Feature comparison

Dragon NaturallySpeaking set the standard for dictation software. Here is how it compares to EnviousWispr in 2026.

EnviousWispr Dragon (Nuance)
Price Free. No subscription, no license fee. ~$699 one-time (Professional)*
Account required No Yes, Nuance account + product activation
Platform macOS (Apple Silicon) Windows only (Mac version discontinued 2018)
Audio processing On-device (Neural Engine) On-device (after installation)
Audio leaves your computer Never. If you enable cloud AI polish, only the text transcript is sent. No (local processing)
Offline transcription Yes (after model download) Yes (after installation)
Offline AI polish Yes. Apple Intelligence or local Ollama models. No AI polish at all
AI polish / LLM providers 5 providers: Apple Intelligence, Ollama, OpenAI, Gemini, Groq None
Writing style control Context-aware Smart Polish adjusts for target app None. Raw transcript only.
Speech engines Parakeet TDT + WhisperKit Proprietary Nuance engine (decades of refinement)
Voice training required No. Works immediately. Improves with use; Nuance claims 99% from first use
Custom vocabulary Custom words for names and terms Custom vocabularies + macro commands + auto-learning
Voice commands Not supported Yes. Edit by voice, navigate, format, custom macros.
First-word capture Yes. VAD detects speech onset; no clipping. Yes. Always-listening mode with wake word.
Starts listening instantly Hotkey press, immediate recording. Always-listening or manual microphone toggle
Text lands in the right app Yes. Three-tier paste with app reactivation. Yes. Types into the active window cursor position.
Clipboard preservation Yes. Clipboard saved and restored after paste. Not applicable (types directly, does not use clipboard)
Auto-punctuation Yes. Neural model + AI polish handle punctuation. Yes. Built-in auto-punctuation.
Auto-stop on silence Yes. Configurable silence timeout. Yes. Configurable pause settings.
Accessibility Menu bar app; global hotkey; works in any text field Full accessibility suite; navigation by voice
Transcription latency 0.43s median; ~1.5s with AI polish Near-instant (local processing, but older engine)
Source code Source-available on GitHub (BSL 1.1) Closed source
Active development Yes. Shipping updates regularly. Effectively abandoned. Microsoft/Nuance focus shifted to enterprise and healthcare.

*Dragon Professional Individual pricing from nuance.com as of April 2026. Consumer Dragon Home has been discontinued. EnviousWispr latency from production PostHog data on Apple Silicon Macs. Dragon claims source-verified from official Nuance pages and product documentation; not firsthand-tested. Competitor claims last verified: 2026-04-04.

Download Free

Why Mac users choose EnviousWispr over Dragon

Dragon left the Mac behind. EnviousWispr was built for it.

$0
Free, not $700

Dragon Professional costs roughly $699 for a single license. EnviousWispr is free. No subscription, no license key, no activation server. Download and use it.

๐ŸŽ
Actually runs on Mac

Nuance discontinued Dragon for Mac in 2018. If you are on macOS, Dragon is not an option. EnviousWispr is built natively for Apple Silicon and takes full advantage of the Neural Engine. See how the pipeline works.

โœจ
AI polish built in

Dragon gives you a raw transcript. EnviousWispr can clean up filler words, fix punctuation, and restructure sentences using Apple Intelligence or a local Ollama model. No extra software, no cloud dependency unless you choose it.

โšก
Works instantly

Dragon claims up to 99% accuracy from first use, improving further with corrections over time. EnviousWispr works immediately. Modern speech models like Parakeet TDT are accurate out of the box.

๐Ÿ”„
Actively maintained

Dragon has not received meaningful updates since Microsoft acquired Nuance in 2022. Their focus shifted to enterprise healthcare. EnviousWispr ships updates regularly with new models, features, and performance improvements.

๐Ÿ“–
Auditable code

Every line is on GitHub under BSL 1.1. Verify what it does. Report issues directly. Dragon is closed source with no public bug tracker.

