Tap the microphone icon inside ChatGPT's message box (not the waveform icon) and start speaking — your words appear as editable text you can review before sending. This works in the ChatGPT browser app, the macOS and Windows desktop apps, and the mobile apps. FluidVox does the same job system-wide: hold your hotkey, speak, and text lands in that same box, plus every other text field on your device.
Use case · Updated July 2026
How to dictate to ChatGPT: Voice Mode vs. Dictation, limits, and faster prompts with FluidVox
ChatGPT already has two different voice features hiding behind two different icons, and mixing them up is the most common source of confusion. This page explains exactly what each one does, where they fall short, and how FluidVox adds fast, accurate, system-wide dictation into ChatGPT's browser tab, desktop app, and iPhone app — not just its message box.
Start 14-day free trialTL;DR
Use ChatGPT's built-in dictation if…
- You only ever dictate inside ChatGPT and nowhere else
- You're fine with no custom dictionary or per-app formatting
- You don't mind the mobile app's auto-send quirk
- You want it with zero setup, already built in
TL;DR
Use FluidVox if you want…
- Dictation that works in ChatGPT and every other app
- A personal dictionary for API names, acronyms, and jargon
- A style automatically matched to ChatGPT (Natural or Code)
- Text that never auto-sends — it lands in the field, editable
- On-device transcription with no cloud round-trip required
ChatGPT's own voice features
Voice Mode vs. Dictation — ChatGPT actually has both
Dictation (microphone icon)
Tap the small microphone icon inside the message box and ChatGPT converts your speech into text you can read, edit, and correct before sending — the same idea as speech-to-text in any other app. It's available in the browser at chatgpt.com, the macOS and Windows desktop apps, and the mobile apps, and it's free on every plan since it's just a way of filling in the same box you'd otherwise type into.
Voice Mode (waveform icon)
Tap the waveform icon instead and you get a live, two-way spoken conversation — you talk, ChatGPT answers out loud in real time, more like a phone call than a text composer. It doesn't show you an editable transcript first; the audio conversation is the interaction. On the free plan, Voice Mode is rationed to roughly 15 minutes a day as of mid-2026 (OpenAI doesn't publish an exact number, and it has changed before), after which you're bumped to a lighter experience until the daily reset.
If your goal is writing a careful, detailed prompt — not having a spoken conversation — dictation is the feature you want, and that's also exactly where FluidVox fits: a faster, more accurate way to fill that same box, plus every other text field you use.
Limits, pricing, and the auto-send bug
What actually limits ChatGPT's voice input
It's free, but not unlimited
Dictation itself costs nothing extra — it's included on ChatGPT's free tier and every paid plan. The catch is that whatever text you dictate still counts as one message against your plan's regular rolling message limit, and Voice Mode's daily minutes are separately capped for free accounts. Paid plans (Plus and above) raise both ceilings substantially but current published caps vary and OpenAI adjusts them periodically, so treat any specific number as approximate as of July 2026.
The 2026 auto-send regression
Throughout 2026, users on OpenAI's developer community forum have reported that the ChatGPT mobile app's dictation no longer reliably shows transcribed text in an editable box before sending — a detected pause in speech can trigger an immediate send instead, even with the auto-send setting turned off for some users. The practical workaround people report is tapping the stop button manually rather than waiting for silence to end the recording. It's a real, documented usability complaint, not a rumor — and it's exactly the kind of premature-submit risk FluidVox is built to avoid, since it always injects text into the field for you to review rather than treating a pause as a send signal.
Side-by-side
ChatGPT's built-in dictation vs. FluidVox
|
|
ChatGPT dictation | |
|---|---|---|
| Works inside ChatGPT | Yes | Yes |
| Works in other apps too | Yes — any app on Mac, Windows, iPhone | No — ChatGPT's own composer only |
| Custom dictionary / vocabulary | Yes, with auto-learned corrections | No |
| Per-app tone / formatting | 6 styles, automatic per-app matching | No — plain transcript |
| Auto-send risk | None — text lands in the field, editable | Reported on mobile app in 2026 |
| Daily voice minute cap | None | None for dictation; Voice Mode capped on free plan |
| Works offline | Yes — Local plan is fully on-device | No — requires an internet connection |
| Cost | $2.99/mo or $39 one-time | Free, included in every ChatGPT plan |
To be fair to OpenAI: their dictation is free, requires zero setup, and is genuinely good for quick prompts. FluidVox's case is for people who dictate into ChatGPT and a dozen other apps in the same day and don't want to learn a different voice tool for each one.
Setup
How to dictate to ChatGPT with FluidVox
Open ChatGPT
In your browser, the desktop app, or the iOS app — click into the message box like normal.
Hold your hotkey and speak
Fn on Mac, Ctrl+Shift+Space on Windows, or the FluidVox keyboard on iPhone. Describe your prompt in full sentences.
Release — text appears, editable
Clean, styled text lands in the composer. Nothing sends automatically; review it, then hit enter yourself.
Let per-app style handle formatting
Set ChatGPT to Natural or Code style once, and FluidVox applies it automatically every time you dictate there.
Smart Formatting
Better prompts through speaking
Speaking naturally tends to produce more detailed, nuanced prompts than typing shorthand — you describe constraints, context, and desired output the way you'd explain them to a colleague. FluidVox's Natural style keeps that conversational phrasing intact instead of over-compressing it.
