For years, voice assistants have promised the future. They would answer trivia, set timers, and tell you the weather. But when it came to real work -- the kind that fills your day, demands your focus, and earns your living -- they fell short. And when it came to privacy, they fell off a cliff.
We built Yaps to change both of those things.
Yaps is a privacy-first, offline voice assistant for macOS that processes everything on your device. No cloud. No subscriptions. No internet connection required. Your voice never leaves your Mac.
Today, we are making it available to the world.
Why We Built a Voice Assistant That Works Offline
The voice technology industry has a trust problem. A 2024 survey found that 40% of voice assistant users worry about who is listening to their conversations. And those worries are not unfounded.
Google paid a $68 million settlement over recording private conversations through its voice assistant. Fireflies.AI was sued for illegally harvesting biometric voice data without consent. Over 276 million healthcare records were breached in 2024 alone, many through cloud-connected productivity tools. Even Apple's built-in Dictation sends audio data to Apple's servers by default.
The pattern is clear: if your voice data touches a server, it is at risk. Every cloud API is a potential breach point. Every uploaded audio file is a liability.
We believe there is a better way. Yaps processes all speech recognition and text-to-speech entirely on your Mac using on-device machine learning models. Your audio is never uploaded, never stored remotely, and never used to train anyone's AI. The simplest way to protect voice data is to never let it leave the device in the first place.
What Is Yaps?
Yaps is a native macOS voice assistant that lives in your menu bar and works with every application on your Mac. It provides speech-to-text dictation, natural text-to-speech, voice notes with automatic transcription, a studio editor, voice commands, and smart history -- all processed locally on your device.
It is built as a true macOS app, not an Electron wrapper or a web view. It starts in under a second, uses less than 200MB of memory, and integrates with macOS Shortcuts and system preferences.
Here is what Yaps does:
Speech-to-Text Dictation That Actually Works
Hold the Fn key and speak. Your words appear wherever your cursor is -- polished, punctuated, and ready. Whether you are composing an email in Mail, writing a document in Pages, drafting code comments in VS Code, or filling out a form in Safari, Yaps turns your speech into clean text that reads like you typed it.
The accuracy is built on models trained on real-world dictation patterns -- the pauses, the corrections, the stream-of-consciousness flow that makes human speech messy and beautiful. Yaps does not just transcribe. It understands context, adds punctuation naturally, and adapts to your speaking rhythm.
Because everything runs on-device, there is zero latency from network round trips. You speak, the text appears. No buffering, no waiting for a server to respond, no degraded performance when your WiFi drops.
Natural Text-to-Speech
Select any text, hold Option+Fn, and hear it read aloud in a natural, warm voice. Not the flat, mechanical voice you might expect from a screen reader. These are voices with rhythm and personality, generated locally using neural speech synthesis.
Hearing your writing read back to you is one of the most effective editing techniques available. You catch errors your eyes skip over. You notice rhythm problems that look fine on screen. For anyone processing large volumes of text -- researchers, lawyers, students, writers -- this feature alone is worth the download.
Voice Notes with Instant Transcription
Press Ctrl+Fn and capture a thought. No opening an app, no finding the right note, no friction. Speak, and Yaps saves it -- timestamped, searchable, and always accessible.
These are not audio recordings you will never revisit. They are full text transcriptions, which means you can search them, copy them, and integrate them into your workflow. Think of voice notes as a voice-powered scratch pad that never runs out of pages and never needs you to re-listen to find what you said. We have written extensively about why voice notes are the best way to capture ideas, including practical workflows for meetings, brainstorming, and writing first drafts.
Studio Editor
The Yaps Studio gives you a dedicated space to work with voice-generated content. Edit transcriptions, refine dictated text, and polish your voice notes in an environment designed specifically for voice-first workflows.
Voice Commands
Control your Mac with your voice. Yaps voice commands let you trigger actions, navigate between apps, and automate repetitive tasks -- all hands-free, all processed locally.
Smart History
Every dictation, voice note, and text-to-speech session is saved in a searchable history. Find what you said last Tuesday. Re-use a dictation from this morning. Your voice work builds a personal archive that stays on your device and remains private.
How Yaps Compares to Other Mac Dictation Apps
The macOS voice assistant landscape has grown in recent years, and choosing the right tool matters -- especially when privacy and workflow integration are priorities. Here is how Yaps compares to the other options available today.
Yaps vs. Cloud-Based Voice Assistants
Cloud-dependent voice tools like Wispr Flow offer speech-to-text by routing your audio to external servers (typically OpenAI or Meta infrastructure) for processing. This approach has inherent trade-offs.
Audio sent to external servers. Requires internet. ~800MB RAM. $10-15/month subscriptions. Privacy depends on policy language.
Audio never leaves your Mac. Works fully offline. Under 200MB RAM. No subscription required. Privacy guaranteed by architecture.
