How to Switch from Dragon to Hearsy: A Migration Guide

Switching dictation software after years of Dragon is harder than switching almost any other tool. Dragon isn't just software — it's a workflow built up through thousands of hours of dictation. You've trained it on your voice, built custom vocabulary, created Auto-Text commands, and developed muscle memory around phrases like "scratch that" and "cap that."

This guide is for people who've made the decision to move on. Maybe Dragon for Mac was discontinued and you've been limping along on an old machine. Maybe the $699 price for aging Windows software finally stopped making sense. Maybe you want local AI processing on Apple Silicon without enterprise overhead. Whatever the reason, this covers exactly what to bring with you and what to leave behind.

One disclosure: I make Hearsy. I've tried to write this guide honestly, including the places where Dragon did things Hearsy doesn't.

---

What Dragon actually gave you (and what matters most)#

Before migrating, it's worth being clear about which Dragon features you actually used regularly. Most Dragon users fell into one of two camps:

Power users with deep customization — thousands of Auto-Texts, custom command macros, years of acoustic training, specialized vocabulary for medical or legal terminology.

Regular dictation users — primarily used Dragon for continuous speech-to-text, maybe with a handful of Auto-Text shortcuts.

The migration path looks different for each. The good news for the second group: the switch is mostly painless. Hearsy's Parakeet TDT engine handles continuous dictation out of the box, with no training required. The accuracy gap that once justified Dragon's complexity has largely closed.

For power users, the migration takes more planning but is still straightforward once you've inventoried what you actually need.

---

Step 1: Export your Dragon custom vocabulary#

Dragon's Vocabulary Center stores the custom words and phrases you've added over time — specialized terminology, proper nouns, unusual spellings, domain-specific jargon. This is the most transferable part of your Dragon setup.

How to export:

1. Open Dragon and go to DragonBar > Tools > Vocabulary Center
2. Click "Export custom word and phrase list"
3. Choose a destination folder and filename
4. Select TXT format (most portable) or XML (preserves written/spoken form pairs)
5. Save the file

The exported TXT file is a plain list of your custom terms, one per line. Keep this file — it's your reference for terms that might need attention in Hearsy.

What to do with the list in Hearsy:

Hearsy uses Parakeet TDT and Whisper Large V3, both of which have broad vocabulary coverage that handles most professional terminology without explicit training. Go through your exported word list and test dictation of your most critical terms in Hearsy. In most cases they'll work correctly out of the box. Terms that don't, you can address through Hearsy's text correction workflow or by adjusting how you pronounce them for clarity.

Hearsy doesn't have a formal vocabulary import system — the underlying models handle most terminology without it. If you have highly specialized jargon (rare chemical compounds, obscure legal citations, unusual proper nouns), note those and test them specifically.

---

Step 2: Inventory your Dragon custom commands#

Dragon's custom command system — Auto-Texts, step-by-step commands, command macros — is the hardest part to migrate because there's no direct import path. Dragon commands are Dragon-proprietary and don't transfer to any other application.

First, export your command list for reference:

1. Go to DragonBar > Tools > Command Browser
2. Select Manage > Export (the export icon)
3. Choose XML format (portable and readable)
4. Save the file

Open the exported XML and review what's actually there. Most long-term Dragon users find they have many more commands than they actively use — Auto-Texts created once for a specific project and never touched again, macros for application shortcuts that have since changed, commands for software they no longer run.

Sort your commands into three buckets:

1. High-value, frequent — Auto-Texts you trigger every day (email signatures, standard disclaimers, frequently typed phrases, document templates)
2. Occasional — Commands you use a few times a month; worth rebuilding
3. Rarely or never used — Skip these entirely

Most users find 10–20 commands account for 80% of their actual command usage. Rebuilding those is an afternoon of work, not a weeks-long project.

Where to rebuild them:

For text expansion macros (the most common Dragon use case), a dedicated text expander works better than any dictation app's built-in system:

• Raycast (free, macOS) — trigger text snippets with keywords, system-wide
• Espanso (free, open source) — cross-platform text expander with YAML config
• TextExpander (subscription, $3.33/mo) — polished option with team sync

These work alongside Hearsy and are actually more reliable than Dragon's built-in command system for pure text expansion, since they're system-level tools that work in every app regardless of what you're dictating.

For document templates you dictate into regularly, Hearsy's AI enhancement feature handles this differently: rather than triggering a canned template, you dictate your content and Hearsy's LLM cleanup (via local Qwen 2.5 3B or Claude) can reformat it according to a prompt template you specify. The result is often cleaner than template insertion because it adapts to what you actually said rather than inserting boilerplate.

---

Step 3: Set up Hearsy#

Hearsy is a macOS menu bar app. Setup is straightforward:

1. Download and install Hearsy from hearsy.app
2. Run through onboarding — it requests Microphone and Accessibility permissions (Accessibility is required for the paste-to-app feature)
3. Choose your STT engine in Settings:
   • Parakeet TDT — faster (~200ms), better for real-time dictation, excellent English accuracy
   • Whisper Large V3 — slightly slower (1–2s), better for 99+ languages and accented speech
4. Set your hotkey — the default is customizable; pick something that doesn't conflict with your existing shortcuts
5. Configure AI enhancement (optional) — Hearsy can run a local Qwen 2.5 3B cleanup pass on your dictation, or use the Claude or OpenAI API if you prefer cloud-based cleanup

For most Dragon migrants, Parakeet is the right starting point. It's the closest to Dragon's real-time feel — text appears quickly as you finish dictating rather than after a multi-second processing delay.

