Mac app. Since release 1.0.23 it also turns casing commands and a few symbol words into text in terminals and code editors.
Pepys Dictate does not write or run code, has no coding-agent or editor integration, and has not been tested in Cursor or VS Code. It is Mac only.
What changes in terminals and code editors
Pepys Dictate recognises these apps by their identifiers and treats them differently from a notes app or a browser.
- Terminals: Terminal, iTerm2, Warp, kitty, WezTerm, Alacritty and Hyper.
- Editors: Visual Studio Code and VS Code Insiders, VSCodium, Cursor, Xcode, Zed, Sublime Text and the JetBrains IDEs.
What happens there
- Formatting is limited to the rules: spoken punctuation, new line and paragraph commands, web and email addresses, and your dictionary. The on-device model that fixes self-corrections, numbers and lists is not used, so “Friday, no wait, Monday” stays as you said it.
- The developer syntax (from 1.0.23) is applied: casing commands and symbol words.
- Pro users can set a style for a specific app, which overrides the rules-only default and brings the model back for that app. The developer syntax still applies.
Everything else, such as a browser-based tool or a chat app, gets the normal formatting.
Voice dictation in Cursor
In Cursor, text goes wherever the cursor is: the agent chat, an inline edit prompt, a comment or the integrated terminal. Pepys Dictate pastes into whichever field has focus.
We have not tested Pepys Dictate in Cursor. This is general text-field support, not a tested Cursor integration, and Pepys Dictate has no Cursor plug-in, no file tagging and no index of the open project. It does recognise Cursor as a code editor, so the rules above apply there, including to prompts.
Cursor also has its own microphone button in the agent chat input.
Can you use voice typing in Cursor?
Yes. Cursor's documentation describes a microphone button in the agent chat input. A system-wide dictation app such as Pepys Dictate, Superwhisper or Wispr Flow can also type into Cursor's fields.
Pepys Dictate has no Cursor integration and has not been tested in Cursor; it pastes into whichever field has focus.
Voice dictation in VS Code
VS Code now includes dictation. Its documentation describes an on-device model on desktop after a first download, used in chat, the editor and the terminal, with an optional cleanup step that sends the transcript text to a Copilot model.
Pepys Dictate has no VS Code extension and no VS Code-specific integration. It pastes into the focused field, including the integrated terminal. We have not tested it in VS Code.
If you only dictate inside VS Code, the built-in option may be enough. A system-wide app matters when you also dictate in a browser, a chat app and documents, and want one behaviour everywhere.
Can you use voice typing in VS Code?
Yes. Visual Studio Code documents built-in dictation for chat, the editor and the terminal, and an on-device model on desktop after a first download.
A system-wide app such as Pepys Dictate also types into VS Code's fields. Pepys Dictate has no VS Code extension and has not been tested there.
Claude Code and other coding agents
Claude Code has its own voice dictation, switched on with /voice. Its documentation says audio streams to Anthropic's servers for transcription and needs a claude.ai account. Superwhisper goes further with agent plug-ins: it pops up when an agent asks a question or needs approval, and you answer by voice.
With Pepys Dictate you dictate into the agent's prompt the way you would into any text field. We have not tested it in Claude Code, and it adds no voice answers, approvals or notifications.
Long natural-language instructions are where voice helps most with an agent: constraints, edge cases and the thing you already tried. The demand is still small. Searches that combine voice with a named coding agent run in the tens a month in the United States, so this page covers agents instead of splitting into one page per tool.
Can you dictate into Claude Code?
Yes. Claude Code has its own /voice dictation, which the vendor says streams audio to Anthropic's servers and needs a claude.ai account. Any system-wide dictation app can also type into the Claude Code prompt in a terminal.
Superwhisper adds a Claude Code plug-in for voice answers and approvals. Pepys Dictate dictates into the prompt and adds no plug-in.
Local-first dictation for developers
Pepys Dictate recognises speech and formats text on your Mac. After the models download once, it works offline, and by default the audio and text you dictate are not uploaded. An optional advanced setting can send a transcript to an AI provider you configure yourself; it is off by default.
That describes the dictation step only. Once the text is pasted into Cursor, Claude Code, ChatGPT or GitHub, that app handles it under its own terms. A prompt that goes to a hosted coding agent leaves your machine through the agent, not through Pepys Dictate. Local dictation does not make a hosted assistant local.
