All four AI app builders document their own voice input for prompts. In Lovable, click the microphone or press Option-V; in Bolt, click Dictate or hold Control-Shift-D; in v0, click the microphone icon in the prompt bar; in Replit, click the microphone in the Agent chat box. Each one transcribes your speech into the prompt box and waits for you to send it. For prompts, the built-in button is the simplest choice. Apple's Dictation and system-wide apps matter when you also want voice in other fields or the same key everywhere.
Publisher note: SuprFlow publishes this guide and is one of the products discussed. Competitor details come from each company's own pages, checked on 3 October 2026. Prices and features change; confirm before you buy.
Which AI app builders have built-in voice input?
| Tool | How to start | What the docs say |
|---|---|---|
| Lovable | Microphone in the project chat, or Option-V on a Mac | Click the check when done; Lovable transcribes into the prompt box "removing filler words and cleaning up the text as it goes" (project chat) |
| Bolt | Dictate (microphone icon), hold Control-Shift-D, or right-click and hold | Text is added to the chatbox "but it doesn't send the prompt." Dictation "does use tokens from your allocation", even if you never send (Chat tools and shortcuts) |
| v0 | Microphone icon in the prompt bar, then the checkmark | "Your speech is converted to text and inserted into the prompt editor." Voice input requires being signed in (Text Prompting) |
| Replit | Microphone in the Agent chat box | Records up to 6 minutes; stops after about 30 seconds of silence or when you switch tabs; "nothing is sent automatically." Available on all Replit plans (Voice Mode) |
None of these four pages says where the audio is transcribed. Bolt says it has to "process your speech to transcribe it", which is why it uses tokens. If that matters to you, check each company's privacy policy before you speak client or company details.
Lovable and Replit both say they drop filler words. That makes a cleaner prompt, but it also means the text is not a word-for-word record of what you said. Read it before you send.
How do you dictate a prompt with Apple Dictation?
Apple's Dictation guide for macOS 27 says "you can speak to enter text anywhere you can type it." That includes these tools' prompt boxes in Safari or Chrome.
- Turn on Dictation once in System Settings → Keyboard → Dictation. On older macOS versions the wording may differ.
- Click in the prompt box so the cursor is blinking.
- Press the Microphone key or your Dictation shortcut, and speak.
- Stop with the same key. Read the prompt, then send it.
Apple Dictation is useful when you are writing somewhere the builder's microphone does not reach: a project description, a support form, a commit message, or the code editor. Setup is in How to use voice to text on a Mac. See the Mac dictation shortcut guide, Mac dictation commands and Mac Dictation not working.
When is a push-to-talk dictation app worth it?
If you only dictate prompts inside one of these builders, use its button. A system-wide app earns its place when you move between a builder, a code editor, a terminal, a chat app and email all day, and want one hold-to-talk key with the same dictionary everywhere.
SuprFlow is one example. It is a menu-bar app for Apple Silicon Macs on macOS 14 or later. Hold the Fn key, speak, release, and it pastes the text into whichever text field has focus. Speech recognition runs on the Mac with an on-device model; optional cloud formatting sends transcript text, not audio, to the provider you pick. Default formatting is local cleanup of punctuation, capitalization and numbers. A personal dictionary helps with library names and acronyms. It costs $125 for lifetime access on up to 5 Macs you own, with no subscription and a 7-day refund on a first purchase (as of 3 October 2026).
Because it pastes at the cursor, it should work in an ordinary text field; test it in each builder's prompt box with a harmless sentence. It does not send the prompt for you. We have not run a controlled benchmark against any of these built-in features.
Which route should you use?
| Builder's own voice input | Apple Dictation | System-wide app (SuprFlow as the example) | |
|---|---|---|---|
| Prompt box | Yes, in all four | Yes | Should work; test first |
| Other fields and other apps | No | Yes | Yes; test first |
| Cost | Included; Bolt counts tokens | Free with macOS | $125 once |
| Start | Click the microphone, or a shortcut in Lovable and Bolt | Microphone key or shortcut | Hold Fn, release to insert |
| Cleans up filler words | Lovable and Replit say so | No | Default formatting tidies punctuation and numbers |
| Best for | Prompting one builder | Occasional prompts and forms | Voice across many apps a day |
What goes wrong when dictating prompts?
- Enter sends. Lovable says to "press Enter or click the send button", and Bolt's homepage chatbox starts building when you press Enter. Apple says the spoken "new line" is "equivalent to pressing the Return key once", so in these boxes it may send a half-finished prompt. Pause and type a line break by hand if the box supports one.
- Two dictation tools at once. Do not start Apple Dictation while a builder's microphone is recording. Use one at a time.
- Shortcut clashes. Bolt's hold-to-talk is Control-Shift-D and Lovable's is Option-V. If you set a custom Dictation shortcut or another app's hotkey, check it does not collide.
- Code words. File names, package names, environment variables and error messages are better pasted than spoken.
- Tokens and credits. Bolt's transcription uses tokens even if you discard the text. Lovable and v0 bill prompts by credits; check your plan.
- Silence. Replit stops after about 30 seconds of silence; Apple Dictation also stops after 30 seconds without speech.
What should you check before you send the prompt?
- That the transcript says what you meant, especially if filler words were removed.
- Component, page and field names, spelled exactly as in your project.
- Negatives such as "don't change the header".
- No API keys, passwords or customer data spoken into the prompt.
- The mode: Lovable's Build mode changes code, while Chat and Plan do not.
On privacy: your prompt becomes part of the project's chat. Lovable lets you copy a link to any message "to show a collaborator or the support team", and Bolt restores earlier versions from version history in the chat. See where dictated text goes after insertion.
Questions people ask
Does Lovable have voice input?
Yes. Click the microphone in the project chat or press Option-V on a Mac, then click the check to transcribe.
Does Bolt dictation cost tokens?
Bolt says yes. Transcribing your speech uses tokens from your allocation, separate from the tokens for the prompt.
Is there voice input in v0?
Yes. Click the microphone icon in the prompt bar, then the checkmark. You must be signed in.
Does Replit Voice Mode read answers aloud?
No. Replit says Voice Mode turns your speech into text in the chat box and does not read Agent's replies aloud.
Related guides
- Vibe coding by voice on a Mac
- How to dictate prompts to Claude Code, Cursor and other AI coding tools
- Dictate to ChatGPT or Claude on a Mac
- How to use voice dictation in VS Code on a Mac
- Teach SuprFlow names and acronyms
- How to use voice to text on a Mac