Speech-to-text for VS Code: dictate prose, not syntax
Developers write more prose than they realise. SpeechFlow puts clean dictation in VS Code — commit messages, docs, AI prompts — at your cursor. Free to start.
Every day as a developer you write thousands of words that aren't code: commit messages, docstrings, inline comments, README sections, PR descriptions, and prompts to Copilot or Cursor. Your hands are already tired from typing logic — dictating that prose instead can cut the friction in half. SpeechFlow lets you hold a key, speak, and get clean, punctuated text right at your cursor inside VS Code.
The prose problem in a developer's day
A rough audit of a typical coding session tells the story: meaningful commit messages, thorough docstrings, useful inline comments, clear PR descriptions, and thoughtful prompts to your AI assistant together easily account for 30–50% of what you actually write. None of it is syntax. All of it benefits from fluent, complete sentences. And yet most developers type it out, slowly, in stolen minutes between context switches.
Apple's built-in dictation can drop raw audio-to-text anywhere, but it won't strip your filler words, add punctuation, or adapt tone. You end up spending as long cleaning the output as you would have spent typing — which defeats the point. See the broader dictation for developers on Mac guide for the full picture.
How SpeechFlow works inside VS Code
SpeechFlow is a native macOS app (Apple Silicon, ~50 MB). It operates at the system cursor level, so it has no VS Code extension to install and nothing to configure. Here's the full flow:
- Open VS Code and position your cursor — in the editor, the integrated terminal, or a Copilot/Cursor chat panel.
- Hold Control and speak naturally. Pause, ramble, restart a sentence — it's fine.
- Release. A cleanup LLM strips fillers (“um”, “like”, false starts), adds punctuation, and inserts the finished text exactly where your cursor sits.
- Keep coding. The keyboard shortcut doesn't leave VS Code focus, so you never context-switch.
Because it inserts at the cursor rather than via a clipboard, it works in every VS Code surface: the editor buffer, the terminal, the Git commit-message box, the search/replace field, and any chat panel from GitHub Copilot, Cursor, or Continue.
What to dictate in VS Code
Honest framing first: dictation is not the right tool for writing raw code. Syntax, brackets, indentation — the precision required makes voice input slower than typing. Where dictation pays off is every piece of prose that lives around the code:
| What you write | Why dictation helps |
|---|---|
| Commit messages | Speak a complete, descriptive message in five seconds instead of hammering out a terse one-liner under pressure. |
| Inline comments | Explain why a block works the way it does — the kind of prose you skip when typing feels slow. |
| Docstrings | Describe parameters and return values in full sentences; clean output means you skip the edit pass. |
| README & docs | First-draft a full section by talking through it; faster than staring at a blank buffer. |
| PR descriptions | Summarise the change, the reasoning, and the test plan by voice directly in the VS Code Git panel. |
| AI prompts | Dictate detailed, precise context to Copilot Chat or Cursor — richer prompts, better output. |
Technical terms and code-switching
A fair question: can dictation handle words like useState, fetchUserById, or --force-with-lease? Mostly, no — and that's fine. The workflow isn't “dictate everything.” It's a fluid switch: hold Control when you're writing prose (the comment body, the commit subject, the Copilot prompt), release it, then type the code as normal. You stay in one file, one window, with one hand on the keyboard. The context-switch cost is near zero. For a deeper look at how this pairs with AI coding tools, see dictation for Cursor and the AI dictation overview.
Privacy and pricing
- Zero data retention — no audio, no text stored on SpeechFlow servers.
- BYOK — bring your own OpenAI, Gemini or Groq key; audio goes straight from your Mac to your provider, nothing in between.
- Free — 2,500 words/week, no credit card. Pro is €10/month or €70/year. BYOK is €69 once, lifetime.
FAQ
Does SpeechFlow need a VS Code extension?
No. It works at the macOS system level and inserts text at the cursor in any app — VS Code, the terminal, or a browser-based editor — with no plugin required.
Can I dictate code syntax, not just prose?
You can, but it's rarely faster than typing. Dictation shines for natural-language prose: comments, docstrings, commit messages, documentation, and AI prompts. For raw code, keep typing.
Does it work in the integrated terminal and Copilot chat panel?
Yes. SpeechFlow inserts wherever the cursor is in VS Code — editor buffer, terminal, Git commit box, or any chat panel like GitHub Copilot, Cursor, or Continue.
Does it store my voice or code context?
No. SpeechFlow retains zero data. In BYOK mode your audio travels directly to your chosen provider (OpenAI, Gemini or Groq) and nothing passes through or is stored on a SpeechFlow server.
Is there a free plan?
Yes — 2,500 words per week, no credit card needed. Pro (€10/month or €70/year) removes the limit. BYOK is €69 once.
Stop typing your commit messages. Try SpeechFlow free — 2,500 words a week, no card required.