Features

CyberWhisper voice-to-text interface on macOS showing real-time transcription and mode switching — local processing, 100+ languages supported
Push to Dictation Hands-free Mode Modes & Dictionary

How CyberWhisper Works

From speaking to polished text in three steps. Works in any app on your Mac.

Step 1

Speak Naturally

Press a shortcut or push a key, then speak freely. Your Mac captures audio entirely locally — nothing is sent to the cloud at this stage.

Step 2

AI Processes Locally

On-device AI models transcribe your speech in real time. Filler words, repetitions, and mid-sentence corrections are removed automatically.

Step 3

Text Appears Anywhere

Clean, formatted text is inserted directly into any app you're using — Notes, Slack, VS Code, Mail, or any text field on your Mac.

100% local processing
No audio uploaded
Works in any app
Quick Triggers

Two ways to speak — depending on how you think.

Choose your preferred shortcut method.

[Push to Dictation]

For short, precise inputs.

[Hands-free Mode]

For long thoughts and open-ended thinking.

Learn more →

Push to Dictation

Press & Release

Hands Free

Hands-free

Keyboard & Mouse Shortcuts

Fn

Keyboard

⌃⌥

Hands-free

Modes

The same words. The right output — everywhere.

Better context engineering with intelligent output adaptation across different applications.

Automatic context adaptation with the ability to define work modes for different contexts.

Use different language models, different voice models, and configure different output styles. Intelligently adapt appropriate outputs across different applications. Switch between different modes instantly using keyboard shortcuts for better context engineering.

Different scenarios require different output requirements: writing emails, replying to client messages, vibe coding, interacting with AI tools. Each scenario has specific format requirements and professional terminology enhancements. Different applications need different output formats, and different recipients (boss, client, colleague, family) require different tones and styles.

You say:

"Can you take a look at this?"

In Slack

Short & casual

In Email

Polite & structured

In IDE

Technical & concise

Applications

Voice

Slack

Message

Word

Email

IDE

Teams

Discord

Scenarios

Boss

Professional

Client

Formal

Colleague

Casual

Family

Friendly

Playful

Humorous

Creative

Expressive

100+ languages

Supports transcription and translation in over 100 languages, from widely spoken languages like English, Chinese, and Spanish to regional dialects and less common languages.

🇺🇸
🇬🇧
🇨🇳
🇯🇵
🇰🇷
🇩🇪
🇫🇷
🇪🇸
🇮🇹
🇷🇺
🇵🇹
🇳🇱
🇵🇱
🇹🇷
🇸🇦
🇦🇪
🇮🇳
🇹🇭
🇻🇳
🇮🇩
🇲🇾
🇸🇬
🇵🇭
🇧🇷
🇲🇽
🇦🇷
🇨🇴
🇨🇱
🇵🇪
🇻🇪
🇨🇦
🇦🇺
🇳🇿
🇿🇦
🇪🇬
🇳🇬
🇰🇪
🇸🇪
🇳🇴
🇩🇰
🇫🇮
🇮🇸
🇮🇪
🇨🇿
🇭🇺
🇷🇴
🇬🇷
🇺🇦

And you can switch languages on the fly.

Hybrid LLM Engine

Local when privacy matters.

Cloud when intelligence matters.

Your key when control matters.

OpenAI

OpenAI

Grok

Grok

Google

Google

Gemini

Gemini

DeepSeek

DeepSeek

Meta

Meta

Mistral AI

Mistral AI

Qwen

Qwen

Ollama

Ollama

On-Device Large Models

Voice models and large language models run entirely on-device

Complete data security, lower latency, lower cost.

You can customize and download different language models based on your needs. Some models guarantee low latency, ideal for scenarios requiring quick responses; others may have slightly higher latency but ensure higher accuracy, perfect for scenarios with high precision requirements. You can choose the model that best fits your needs for different scenarios.