Privacy considerations: When audio is sent to third-party cloud servers, you are trusting multiple organizations with your voice data. Some cloud-based tools also capture screenshots of your active window for "context understanding," which means your screen content -- including sensitive documents, private messages, and proprietary code -- may also be transmitted externally. Original privacy policies for some of these services have included provisions for using customer content for AI model training, though policies can change over time. The risks are more significant than most people realize -- your voice carries biometric data, emotional state, and health indicators, as we detail in our article on why your voice data is more sensitive than you think.
Performance and resource usage: Cloud-dependent voice tools typically require significantly more system resources. Independent reports on cloud-based alternatives show memory usage around 800MB (compared to Yaps' sub-200MB footprint), CPU usage of 8% even when idle, and startup times of 8 to 10 seconds. Because Yaps runs locally with optimized native code, it starts in under a second with minimal system impact.
Offline availability: Cloud voice tools simply do not work without an internet connection. If you are on an airplane, in a coffee shop with unreliable WiFi, working from a cabin, or operating in any environment without stable connectivity, cloud-based dictation stops entirely. Yaps works everywhere your Mac goes, regardless of network conditions.
Cost: Many cloud-based voice assistants use subscription models (commonly $10-15 per month) that create ongoing costs. Yaps offers its core features without requiring a monthly subscription.
Yaps vs. Offline-Only Dictation Tools
There are a handful of other offline dictation tools for macOS, such as ParaSpeech, which offers on-device speech recognition at a one-time price. These tools validate the core premise that on-device processing is the right approach for privacy.
However, most offline dictation tools are limited to dictation only. They do not include text-to-speech, voice notes, a studio editor, voice commands, or smart history. If you need a complete voice workflow rather than a single feature, the scope difference is significant.
Some offline tools also note that "future cloud features are not included in the lifetime license," which suggests a potential shift toward cloud dependency over time. And while marketed as fully offline, some collect anonymous telemetry data by default -- a small but meaningful distinction for users who prioritize complete privacy.
Yaps provides the full spectrum of voice productivity features while keeping everything on-device. No telemetry. No cloud upsell. No asterisks.
Yaps vs. Apple Dictation
Apple's built-in Dictation has improved significantly, especially on Apple Silicon Macs. It is free, it is integrated, and recent versions offer on-device processing for many languages.
That said, Apple Dictation is a system-level feature, not a productivity tool. It does not offer voice notes, text-to-speech, voice commands, a studio editor, smart history, or a customizable workflow. And by default, Apple Dictation sends audio to Apple's servers for processing unless you specifically enable on-device mode in System Settings.
Yaps is designed for people who want voice to be a primary input method for serious work, not an occasional convenience feature.
Is Yaps Safe? Understanding On-Device Voice Processing
This is a question we hear often, and we welcome it. Trust should be earned, not assumed.
Here is exactly what happens when you use Yaps:
- Audio capture: When you activate dictation (Fn key), your microphone captures audio through macOS system APIs.
- On-device processing: The audio is processed by machine learning models running locally on your Mac's CPU and Neural Engine. No network requests are made.
- Text output: The recognized text is inserted at your cursor position or saved as a voice note.
- Audio disposal: The raw audio data is released from memory. It is not saved to disk, not uploaded, and not retained.
At no point does audio data leave your device. There is no "phone home" behavior, no analytics on your speech content, and no mechanism for remote access to your voice data. You can verify this yourself -- Yaps works identically with your network connection completely disabled.
This architecture is possible because modern Apple Silicon chips (M1, M2, M3, M4) include a dedicated Neural Engine capable of running speech recognition models locally at speeds that match or exceed cloud-based alternatives. The era when cloud processing was necessary for good voice recognition is over.
You can verify Yaps' privacy yourself: disconnect from the internet entirely and use every feature. It all works identically, because no network connection is ever needed.
Who Is Yaps For?
Yaps is built for anyone who works on a Mac and wants to use their voice more effectively. Some specific groups that benefit most:
Writers and content creators who dictate drafts, use text-to-speech for editing, and capture ideas on the fly with voice notes.
Professionals handling sensitive information -- lawyers, healthcare workers, financial advisors, therapists -- who cannot risk client data being transmitted to external servers.
Remote workers and travelers who need dictation that works reliably on airplanes, in hotels with poor WiFi, and in locations without stable internet access.
Developers who want to dictate comments, documentation, and messages without switching away from their editor. Our practical guide to voice input for developers covers specific workflows for commit messages, PR descriptions, code reviews, and more.
Accessibility-focused users who rely on voice input and audio output as primary interaction methods and need tools that are fast, reliable, and always available. This includes professionals managing RSI, carpal tunnel, tendinitis, or other repetitive strain conditions where reducing keyboard use is part of injury management. If that describes your situation, our guide on voice input as assistive technology for RSI and repetitive strain injuries covers how to build a voice-first workflow around an existing injury.
Privacy-conscious professionals who have read one too many headlines about voice data breaches and want a tool that respects their boundaries by design.
Built Natively for macOS
We chose to build Yaps exclusively for macOS because focus enables excellence. Rather than building a mediocre cross-platform tool, we built something that feels like it belongs on your Mac.