---

Step 4: Adapt your dictation style#

This is where most Dragon veterans hit friction. Dragon trained you on specific verbal behaviors that don't translate directly to Hearsy:

Dragon commands that don't exist in Hearsy:

• "Scratch that" — deletes the last thing Dragon transcribed. In Hearsy, use Cmd+Z after pasting, or just rephrase and retype.
• "Cap that" / "all caps that" — Dragon would capitalize the last word on command. In Hearsy, type it correctly or use your text editor's case-change shortcut.
• "Select [phrase]" — Dragon could select arbitrary text by voice. Hearsy doesn't offer this; use standard keyboard shortcuts or mouse selection.
• "New line" / "new paragraph" — these work naturally in Hearsy by saying "new line" or "new paragraph," which Whisper and Parakeet both handle as punctuation cues when you pause appropriately.

What works better in Hearsy:

• Longer dictation sessions — Hearsy records until you release the hotkey, handling multi-sentence and multi-paragraph chunks cleanly. Dragon's continuous mode required you to keep speaking to avoid timeouts; Hearsy's press-and-hold model is more deliberate and often produces cleaner output.
• AI cleanup pass — if you enable AI enhancement, rough dictation gets polished automatically. Dragon transcribed verbatim; every "um," "uh," and rephrasing appeared in the output. Hearsy's optional LLM pass strips filler and reformats on the fly.
• No training required — Dragon improved with use, but fresh installs were noticeably worse. Hearsy's Parakeet and Whisper engines perform at full accuracy on the first use, with no correction sessions required.

---

What to expect on accuracy#

Dragon marketed "up to 99% accuracy right from first use." Independent reviews put real-world out-of-box accuracy closer to 95–97% for most users, improving with correction and training over time.

Parakeet TDT and Whisper Large V3 achieve comparable or better real-world accuracy without training, on most voices and microphones. The training advantage Dragon once had — years of personal acoustic adaptation — has been largely closed by transformer-based models trained on orders of magnitude more diverse audio data.

Where Dragon still has an edge: deeply specialized domain vocabulary in dedicated editions. Dragon Medical One and Dragon Legal have vocabulary databases built specifically for those fields. Parakeet and Whisper handle most medical and legal terminology well, but if you're a radiologist dictating 50+ anatomical terms per session or a lawyer whose work involves highly specific legal citations, test your specific vocabulary carefully before committing to the switch.

For general professional and personal use, the accuracy difference will be imperceptible within a few days of use.

---

What Dragon had that Hearsy doesn't#

An honest migration guide includes this. Dragon offered capabilities that Hearsy doesn't currently replicate:

Deep voice command system. Dragon let you voice-control your entire Windows environment — navigate menus, click buttons, scroll, select text by voice, control applications. Hearsy is focused on dictation-to-text; it doesn't offer full voice control of the Mac OS. For that, macOS Voice Control (built in, free) fills some of this gap, and Talon Voice fills more.

Custom vocabulary training depth. Dragon's Vocabulary Center allowed fine-grained control: setting spoken/written form pairs, adjusting recognition weights for terms, training specific words through repetition. Hearsy's underlying models handle vocabulary implicitly and don't expose per-word tuning.

Windows integration. Dragon had tight integrations with specific Windows applications — dictating into fields in Dragon Anywhere, Word integration for formatting commands, etc. Hearsy's integration model is simpler: press hotkey, dictate, paste. Powerful but less application-aware.

If any of these were core to your workflow, weigh them against the Hearsy benefits before switching. For most Dragon migrants coming from the Mac side, none of these are dealbreakers — Dragon for Mac was already discontinued and didn't offer most of these Windows-specific features anyway.

---

The migration in practice: a realistic timeline#

Day 1: Install Hearsy, grant permissions, run onboarding. Do a 30-minute dictation session. Note anything that doesn't work.

Days 2–3: Rebuild your 5–10 most-used Dragon Auto-Texts in Raycast or Espanso. This takes a couple of hours at most.

Week 1: Adjust dictation habits — breaking the "scratch that" reflex takes a few days. Accuracy will feel slightly different; give it a week before judging.

Week 2: Most users are at full productivity. The Dragon-specific habits (global commands, verbose formatting commands) are replaced by the simpler Hearsy workflow.

The emotional part — leaving software you've used for years — usually resolves faster than expected once you experience Parakeet's speed on Apple Silicon. Dragon's latency became normalized over years of use; the sub-200ms response of Parakeet feels qualitatively different.

---

Summary#

The vocabulary export takes 10 minutes. The command rebuild takes an afternoon. The dictation habit adjustment takes a week. After that, most people don't look back.

---

For context on what Dragon's current state looks like, see Dragon NaturallySpeaking in 2026: Is It Still Worth It?. For medical-specific alternatives, see Dragon Medical One alternative. For the broader Mac dictation landscape, see best dictation software for Mac.