Runs fully on-device

Biometric voice data never leaves your device

Millisecond-level latency

Biometric Data Protected

Latency vs Accuracy

Latency Low
Accuracy High

Low Latency

High Accuracy

Biometric Voice Data

Highly Sensitive • Fully Protected

Encrypted
On-Device
Adaptation

It adapts to you.

Why it gets better with use.

Dictionary

Vocabulary

→ Fewer corrections

Define custom vocabulary to greatly improve speech recognition accuracy and resolve spelling errors. Perfect for personal names, specific professional terms, and domain-specific vocabulary.

Snippets

→ Faster replies

Quickly define frequently used content such as postal addresses, product descriptions, and common message phrases. All vocabulary and snippets can be used across different applications.

Audio

→ Works anywhere

Works seamlessly across diverse scenarios and adapts to different input devices: AirPods, Bluetooth headphones, built-in microphones, laptop lid closed mode, wired microphones, and recording devices.

When your laptop lid is closed, the system intelligently adapts to the different microphone configuration, ensuring optimal audio capture even in clamshell mode.

Advanced noise reduction algorithms automatically filter out background noise in noisy environments like cafes, while preserving clear speech in quiet offices.

Intelligent silence detection handles long pauses and blank spaces in speech, automatically identifying when you're thinking, pausing, or finished speaking. The system understands natural speech patterns with frequent pauses and extended silence periods.

Dictionary

Personal Names

John Smith, Olivia Zhang, Neo Ruan, Jim Sang

Professional Terms

API, Kubernetes, CyberWhisper, LLM, Transformer, GPT, Claude

Input Devices

Built-in

Clamshell

Wired

Bluetooth

USB

Loopback Input

BYOK LLM integration

Bring Your Own Key (BYOK) support for LLM services. Use your own API keys from OpenAI, Anthropic, or other providers to maintain full control over your AI costs, usage, and data routing.

Perfect for teams and power users who need full control.

Bring your own key

OpenAI
OpenAI
Gemini
Gemini
Grok
Grok
DeepSeek
DeepSeek
Playground

See the difference — instantly.

Compare modes before you commit.

A place to shape how your voice becomes text.

Switch between different models, microphones, and parameters, see results change in real-time.

Compare output styles and speeds of different Modes. Experiment and save your final mode combination.

Mic
Whisper
LLM

Real-time output area

Ready to try CyberWhisper?

Speak your thoughts. Let the text follow.

Runs locally. No audio upload. No lock-in.

Frequently Asked Questions

Everything you need to know about CyberWhisper.

How does CyberWhisper work on Mac?

CyberWhisper runs as a background app on your Mac. Once activated (via keyboard shortcut or push-to-talk), it captures your voice, processes it locally, and inserts formatted text directly into any active text field — whether you're in Safari, Notes, Slack, VS Code, or any other app.

Does CyberWhisper work offline?

Yes. Core voice-to-text transcription works entirely offline using on-device AI models. Cloud AI features like advanced formatting, vocabulary enhancement, and cross-language translation require an internet connection.

What languages does CyberWhisper support?

CyberWhisper supports over 100 languages and dialects, including English, Chinese (Simplified & Traditional), Japanese, Korean, French, German, Spanish, and many more. You can also switch languages mid-session and translate between languages on the fly.

How is my privacy protected?

Your voice data never leaves your device during local processing. CyberWhisper uses on-device AI models for transcription. Only when you explicitly use cloud AI features (advanced formatting, translations) does data leave your Mac — and only to the AI provider you choose. You can also BYOK to maintain full control over your data.

Can I use my own API keys?

Yes. Pro and Enterprise plans support BYOK (Bring Your Own Key). Connect your own OpenAI, Anthropic, Google Gemini, DeepSeek, or any OpenAI-compatible API key. This gives you full control over costs, model selection, and data routing.

What apps does CyberWhisper work with?

CyberWhisper works anywhere you can type on your Mac — any app that accepts text input. This includes Safari, Notes, Mail, Slack, Discord, Notion, Obsidian, VS Code, Cursor, Linear, GitHub, and thousands more.