So if your prompts contain proprietary code, the destination matters as much as the dictation app. Check where your editor or agent sends what you give it.
Where each option processes your speech
- Pepys Dictate. On your Mac. Works offline after the models download.
- Superwhisper. The vendor states that it can run with no audio or text leaving your device when you pair a local voice model with a local language model, and that cloud models are optional per mode.
- Wispr Flow. The vendor states that transcription happens on its servers.
- Willow Voice. Cloud by default, per the vendor. Offline dictation is listed under paid plans.
- Aqua Voice. The vendor states that audio is processed in the cloud and that there is no offline mode.
- VS Code built-in dictation. On-device on desktop by default; an optional cleanup step sends transcript text to a Copilot model.
- Claude Code /voice. The vendor states that audio streams to Anthropic's servers.
Pepys Dictate and four other Mac dictation apps for developers
Every cell is Yes, No, Unknown or Not stated in public docs, with a short qualifier. Not stated means the vendor's pages do not say, which is not a claim that the feature is absent. Nothing is scored.
| Point | Pepys Dictate(our app) | Superwhisper | Wispr Flow | Willow Voice | Aqua Voice |
|---|---|---|---|---|---|
| Mac | Yes | Yes | Yes | Yes | Yes |
| Windows | No (planned, not available) | Yes (no agent plug-ins) | Yes | Yes | Yes |
| Local processing | Yes, speech and formatting | Optional, per mode | No (vendor states cloud) | Paid plan only (offline mode) | No (vendor states cloud) |
| Works offline | Yes, after models download | Yes, with local models | No | Paid plan only | No (vendor says no offline mode) |
| Natural-language prompt dictation | Yes, any Mac text field (not tested in Cursor or VS Code) | Yes (vendor states) | Yes (vendor states) | Yes (vendor states) | Yes (vendor states) |
| Coding-specific handling | Partial: rules-only formatting and spoken casing and symbol words in terminals and code editors (from 1.0.23) | Yes: Super Mode adapts to the app on screen (vendor states) | Yes: developer terms, camelCase, snake_case and acronyms (vendor states) | Not stated in public docs | Yes: technical vocabulary and CLI syntax (vendor states) |
| Cursor integration | No (general text-field support) | Yes, through its Agent Plugins pane | Yes: file tagging (vendor states) | Not stated in public docs (names Cursor as a place to dictate) | Dictates into Cursor; its MCP server connects to Cursor |
| Claude Code integration | No (general text-field support) | Yes: plug-in for voice answers, approvals and notifications | Not stated in public docs | Not stated in public docs (names Claude Code as a place to dictate) | Yes: command-line skill and MCP server |
| Terminal integration | No dedicated feature; pastes into the prompt; code-editor rules apply | Through its agent plug-ins | Works in terminals (vendor states) | Not stated in public docs | Terminal dictation listed (vendor states) |
| Model choice | Limited: speech model follows your language; one optional higher-accuracy formatting model (Pro) | Yes: local and cloud models, own API keys | Not stated in public docs | Not stated in public docs | Not stated in public docs |
| Custom vocabulary | Yes: Dictionary (10 entries free, unlimited with Pro); optional on-screen words | Yes: vocabulary and replacements | Yes: personal dictionary | Yes: auto-learning dictionary | Yes: custom dictionary |
| One-time purchase | Yes | Yes (Lifetime licence) | Not stated in public docs | Not stated in public docs | Not stated in public docs |
| Current price | $9.99 once (1 Mac); free 2,000 words a week | Free tier; $8.49 a month; $84.99 a year; $249.99 one-time lifetime licence | Free tier; $12 a month, billed annually; $15 a month, billed monthly | Free tier; $12 a month, billed annually; $15 a month, billed monthly | Free tier; $8 a month, billed annually; $10 a month, billed monthly |
| Primary trade-off | Mac only, one Mac per licence, no coding-agent integrations and no Cursor or VS Code plug-ins | Configurable and on four platforms, but AI features and cloud models need Pro | Four platforms and developer features, but cloud transcription and a subscription | Cloud by default and a subscription; coding use is described, not integrated | Agent hooks through MCP and a command-line tool, but cloud only and a subscription |
Scroll sideways to see every column. Rows about Pepys Dictate come from its source code and release notes. Rows about other products come from their own pages, read on October 8, 2026. A vendor marketing statement shows what it claims, not what we measured.
