Voice dictation for Replit means speaking your prompts instead of typing them: tap the microphone in Replit Agent’s chat box for built-in transcription, or add Genie 007 alongside it so every prompt, commit message, and code comment across Replit, GitHub, and Slack gets the same think-to-text treatment, in 140+ languages, from £5 a month.
Replit built a microphone into the Agent chat box in 2026, covering the most common need: getting a prompt in faster than you can type it. But it only lives in that one box. This guide covers how to use voice in Replit Agent’s built-in mode, what breaks it, the system-level options worth pairing with it, and where Genie 007 picks up the slack.
If you split your build time between Bolt and Replit, see our guide to voice dictation for Bolt for how Bolt’s own dictation compares to Genie 007 across both tools.
How to Use Voice in Replit Agent (Replit’s Built-In Voice Mode)
Replit’s own answer to “can I talk to Replit Agent instead of typing” is Voice Mode: a mic icon in the Agent chat box. Tap it, speak, and Replit transcribes your prompt into editable text before you send it.
- Open Agent inside the Project Editor, or from the prompt box on the home screen when starting a new app.
- Select the microphone icon next to the chat box and allow microphone access if asked.
- Speak your prompt — a waveform shows Replit is recording.
- Stop manually, or let it stop after roughly 30 seconds of silence or when you switch tabs.
- Review the transcribed text, fix anything it misheard, then send it like any typed message.
Transcription captures meaning rather than a literal word-for-word record, so filler words like “um” get dropped automatically. Recordings are capped at six minutes per take — Replit’s own documentation covers the exact limits. It’s free on every plan, on desktop web, mobile web, and the mobile app — a generous replit voice input option. The catch is scope: it only transcribes into the Agent chat box, not the editor, shell, commit messages, or PR fields.
Replit Voice Mode Not Working? Quick Fixes
Most “replit voice mode not working” reports trace back to permissions, not a bug:
- Microphone blocked. Re-enable microphone access for Replit in your browser’s site settings.
- Recording stopped after ~30 seconds of silence. Expected behaviour — keep talking or restart for a longer prompt.
- You switched tabs mid-recording. Voice Mode stops the moment you background the tab.
- Mobile permission. On mobile web or the app, the phone’s own settings need mic access granted, not just the browser.
- Managed devices. Some corporate policies block mic access site-by-site; check with IT if the icon is greyed out.
Talk to Replit Agent With System-Level Dictation
Because Replit’s voice input is scoped to the Agent box, developers often pair it with dictation built into their OS. Here’s how the realistic options compare if you want to talk to Replit Agent, and everything else, by voice.
| Tool | Price/month | Free plan? | Platforms | Works in | Best for |
|---|---|---|---|---|---|
| Genie 007 | From £5 | Yes — free trial, no card | Windows, Mac, mobile, browser (4 platforms) | Everywhere you type — Replit, GitHub, terminal, Slack, docs | Think-to-text prompts and code comments everywhere, in 140+ languages |
| Replit Voice Mode | Free (included) | Yes — all plans | Desktop web, mobile web, mobile app | Only the Agent chat box | A quick spoken prompt with nothing to install |
| Apple Dictation / Siri | Free | Yes — built into macOS/iOS | Mac, iPhone, iPad | Any native text field, system-wide | Mac/iPhone users dictating short prompts offline |
| Google voice typing | Free | Yes | Chrome (Docs); Gboard on Android | Google Docs; any field with Gboard | Drafting a spec in Docs before pasting into Replit |
| Windows Voice Access | Free | Yes — built into Windows 11 | Windows 11 | System-wide, including Replit’s browser tab | Windows users wanting hands-free navigation too |
| Dragon Professional Anywhere | ~$55 (~£43) | No | Windows (cloud) | Windows apps and browser | Enterprise teams needing custom vocabularies |
Genie 007 is the only option that works across all four platforms and follows you out of the Agent chat box into everything else you type — the built-ins are worth keeping for quick, no-install jobs, but none scale from prompt to commit to Slack reply the way Genie 007 does.
Apple Dictation/Siri run on-device, offline, but can pause mid-sentence and mishear names like “Supabase”. Google’s voice typing (Docs → Tools → Voice typing, or Gboard) covers 120+ languages for dictation, though commands are English-only (Google’s support documentation has the list) — fine for a spec, not code. Windows Voice Access adds system-wide dictation and navigation on Windows 11 but isn’t built for developer vocabulary. Dragon Professional Anywhere runs ~$55/month with team admin controls — overbuilt for prompting Replit Agent faster.
Dictate Replit Prompts on Mobile
Thumb-typing a detailed Agent prompt is where dictation earns its keep fastest. Replit’s Voice Mode already works on mobile web and inside the app, so its mic icon is the first thing worth reaching for. Outside the Agent box, your phone’s keyboard mic (Gboard, or iOS dictation) works in any field, including GitHub’s app. For that same experience across every app on your phone, the Genie 007 mobile app applies the same think-to-text cleanup on mobile that it does on desktop.