Speed
Prompt roughly 3x faster
Most people speak far faster than they type. FluidVox streams your words in real time and cleans them up instantly, so long, detailed prompts take seconds instead of minutes to write out.
Accuracy
Technical context preserved
Add API names, framework terms, and acronyms to your personal dictionary once and FluidVox will always transcribe them correctly — something ChatGPT's own microphone icon has no equivalent for. Corrections you make are auto-learned over time, with an AI validation gate to keep bad guesses out.
A day dictating to ChatGPT with FluidVox
Research prompt
Dictate a detailed research question with specific constraints and desired output format, in Natural style.
Code review
Switch to Code style and voice-describe the code you want ChatGPT to review, including architecture context.
Writing assistance
Dictate a detailed prompt for content generation with tone and style guidance, hands-free with Fn+Space.
On the go, from an iPhone
Open ChatGPT's iOS app, switch to the FluidVox keyboard, and dictate the same way you would at your desk.
Dictate to ChatGPT — FAQ
It depends what you mean. ChatGPT can transcribe audio you upload or record in a conversation, and its Voice Mode can read responses back to you out loud. What it can't do is dictate your typing into other apps — its dictation only fills ChatGPT's own message box. FluidVox is built for that second job: it dictates into any app on your Mac, Windows PC, or iPhone, ChatGPT included.
In ChatGPT, "dictate" refers to the microphone-icon feature that converts your speech into typed text in the message composer, which you can edit before sending. It's distinct from Voice Mode, which is a spoken back-and-forth conversation where ChatGPT talks back. Both live behind different icons in the same input bar.
Dictation (microphone icon) turns your speech into editable text in the composer — you review it, then send it like any typed message. Voice Mode (waveform icon) is a live, two-way spoken conversation where ChatGPT talks back in audio and doesn't show a transcript-first composer step. Use dictation when you want to write a careful prompt; use Voice Mode when you want a quick spoken back-and-forth.
The most common causes are blocked microphone permissions in the browser or OS, an unsupported browser, or a known 2026 regression where the mobile app auto-sends your message on a pause instead of leaving it editable in the box, which people mistake for the feature being broken. Manually tapping stop instead of waiting for silence avoids most premature sends. FluidVox is unaffected by that bug — it always drops text into the field for you to review, never auto-submits.
Yes, dictation itself (the microphone icon) is free on every ChatGPT plan, though the message you send still counts against your plan's regular message limits. Voice Mode, the spoken-conversation feature, is more restricted on the free tier — around 15 minutes a day as of mid-2026, though OpenAI doesn't publish an exact figure and it can change. FluidVox's dictation has no daily voice-minute cap on any plan.
Yes, the ChatGPT iOS app has its own microphone-icon dictation. FluidVox also works here: its iPhone app includes a custom keyboard, so you can dictate into the ChatGPT app (or any other iOS app) using FluidVox's on-device transcription instead of Apple's or OpenAI's built-in speech recognition.
Several browser extensions add voice input to ChatGPT's website, but they only work inside the browser tab and stop the moment you switch to a native app. FluidVox isn't a browser extension — it's a system-level Mac and Windows app, so the same dictation works in ChatGPT's browser tab, its desktop app, your code editor, and everywhere else, with no per-site extension to install.
Dictation itself isn't capped by length, but very long recordings can time out or fail to transcribe cleanly in ChatGPT's interface, and whatever you send still counts as one message against your plan's rolling message limit. FluidVox has no dictation length cap — its hands-free mode (Fn+Space on Mac, Ctrl+Shift+H on Windows) is built for continuous, multi-minute dictation with real-time streaming.
Dictation is included free with every ChatGPT plan, including the free tier — it's not a separate paid feature. What costs money is ChatGPT itself (Plus, Pro, etc.) if you need higher message or Voice Mode limits. FluidVox is a separate one-time or subscription purchase — $2.99/month or $39 one-time for the on-device Local plan — that adds system-wide dictation to ChatGPT and every other app you use.
The ChatGPT API itself is text-in, text-out — OpenAI's separate Whisper and Realtime APIs handle speech, and you'd need to wire that up yourself in a custom app. That's a developer integration, not something the consumer ChatGPT app exposes. FluidVox solves the everyday version of this problem without any code: it dictates into whatever interface you're already using to talk to the API, including tools like Cursor or VS Code.
Not as a standalone product — but ChatGPT's Voice Mode and its "read aloud" option both generate spoken audio from text using OpenAI's text-to-speech voices. That's the reverse of dictation: text becoming speech, not speech becoming text. FluidVox only works in the speech-to-text direction — it doesn't generate audio.
Yes. FluidVox injects text into any focused text field on macOS and Windows, including the ChatGPT desktop app's composer, ChatGPT open in a browser tab, and any other app you switch to — all with the same hotkey.
Natural style suits most prompts — it keeps your conversational phrasing intact, which tends to give ChatGPT more context to work with. If you're pasting requirements or dictating around code, switch to Code style to keep technical terms and formatting intact. FluidVox's per-app style matching can apply your chosen style to ChatGPT automatically every time you dictate there.
Start dictating to ChatGPT today
Install FluidVox once and it works across ChatGPT, your browser, your code editor, and every other app. 14-day free trial, full access, no credit card.
Start free trial