Just speak. Write faster.
Turn your voice into polished text, anywhere you can type.
Superwhisper is voice-to-text for macOS, Windows, and iOS. Speak naturally, and your voice becomes polished text, pasted right where your cursor is. Slack, Gmail, Cursor, Notion, your terminal, anywhere. No copying, no editing pass, no friction.
The free tier runs three on-device Whisper models with no time limit. Pro adds faster local models, cloud transcription, Super Mode with language models from OpenAI, Anthropic, Google, and more, plus meeting recording and file transcription.
| Platform | Link |
|---|---|
| macOS | Download for Mac |
| Windows (Intel) | Download for Windows x64 |
| Windows (ARM) | Download for Windows ARM |
| iOS | Download on the App Store |
One license covers all your devices.
- Press your hotkey. A global shortcut triggers recording from anywhere on your system. Push-to-talk (hold to record) or toggle mode (press once to start, again to stop).
- Speak naturally. Talk the way you normally would. Filler words, self-corrections, half-finished thoughts: it handles all of it.
- Text appears. Superwhisper transcribes, runs it through your chosen mode, and pastes the result.
No app switching, no clipboard dance, no editing pass.
Your audio never leaves your machine. Speech recognition runs entirely on your device, with no internet required. The free tier includes three Whisper models (Fast, Nano, Standard) that cover 100+ languages. Pro adds Parakeet V2 and V3 from NVIDIA for near-instant English dictation, plus the larger Whisper Ultra variants.
| Model | Languages | Size | Tier |
|---|---|---|---|
| Fast (Whisper) | 100+ | 75 MB | Free |
| Nano (Whisper) | 100+ | 150 MB | Free |
| Standard (Whisper) | 100+ | 500 MB | Free |
| Parakeet V2 (NVIDIA) | English | 476 MB | Pro |
| Parakeet V3 (NVIDIA) | 24 | 494 MB | Pro |
| Ultra Turbo V3 (Whisper) | 100+ | 1.6 GB | Pro |
| Ultra V3 (Whisper) | 100+ | 3.0 GB | Pro |
For highest accuracy or fastest turnaround, cloud models run through the Superwhisper proxy. Providers never see your account. Requests are anonymized.
| Model | Provider | Languages | Tier |
|---|---|---|---|
| Ultra | Superwhisper | 100+ | Pro |
| S1-Voice | Superwhisper | 100+ | Pro |
| Scribe V2 | ElevenLabs | 99 | Pro |
| Nova 3 | Deepgram | 36 | Pro |
| Nova Medical | Deepgram | English | Pro |
Super Mode goes beyond transcription. A language model rewrites your text, fixing grammar, adjusting tone, and formatting for context. An email comes out like an email. A code prompt stays technical. A casual Slack message sounds like you.
Choose from cloud language models (Claude Sonnet, GPT-5, Gemini, Grok, Llama) or on-device models that run offline through llama.cpp:
Cloud language models: Claude Sonnet 4.6, GPT-5.4 mini, Gemini 3 Flash, Grok 4.1 Fast, and more.
On-device language models: GPT OSS 20B, DeepSeek R1 Distill, Llama 3 8B, Mistral 7B, Phi-2 3B, and more. Runs fully offline.
Each mode shapes your output for a different context:
| Mode | What it does |
|---|---|
| Voice | Pure transcription, exactly what you said |
| Super | Context-aware rewriting that adapts to your screen |
| Message | Casual, conversational tone for chat apps |
| Professional formatting with greetings and structure | |
| Note | Clean, organized notes from stream-of-consciousness speech |
| Meeting | Record and transcribe meetings with automatic notes |
| Custom | Write your own system prompt: format as JIRA tickets, JSON, legal briefs, whatever you need |
When enabled, Superwhisper reads your screen and adapts its output to match your context. Dictating into a code editor gives different results than dictating into Gmail. No setup needed.
Hold your hotkey, speak, release. Text appears. Configurable per-mode keyboard shortcuts let you jump straight into the right mode.
Enter names, abbreviations, and specialized terms once. Superwhisper remembers them and gets your domain-specific language right every time. Medical terminology, codebase names, product acronyms, you name it.
Dictate in any of 100+ supported languages and dialects. Translate to English on the fly, or keep the original language.
Drop in audio or video files and get reliable transcripts. Process interviews, lectures, podcasts, or meeting recordings after the fact.
Use Superwhisper with Cursor, Claude Code, OpenCode, Amp, Codex, or any agentic coding app. Speak your prompts, describe bugs, dictate commit messages. Drive your entire workflow without touching the keyboard.
One tool across your devices. Most voice-to-text apps are macOS-only. Superwhisper runs on Mac, Windows (Intel and ARM), and iPhone, so your workflow doesn't break when you switch platforms.
Add your own API keys from OpenAI, Anthropic, Google, or any compatible provider. Use the models you already pay for.
On-device models mean your audio never leaves your machine. No logging, no data collection, no third-party access. Cloud requests go through the Superwhisper proxy, so providers never see your account, and nothing is retained for training. SOC 2 Type II certified and HIPAA compliant.
Superwhisper is built for teams that handle sensitive data:
- SOC 2 Type II certified and HIPAA compliant
- Centralized billing and authentication with SSO
- Model access control: choose which models your team can use
- Self-hosted model support: run against private endpoints or your own VPC
- Volume discounts for teams of any size
Contact us. We respond within 24 hours.
| Free | Pro | |
|---|---|---|
| On-device Whisper models (Fast, Nano, Standard) | Included | Included |
| Parakeet, Ultra, and all cloud transcription | — | Included |
| Super Mode with cloud and local language models | — | Included |
| File transcription | — | Included |
| Meeting recording | — | Included |
| Translation to English | — | Included |
| Custom modes (unlimited) | 3 modes | Included |
| 100+ languages | Included | Included |
| Custom vocabulary | Included | Included |
| Bring your own API keys | — | Included |
New users get 15 minutes of free access to all Pro features. See full pricing.
Full docs at superwhisper.com/docs. Want to contribute? See the docs repo.
| Platform | Minimum |
|---|---|
| macOS | 14.0 (Sonoma) or later, Apple Silicon recommended for local models |
| Windows | Windows 10/11 (Intel x64 or ARM) |
| iOS | 17.0 or later |
On-device models run best on Apple Silicon. Intel Macs work with cloud models.
Superwhisper is built on OpenAI Whisper, NVIDIA Parakeet, llama.cpp, and models from Anthropic, OpenAI, Google, Meta, Mistral, and others.
Made with care by the Superwhisper team.