Pepys publishes this page and also makes Pepys Dictate, one of the apps compared. The other tools are competing products, so read our comparison with that in mind. Every competitor fact links to the page it came from.
Pepys Dictate compared with each one
Voice built into Cursor, VS Code and Claude Code
Three of the tools developers live in now have their own voice input. They are free with the tool, and each works only inside it.
| Tool | What it does | Processing | Needs | Read from |
|---|---|---|---|---|
| Cursor | A microphone button in the agent chat input dictates your prompt. Its documentation says to review the transcription before sending. | Not stated on the page we read | Cursor | Cursor documentation: prompting agents |
| Visual Studio Code | Built-in dictation into chat, the editor, the commit box and the terminal, plus a spoken Voice Mode for agent conversations. Terminal dictation adapts speech for command lines, for example by converting spoken symbol names. | On desktop, dictation uses an on-device model by default after a first download. An optional cleanup step sends the transcript text to a Copilot model; Voice Mode needs an eligible individual Copilot plan. | VS Code on Apple silicon for the on-device model | Visual Studio Code documentation: voice support |
| Claude Code | Turn on /voice, then hold a key or tap to dictate into the prompt. The vendor says transcription is tuned for coding vocabulary such as regex, OAuth, JSON and localhost. | The vendor states that audio streams to Anthropic's servers for transcription and is not processed locally. | A claude.ai account and a local microphone; not available in SSH or cloud sessions | Claude Code documentation: voice dictation |
Scroll sideways to see every column.
Where the other apps do more than Pepys Dictate
If developer tooling is the deciding factor, these are the places Pepys Dictate is behind.
- Superwhisper has dedicated coding-agent plug-ins for Claude Code, Codex, Cursor and others, voice answers and approvals, and deep model controls. Superwhisper may be the stronger fit if you want dedicated coding integrations and deep model controls.
- Wispr Flow tags files in Cursor and Windsurf, handles developer terms and naming styles, and runs on Windows and phones.
- Willow Voice names Cursor, Claude Code and Codex for prompt dictation and runs on Windows and iPhone.
- Aqua Voice lets an agent read your past dictations and change your dictionary through an MCP server, and adds a command-line tool.
- Talon is a hands-free input system with voice commands, eye tracking and scripting for macOS, Windows and Linux. It is the kind of tool the fourth meaning of voice coding needs.
Pepys Dictate may be the better fit if most of your coding dictation is really prompts, documentation and technical prose, and you want a simpler local-first Mac workflow at a low one-time price.
What is the best dictation app for developers?
There is no single best dictation app for developers, because it depends on the job. Superwhisper, Wispr Flow, Willow Voice and Aqua Voice describe developer features and agent hooks, and run on more platforms than a Mac.
Pepys Dictate fits developers who mostly dictate prompts and technical prose on one Mac and want local processing and a one-time price.
Is Superwhisper better for coding than Pepys Dictate?
Superwhisper may be the stronger fit for coding if you want dedicated coding integrations and deep model controls. Its documentation lists plug-ins for Claude Code, Cursor and other agents, and a choice of local and cloud models.
Pepys Dictate can fit developers who mostly want simple natural-language dictation on a Mac at a low one-time price.
What each costs
Compact only. The dated plan list for every app is on the buyer's guide, and the price here is read from the same records.
| Product | Price | Read on |
|---|---|---|
| Pepys Dictate | $9.99 once, plus tax where applicable, for one Mac. Free: 2,000 words a week. | October 8, 2026 |
| Superwhisper | $8.49 a month; $84.99 a year; $249.99 one-time lifetime licence. Free: unlimited dictation with local Whisper models, up to two modes without AI processing. | October 8, 2026 |
| Wispr Flow | $12 a month, billed annually; $15 a month, billed monthly. Free: 2,000 words a week on desktop and 1,000 on mobile. | October 8, 2026 |
| Willow Voice | $12 a month, billed annually; $15 a month, billed monthly. Free: unlimited use of a weaker model (Frontier Mini), limited personalization, 20 Scribe uses a week. | October 8, 2026 |
| Aqua Voice | $8 a month, billed annually; $10 a month, billed monthly. Free: 1,000 words and the Avalon model, no card. | October 8, 2026 |
Prices exclude tax. The Pepys Dictate price is a founding price that will increase. Competitor prices were read from each vendor's pricing page on October 8, 2026 and can change.
