Searching for "free dictation app Mac" returns a wall of listicles recommending paid apps with free trials. An app that costs $15/month after 14 days isn't free. An app that limits you to 2,000 words per week isn't free in any meaningful sense. This is a guide to options that are actually free - no trial expiration, no credit card required, no word limits that matter.
Apple's built-in dictation
Every Mac has dictation built in. System Settings, Keyboard, Dictation. Toggle it on, press fn twice, and start talking. Text appears wherever your cursor is. Zero setup, zero cost.
By default, Apple sends your audio to Apple's servers. You can switch to on-device mode in the same settings panel, which processes everything locally at the cost of noticeably worse accuracy - especially for uncommon words and proper nouns.
The interaction model is the main weakness. The fn-fn toggle is clunky compared to a dedicated push-to-talk key. There are no voice commands beyond basic punctuation ("period," "new paragraph"). No clipboard management. No AI text processing. No way to extend it. Apple's on-device STT is the most private built-in option, but the accuracy gap versus their cloud mode is steeper than it should be in 2026.
Best for people who want basic dictation with zero setup and aren't particular about the interaction model.
Whisper.cpp
OpenAI's Whisper model compiled for Apple Silicon. Free, open source, fully local.
You clone the repo, compile with make, download a model, and run from the terminal. It processes audio files and outputs text. The accuracy is excellent with medium and large models - legitimately competitive with cloud services. Metal acceleration is well-supported, and the community is active.
The catch is that Whisper.cpp is a transcription engine, not a dictation tool. There's no push-to-talk hotkey. No system-wide cursor insertion. No way to dictate directly into another app without building your own wrapper. You record audio separately, run the CLI on the file, and manually paste the output. For batch transcription of meeting recordings or audio files, it's great. For "hold a key and talk into Slack," you're on your own.
Best for developers and technical users who want batch transcription and are comfortable building around a CLI tool.
MacWhisper (free tier)
A native Mac app that wraps Whisper models in a clean GUI. The free tier includes the tiny and base Whisper models. Drag an audio file in or record directly, and it transcribes locally on Apple Silicon.
MacWhisper bridges the gap between Whisper.cpp's power and most people's tolerance for terminal commands. The interface is simple and the local processing is fast. But the free tier is limited to the smallest models (lower accuracy), and there's no push-to-talk hotkey - you have to open the app and click record. The Pro tier ($29 one-time) unlocks larger models and real-time transcription, but the free version is batch processing only.
Best for people who want Whisper accuracy in a GUI for transcribing recordings, not for real-time dictation into other apps.
ActionPiper
On-device push-to-talk dictation, 142 voice commands, clipboard manager, and AI snippets. Free.
Download the DMG from modelpiper.com, install ToolPiper from the Mac App Store (free, provides the STT engine), hold Right Option, speak, release. Text appears at your cursor in any app. The Parakeet STT model runs on the Neural Engine with 140ms end-to-end latency. The model stays loaded as a keep-warm backend, so there's never a startup delay.
What sets ActionPiper apart from the other free options is scope. Push-to-command mode (Right Command) adds 142 macOS system actions - toggle dark mode, manage windows, control audio, adjust display settings, and more. The built-in clipboard manager tracks 200-2000 items with smart categories and OCR. AI snippets (;fix, ;formal, ;summarize) use a local LLM for text transformation. And 29 MCP tools let developers control their Mac from Claude Code, Cursor, or Windsurf.
The trade-offs: macOS only, 25 languages via Parakeet (English is strongest), and the Right Option/Right Command hotkeys aren't customizable yet. Requires ToolPiper running for voice features.
Best for Mac users who want a complete voice input system - push-to-talk, cursor insertion, voice commands, clipboard, and snippets - without paying for four separate apps.
Comparison
| Apple Dictation | Whisper.cpp | MacWhisper Free | ActionPiper | |
|---|---|---|---|---|
| Push-to-talk hotkey | fn fn (clunky) | No | No | Yes (Right Option) |
| Cursor insertion | Yes | No | No | Yes |
| On-device | Optional (lower accuracy) | Yes | Yes | Yes |
| Voice commands | Punctuation only | No | No | 142 system actions |
| Clipboard manager | No | No | No | Yes (200-2000 items) |
| AI text expansion | No | No | No | Yes (;fix, ;formal, custom) |
| Offline capable | On-device mode only | Yes | Yes | Yes |
| Setup difficulty | None | High (compile) | Low | Low (DMG + ToolPiper) |
| Accuracy | Good (cloud) / Fair (local) | Excellent (large models) | Fair (free tier) | Good (Parakeet) |
What about the paid options?
For context, the paid dictation landscape on Mac in 2026 looks like this. Wispr Flow ($15/month) has the most polished experience with 97%+ accuracy via cloud models and AI text cleanup, but sends all audio and screenshots to OpenAI/Meta servers. SuperWhisper ($249 one-time) runs Whisper locally with a nice GUI and push-to-talk. Raycast Pro ($8/month) includes AI text features and clipboard history, but the AI is cloud-dependent and there's no dictation.
If you're willing to pay, these are solid products. But think about what you're actually paying for. Someone else's GPU time to run a model that fits on hardware you already own. Your Mac has a Neural Engine that processes speech at 210x realtime. A $15/month subscription on top of that, to do the same work slower while sending your audio to someone else's servers, is a bet that local inference won't catch up. It's catching up fast.
Paying a subscription for speech-to-text is a transitional model. Local models get smaller and faster every quarter. The accuracy gap shrinks with each generation. This is headed the same direction as long-distance phone charges - a relic of when the infrastructure hadn't caught up to the obvious end state.
Our recommendation
If you want the simplest possible setup and don't care about privacy, Apple's built-in dictation in cloud mode is fine. It's already on your Mac.
If you want to transcribe recorded audio files with maximum accuracy and you're comfortable with the terminal, Whisper.cpp with a medium or large model is excellent.
If you want push-to-talk dictation that works everywhere, processes on-device, adds voice commands and clipboard management, and costs nothing, ActionPiper is the answer.
ActionPiper is part of the ModelPiper family of local AI tools for Mac. See also: Wispr Flow Alternative, Private Voice Dictation, Push-to-Talk AI on Mac.