Where Built-In Replit Voice Input Hits a Ceiling
Replit’s Voice Mode and the system tools above solve one problem: getting words into a box faster than you can type them. A Replit session isn’t just prompting Agent — it’s a commit message, a GitHub PR reply, a README section. None of the built-ins follow you between those apps, so switching tools mid-task defeats the point of dictating at all. There’s also a transcription-versus-intent gap: built-ins give roughly what you said. A developer who wants replit agent voice input to save real time needs it cleaned into a precise spec, not typed word for word.
Genie 007: Voice Dictation for Replit That Works Everywhere
Genie 007 is a think-to-text voice assistant that runs across four surfaces — Windows, Mac, mobile, and a browser extension — active in Replit’s browser tab, your terminal, GitHub, and Slack, not just one chat box. Voice Typing handles fast, formatted dictation anywhere you’d normally type; Genie Mode turns a rough spoken thought into the finished text an app expects — an Agent-ready prompt, a commit message, a PR description — instead of a literal transcript. It’s on-device (AES-256, zero data retention, GDPR compliant, HIPAA ready) with 140+ languages and live translation.
Setting Up Genie 007 Alongside Replit
Install the Windows app, the Mac app, or the browser extension, sign in, and Genie 007 listens in the background for its shortcut. Click into the Agent chat box, editor, or a GitHub PR field, trigger it, and speak; Genie Mode turns the rough idea into a structured, Agent-ready prompt automatically. The same setup covers commit messages, PR replies, or a README during a vibe-coding session.
Spoken Prompt Cheat Sheet for Replit Agent
Detailed prompts get better output from Agent, and speaking is the fastest way to add detail without losing the thread:
| What you want to build | What to say out loud |
|---|---|
| Login page | “Build a login form with email validation, password strength rules, and a forgot-password flow that emails a magic link.” |
| Bug fix | “Find why the checkout API returns a 500 when the cart is empty, fix it, and add a test that covers that case.” |
| Database schema | “Add a Postgres table for orders with user_id, status, and created_at, then write a query that lists a user’s last 10 orders.” |
| Refactor | “Refactor the auth middleware to use JWT instead of session cookies, keep the existing error handling, and add rate limiting.” |
| Deploy prep | “Add environment variables for the Stripe keys, update the README with setup steps, and write a one-line commit message.” |
| Mobile-friendly feature | “Add a dark mode toggle to the settings screen that persists to local storage and matches the existing colour palette.” |
Genie 007 vs Replit Voice Mode: Which Should You Use?
Use Replit’s built-in Voice Mode for a quick, one-off prompt when you don’t want to install anything. Typing averages 40-50 words per minute; speech runs closer to 150 words per minute, so even that basic mic beats typing in full. Reach for Genie 007 once dictation is part of your real workflow — prompting Agent, writing commits, replying on GitHub — from one voice layer on Windows, Mac, mobile, or browser. See voice dictation for Claude Code or our think-to-text explainer.
Frequently Asked Questions
Does Replit have built-in voice dictation?
Yes. Replit Agent has a built-in Voice Mode: tap the microphone in the chat box, speak, and Replit transcribes your prompt into editable text before you send it. It’s free on every plan, on desktop web, mobile web, and the mobile app, but only inside that chat box.
Why is Replit’s voice mode not working?
Usually a permissions issue, not a bug. Check your browser and device have granted microphone access to Replit. Recording also stops after about 30 seconds of silence or if you switch tabs, which people sometimes mistake for a fault.
Can I dictate Replit prompts on my phone?
Yes. Replit’s Voice Mode works on mobile web and in the Replit app, so the mic icon works the same as on desktop. For dictation elsewhere — GitHub, Slack, notes — your phone’s keyboard mic or the Genie 007 mobile app covers the rest.
Does Genie 007 work with Replit?
Yes. Genie 007 runs as a Windows app, Mac app, mobile app, and browser extension, so it works inside Replit’s browser tab — the Agent chat box, editor, or terminal — the same way it works in GitHub, Slack, and everywhere else you type.
Windsurf users can apply the identical approach in Codeium’s AI IDE — read the guide to voice dictation for Windsurf.
Try Genie 007 Free
If Replit’s Agent mic hooked you on talking instead of typing, Genie 007 brings that same speed to everywhere else you write code.
Download Genie 007 free — available for Windows, Mac, mobile and as a browser extension. No credit card required.
Written by Bill Kiani, founder of Genie 007.
Non-coders splitting time between Replit and Lovable should read voice dictation for Lovable for the built-in-versus-Genie-007 breakdown.
Teams pairing Replit with a front-end generator like v0 can apply the same workflow described in our voice dictation for v0 guide.