Two generations of on-device dictation

Both Dragon and EnviousWispr process speech locally. The difference is what happens after the transcript. For more on why on-device matters, read on-device vs cloud dictation privacy.

EnviousWispr
1
You speak into your Mac's microphone.
2
Audio is transcribed by the Neural Engine using Parakeet TDT or WhisperKit. Nothing uploaded.
3
AI polish cleans up the transcript. Runs on-device with Apple Intelligence or Ollama, or with your own cloud API key if you choose.
4
Polished text is pasted into your active app. Audio is discarded.
Dragon (Nuance)
1
You speak into your Windows PC's microphone.
2
Audio is transcribed locally by Dragon's proprietary engine. Processing stays on-device.
3
Raw transcript is inserted. No AI polish, no cleanup, no sentence restructuring.
4
You manually edit the text to fix filler words, punctuation, and phrasing.

The details that matter for Mac users

Dragon was the gold standard for Windows dictation. But two fundamental gaps make it irrelevant for Mac users in 2026.

๐ŸŽ
Actually runs on your Mac

Nuance discontinued Dragon for Mac in 2018. The last Mac version (6.0.8) predates Apple Silicon entirely. It does not run on modern macOS, and Nuance has no plans to bring it back. If you search for "Dragon Mac," you will find forum posts from frustrated users and no download link.

EnviousWispr is built natively for Apple Silicon. It uses Core ML and the Neural Engine directly, meaning transcription runs on dedicated ML hardware rather than general-purpose CPU cores. The app is a lightweight menu bar utility that launches in seconds and uses minimal memory.

This is not a compatibility layer or a port. Every component, from audio capture to paste-back, uses native macOS APIs: AVFoundation for audio, Accessibility API for text insertion, and Core ML for model inference.

Technical detail: EnviousWispr runs Parakeet TDT (600M parameters) and WhisperKit models through Core ML on Apple Silicon's Neural Engine. Dragon's proprietary engine was optimized for x86 Windows with no Apple Silicon path. The last Dragon Mac release cannot even launch on macOS 14+.
โœจ
Modern AI polish, not just raw transcript

Dragon gives you a transcript and expects you to fix it. It has no concept of LLM-based post-processing. Dragon was built in an era before large language models existed, and Nuance has not retrofitted this capability.

EnviousWispr integrates five LLM providers for Smart Polish: Apple Intelligence (fully on-device), Ollama (local models like Llama or Mistral), and cloud options via OpenAI, Gemini, or Groq if you bring your own API key. The polish step removes filler words, fixes punctuation, restructures awkward phrasings, and adapts to the app you are dictating into.

The system is context-aware. Dictating into Slack? The polish keeps it conversational. Writing an email? It formalizes. And all cloud providers only receive the text transcript, never your audio.

Technical detail: Smart Polish uses sandwich prompt framing with XML-tagged transcript boundaries to prevent hallucination. Output validation rejects responses longer than 3x the input. Short transcripts (3 words or fewer) bypass the LLM entirely. Preamble stripping removes "Certainly!" artifacts automatically.

Modern models, modern hardware

EnviousWispr runs neural speech models on Apple Silicon's dedicated ML hardware. The result is sub-second transcription with no training needed.

0.43s
Median transcription
From end of speech to raw text
1.5s
With AI polish
Apple Intelligence on-device
0s
Training required
Accurate from the first sentence

Based on production data from Apple Silicon Macs. Results vary by hardware and settings.

Where Dragon genuinely wins

Dragon pioneered voice dictation and spent decades refining it. These are real strengths, not legacy talking points:

๐ŸŽฏ
Decades of accuracy refinement

Dragon has been tuning its speech engine since the 1990s. Nuance claims up to 99% accuracy from first use, improving further as you make corrections. The engine learns your voice, your vocabulary, and your correction patterns over time. That adaptive accuracy loop is genuinely impressive.