Choose by workflow
Choose Pepys Dictate if
- You use a Mac, mostly for prompts, documentation and technical prose.
- Local recognition and formatting matter to you.
- You want a simple workflow and a one-time price.
Choose a power-user coding dictation app if
- You want dedicated agent or IDE integrations.
- You need deep model configuration, or Windows and phone apps.
- You want voice approvals or terminal automation.
Choose a voice-control system if
- You want to move, select and edit code by voice.
- You need hands-free control of the whole computer.
Use built-in dictation if
- You dictate only occasionally, or only inside Cursor, VS Code or Claude Code, and the built-in option meets the need.
No one app suits every developer. The right choice depends on which of the four jobs you do most.
Try Pepys Dictate for Mac
Talk through the problem. Let your coding agent handle the syntax. $9.99 once, plus tax where applicable, for one Mac, with recognition and formatting on your Mac. Free for 2,000 words a week.
Apple silicon, macOS 13 or later. More about Pepys Dictate
What this page rests on
We read the Pepys Dictate app's source and release notes for this page. We did not run it in Cursor, VS Code or a terminal. Where a row says Not tested, treat it as general text-field support, not a tested integration.
| Claim | Basis |
|---|---|
| Types into any focused Mac text field | Read in the app source: paste with clipboard restore, a typing fallback, then the clipboard |
| Works in Cursor and VS Code text fields | Not tested. General text-field support, with both recognised as code editors |
| Works in terminals | Read in the app source, not tested. Seven terminals are recognised as code apps |
| Rules-only formatting and developer syntax in code apps | Read in the app source and the 1.0.23 release notes |
| Local recognition and formatting, offline after setup | Read in the app source and the privacy statement |
| Edit by voice on selected text | Read in the app source, not tested in editors. It is a prose editing tool, not a code-aware one |
| Use words on my screen (optional, off by default) | Read in the app source, not tested in editors. No accuracy measured |
| Accuracy on programming terms | Not measured. A benchmark is planned |
| Coding-agent or IDE integration, file tagging, voice control of the editor | None in the app |
What we have not measured yet
We have not run an accuracy benchmark on programming speech, so this page makes no accuracy claim for any app. A planned benchmark will cover the categories below, and its results will be added here.
- Programming terms
- Product and library names
- Identifiers (camelCase, snake_case)
- Code punctuation and symbols
- Cursor-style prompts
- Terminal commands
- Natural-language coding prompts
More questions about dictation for coding
Will spoken symbol words change my prompts in a code editor?
Yes. In a terminal or code editor, words such as equals and underscore become symbols, even in a prompt. Use the Paste raw shortcut when you want exactly what you said.
Does Pepys Dictate understand my project's names?
The ones you add to the Dictionary. The optional Use words on my screen setting (off by default) can add words from the focused field and the window title for one dictation, read through macOS Accessibility, kept in memory and never sent. It does not index your codebase or read your files.
Does it work in JetBrains IDEs and Xcode?
They are recognised as code editors, so the same rules apply. We have not tested them.
Can I use it for dictation because of an injury?
Pepys Dictate is not a hands-free programming system, but free licences are available for anyone who needs dictation because of an injury or disability. Email contact@pepys.co and say what you need.
Is there a Windows or Linux version for developers?
No. Pepys Dictate is Mac only today. Windows and Android versions are planned, not available, and not part of the Mac licence. An iOS keyboard is in development.
Can the dictation step send my text anywhere?
By default it does not upload your audio or text. An optional advanced setting can send a transcript to an AI provider you configure. What your editor or agent does with the pasted text is separate.
Sources checked October 8, 2026
Competitor and built-in-tool facts were read from their own pages, and Pepys Dictate facts from the app's source, release notes and our own product and pricing. Tell us if something here is out of date and we will fix it.
This page is based on each vendor's own pages and documentation, read on October 8, 2026, and on a read of the Pepys Dictate app's source and release notes (version 1.0.26). It reports no testing of our own in Cursor, VS Code or a terminal, so it says nothing about accuracy or speed. Prices are dated and exclude tax, and the Pepys Dictate price is a founding price that will increase. Marketing statements are attributed to the vendor.
Superwhisper
Willow Voice
Built-in voice
Talon Voice
Superwhisper: platforms, processing and price
Wispr Flow: platforms, processing and price
Willow Voice: platforms, processing and price
Aqua Voice: platforms, processing and price