1. Define what “free” must include
A permanent no-cost tool is different from a free trial, a capped plan, or open-source code that still costs setup time. Decide whether you need text in any supported Mac field, one specific document editor, imported-file transcription, spoken computer control, or a polished cross-app workflow.
Also separate dictation from transcription. Dictation creates text while you speak into the current workflow. File transcription starts with saved audio. A free engine that transcribes WAV files is not automatically a polished menu-bar app that inserts text at the cursor.
- Permanent free: no purchase after a trial or quota.
- Cross-app: works in more than one destination.
- Local: speech processing happens on the Mac for the active configuration.
- Turnkey: no compiler, model wiring, or custom text-insertion layer required.
2. Start with Apple Dictation for ordinary text
Apple Dictation is included with supported versions of macOS. Turn it on under System Settings → Keyboard, choose a shortcut, place the insertion point in a supported field, and speak. Apple’s current guide says Dictation can accept text of any length without a timeout and supports punctuation, new lines, paragraphs, and some formatting commands.
Do not assume every language and Mac uses the same processing path. Apple tells users to read the note shown beneath Keyboard → Dictation to see whether general voice inputs and transcripts are processed on the device for the current configuration.
This is the best default for someone who wants free basic voice typing and does not yet know which paid-app feature would justify another installation.
3. Use Megaphone for a finished free open-source app
Megaphone is a free MIT-licensed Mac app, not just an engine or source-code sample. Its current first-party documentation says Apple SpeechAnalyzer streams transcription on-device, while Apple’s on-device Foundation Models can remove fillers, resolve self-corrections, and power the optional “Hey Megaphone” inline instruction. A deterministic Basic cleanup mode remains available when Apple Intelligence is unavailable or too slow.
The finished workflow includes hold-to-talk or toggle activation, a personal Dictionary with automatic learning and JSON import or export, active-app and window-title context processed locally, voice macros, and multiple Apple-supported speech languages. The app requires macOS 26 on Apple silicon, downloads the selected Apple speech model on first use, and does not offer saved-file transcription in the product workflow documented here.
The trust tradeoff is explicit. Releases are signed, but the publisher says they are not notarized and may require clearing quarantine or choosing Open Anyway. People who do not want that installation step should build the inspectable source themselves or choose a notarized alternative.
4. Use Google AI Edge Eloquent when file transcription matters
Google’s June 3, 2026 launch announcement and current product page establish a native macOS Google AI Edge Eloquent app with on-device dictation across any application, a customizable hotkey, Voice Edit for selected text plus spoken commands, and local audio or video file transcription. Google also markets the current Mac download as “No cap” and “without subscription.”
That makes Eloquent a stronger no-subscription next step when the missing capability is not basic cursor input, but a dedicated cross-app dictation app, local file transcription, or selected-text editing. Apple Dictation is still the simpler built-in first test because it is already on the Mac and needs no extra installation.
One first-party caveat matters: the current product-page FAQ still contains older desktop-evaluation wording that conflicts with the dated launch announcement and direct Mac DMG. Treat the launch announcement and current downloadable Mac page as the authoritative platform evidence, and do not transfer iOS-only cloud-feature wording to the native Mac app.
5. Use Voice Control for navigation and editing commands
Apple Voice Control is the free built-in option when the goal extends beyond text entry. It can navigate the Mac, interact with interface elements, dictate and edit text, and run spoken commands.
Voice Control and standard Dictation are separate system modes. Apple says standard Dictation is unavailable while Voice Control is on, so choose the feature that matches the current task rather than trying to run both simultaneously.
A dedicated dictation app is not a substitute for Voice Control when accessibility-grade spoken navigation is the requirement.
6. Use Google Docs Voice typing for a free document workflow
Google Docs Voice typing is a strong no-cost choice when the destination is already a Google document. Google currently documents support in current Chrome, Edge, and Safari, along with English commands for selecting, formatting, editing, and navigating text.
The browser controls the speech-to-text service and its processing before text reaches Docs. Check the browser, site microphone permission, Workspace policy, and current Google documentation instead of treating this as a guaranteed local Mac workflow.
It is not a system-wide Mac dictation app. Its advantage is the document-specific command set inside Docs.
7. Use whisper.cpp only when open-source setup is part of the plan
whisper.cpp is an MIT-licensed C and C++ implementation of OpenAI’s Whisper model with first-class Apple-silicon support. Its official repository documents local model downloads, command-line builds, offline inference, and transcription of audio files.
That makes it a valuable free speech-recognition engine for technical users, but the repository is not a finished global Mac dictation product. A cross-app menu-bar workflow still needs microphone capture, shortcut handling, model management, permissions, error states, and safe text insertion.
Choose this route when owning and integrating the stack is worth the setup time. Choose a finished app when the goal is to start dictating rather than build the surrounding product.
8. Treat trials and capped plans honestly
A full trial can be useful for evaluating a paid workflow, but it is not a permanent free option. Compare the post-trial price, word or minute limits, account requirement, network path, and what stops working when the allowance ends.
IraVoice 0.7.6 offers a 3-day full trial with no card or account, then costs $50 once. It should not appear on a permanent-free list. Evaluate it only when local held-key recognition, polished text, or Voice-to-Spec solves a gap that Apple’s included tools do not.
9. Test the free path before paying
Dictate the same short paragraph in the actual app and field you use, then inspect names, punctuation, numbers, corrections, and insertion behavior. Repeat with a longer passage and the microphone you normally use.
If Apple Dictation already produces acceptable text, keep it. If Megaphone covers the finished free open-source job on your Mac, use it. If Google AI Edge Eloquent supplies local file transcription or Voice Edit, use it. If Voice Control supplies the commands you need, use it. Pay for another app only after naming the exact recurring friction it removes.
For sensitive work, verify the processing and retention boundary of the active tool and configuration. “Free,” “offline,” “on-device,” and “zero retention” are different claims.