Yaps is a true native macOS application. It is not built with Electron, not running a hidden browser, and not wrapping a web view. This means:
- Under 200MB memory usage -- a fraction of what Electron-based alternatives consume
- Instant startup -- ready in under one second, every time
- Menu bar integration -- always accessible, never in the way
- macOS Shortcuts support -- automate voice workflows with Shortcuts
- System preference respect -- follows your appearance, accessibility, and language settings
- Apple Silicon optimization -- takes full advantage of the Neural Engine on M1/M2/M3/M4 chips
What Does "Privacy First" Actually Mean?
We use the phrase "privacy first" deliberately, and we want to be specific about what it means in practice:
- No audio is ever uploaded to any server, under any circumstances
- No telemetry is collected on your speech content, usage patterns, or voice characteristics
- No account required to use Yaps -- we do not even have your email address unless you choose to give it to us
- No third-party analytics SDKs are included in the app
- No "anonymous" data collection that could be de-anonymized
- No future cloud features that would compromise the local-first architecture
- Works with network disabled -- unplug your ethernet, turn off WiFi, and Yaps functions identically
This is not a privacy policy buried in legal language. It is an architectural decision baked into the software. The data never leaves your device because the software is designed so that it cannot.
Most apps promise privacy through policy. Yaps guarantees it through architecture -- the software simply has no mechanism to send your data anywhere.
What Is Next for Yaps
Today's launch is the beginning. We are actively working on:
- Voice cloning -- use your own voice for text-to-speech playback
- Meeting transcription -- live, on-device transcription for calls and meetings
- Deeper app integrations -- tighter connections with the tools you already use
- Expanded language support -- bringing on-device voice processing to more languages
- Custom voice commands -- define your own voice-triggered workflows
All future features will maintain the same privacy architecture: everything on-device, nothing uploaded, no compromises.
Getting Started with Yaps
Download Yaps for free from yaps.ai and experience what voice-first productivity feels like when privacy is not an afterthought.
Setup takes less than a minute. Grant microphone access, and you are ready. No account creation, no API keys, no subscription billing -- just a voice assistant that works.
Frequently Asked Questions
Does Yaps require an internet connection?
No. Yaps works entirely offline. All speech recognition and text-to-speech processing happens on your Mac using local machine learning models. You can use Yaps on an airplane, in a location without WiFi, or with your network connection completely disabled.
Is Yaps free?
Yaps offers a free tier with core voice assistant features. Visit yaps.ai for current pricing details on premium features.
What macOS versions does Yaps support?
Yaps is built for modern macOS and is optimized for Apple Silicon (M1, M2, M3, M4) Macs. Check yaps.ai for specific system requirements.
How does Yaps compare to Wispr Flow?
Yaps and Wispr Flow take fundamentally different approaches. Wispr Flow sends audio to cloud servers for processing and requires a monthly subscription ($15/month). Yaps processes everything on-device, works offline, uses significantly less memory (under 200MB vs. ~800MB), and does not require an internet connection. If privacy and offline availability are priorities, Yaps is designed specifically for those needs.
Does Yaps record or store my voice?
No. Yaps processes audio in real-time and releases it from memory immediately after transcription. No audio is saved to disk, uploaded to any server, or retained in any form. Only the transcribed text is saved (in your voice notes and history), and that data stays on your Mac.
Can I use Yaps for dictation in any app?
Yes. Yaps works system-wide with any macOS application. Wherever you can place a text cursor -- email clients, word processors, code editors, web browsers, messaging apps -- Yaps can insert dictated text.
Is on-device speech recognition as accurate as cloud-based alternatives?
Modern Apple Silicon chips include a Neural Engine specifically designed for machine learning tasks like speech recognition. On-device models have reached parity with cloud-based alternatives for most use cases, and they eliminate the latency introduced by network round trips. Many users find that Yaps feels faster because there is no waiting for a server response.
What is the best offline dictation app for Mac?
Yaps is designed to be the most comprehensive offline voice tool for macOS, combining speech-to-text, text-to-speech, voice notes, a studio editor, voice commands, and smart history -- all running on-device. While other offline dictation tools exist, most offer only single-feature dictation without the broader voice productivity workflow that Yaps provides.
How much memory does Yaps use?
Yaps uses less than 200MB of RAM, which is a fraction of what many voice tools require. For comparison, some cloud-based voice assistants use 600-800MB of RAM and consume noticeable CPU resources even when idle. Yaps is designed to be lightweight enough to leave running all day without impacting your other work.
Does Yaps work with Apple Silicon and Intel Macs?
Yaps is optimized for Apple Silicon (M1, M2, M3, M4) to take full advantage of the Neural Engine for on-device speech processing. Check yaps.ai for the latest compatibility information.
Your voice is your most natural, most powerful communication tool. It should not require a cloud server, a subscription, or a leap of faith about who might be listening.
Your voice has been waiting. Let it lead.