๐Ÿ—ฃ๏ธ
Voice command editing

Dragon lets you edit, format, and navigate entirely by voice. Say "select previous sentence" or "bold that" and it works. You can create custom voice macros for repetitive tasks. EnviousWispr is dictation-only; it does not support voice-driven editing or navigation.

๐Ÿฅ
Medical and legal editions

Dragon Medical One and Dragon Legal Individual ship with specialized vocabularies for clinical notes, legal briefs, and industry-specific terminology. These are purpose-built products with EHR integrations. EnviousWispr targets general dictation, not regulated industry workflows.

๐Ÿข
Enterprise deployments

Dragon has decades of enterprise deployment experience with IT admin tools, network licensing, and compliance certifications. Organizations with existing Dragon infrastructure have switching costs that go beyond the software itself.

If you are on a Mac and want free, actively maintained dictation with AI polish, give EnviousWispr a try.

Common questions

Is there a free Dragon alternative for Mac?

Yes. EnviousWispr is free, runs natively on macOS with Apple Silicon, and processes speech on-device. Dragon discontinued its Mac version in 2018, so EnviousWispr fills that gap without costing $700.

Does Dragon still work on Mac?

No. Nuance discontinued Dragon for Mac in 2018. The last Mac version does not run on modern macOS with Apple Silicon. If you need dictation on a Mac, EnviousWispr is a current, actively maintained option.

Is Dragon dictation free?

No. Dragon Professional Individual costs approximately $699 for a one-time license. The consumer version (Dragon Home) has been discontinued. EnviousWispr is completely free with no license fee or subscription.

Do I need to train EnviousWispr to recognize my voice?

No. EnviousWispr uses modern neural speech models that work accurately from the first sentence. Dragon improves accuracy as you use it and make corrections, though Nuance claims 99% accuracy from first use. EnviousWispr skips that entirely.

Does EnviousWispr work offline?

Transcription runs entirely on-device and works without internet after the initial model download. AI polish can also run offline using Apple Intelligence or a local Ollama model. See the getting started guide for setup.

Can EnviousWispr do voice commands like Dragon?

EnviousWispr focuses on dictation: converting speech to text as fast as possible. It does not support macro commands or application control the way Dragon Professional does. If you need voice-driven automation, Dragon's scripting features are more extensive.

What happened to Dragon after Microsoft bought Nuance?

Microsoft completed its acquisition of Nuance in March 2022. Since then, Dragon consumer products have been effectively sidelined. The focus shifted to enterprise healthcare (DAX Copilot). Dragon Professional Individual is still sold but receives minimal updates.

Is EnviousWispr open source?

Source-available under the Business Source License 1.1. You can read, build, and inspect every line of code on GitHub. It is not an OSI-approved open source license, but it gives you full transparency. Dragon is entirely closed source.

Can EnviousWispr match Dragon's accuracy for medical or legal terms?

For general dictation, EnviousWispr's neural models are highly accurate out of the box. For specialized medical or legal terminology, Dragon's purpose-built editions (Dragon Medical One, Dragon Legal Individual) have purpose-trained vocabularies that EnviousWispr does not replicate. EnviousWispr's Custom Words feature lets you add names and terms manually, but it is not a substitute for a medical dictionary with tens of thousands of clinical terms.

I used Dragon on Mac before 2018. Is EnviousWispr a replacement?

For dictation, yes. EnviousWispr handles the core workflow Dragon Mac users relied on: speak, transcribe, paste into any app. It adds AI polish that Dragon never had. The main gap is voice commands. Dragon let you edit and navigate by voice; EnviousWispr is focused purely on converting speech to polished text.

Does EnviousWispr work with external microphones?

Yes. EnviousWispr uses whatever microphone macOS has selected as the input device. USB microphones, Bluetooth headsets, and audio interfaces all work. Dragon similarly supports external microphones on Windows. You can select your preferred input device in System Settings before recording.

Dragon left the Mac. We did not.

Free to download. No account required. No $700 license fee.