Speak. Transcribe. Done.

Beautiful voice-to-text transcription for macOS, Windows, and Linux.
Privacy-first, lightning fast, and completely free.

🔒 100% Private
Lightning Fast
💰 One-Time Payment

How It Works

Three simple steps to perfect transcription

1
⌨️

Press Hotkey

Hold Cmd+Shift+Space (or your custom hotkey) from anywhere on your computer

2
🎤

Speak

Watch the live waveform as you speak. Release the hotkey when finished

3

See Text Appear

Your words instantly appear in any text field - no copying or pasting

Powerful Features

Everything you need for seamless voice-to-text

🎯

Native macOS UI

Beautiful Swift/SwiftUI floating pill that blends perfectly with macOS design language

🔐

Privacy First

All processing happens locally on your device using Apple's MLX Whisper. No cloud, no tracking

Lightning Fast

Transcribe in real-time with Apple Silicon optimization. 5-15x faster than real-time

🌍

99 Languages

Auto-detect or manually select from 99 supported languages worldwide

🌊

Live Waveform

Real-time audio visualization shows your voice being captured with smooth animations

📍

Flexible Positioning

Choose from 7 preset positions or drag the bubble anywhere on your screen

🤖

AI Assistant (Optional)

Say "echo action" to activate intelligent assistance - requires your own OpenAI, Anthropic, or local AI setup

🔧

Smart Tools

Access clipboard, files, web content, and selection with natural language commands

📚

Custom Vocabulary

Add technical terms, names, and acronyms for perfect accuracy in your domain

📜

History Search

Search and manage up to 1,000 past transcriptions with full-text search

⌨️

Modifier-Only Hotkeys

Use ergonomic combinations like Cmd+Ctrl+Alt without reaching for letter keys

🎨

Customizable UI

Choose themes, sounds, and animations to match your workflow

📱

Application Indexing

Automatically indexes installed applications for voice-launching via AI commands

AI-Powered Intelligence (Optional)

Say "echo action" to unlock powerful assistance

⚠️ Important: AI features require your own API key and account with third-party providers (OpenAI, Anthropic, Azure) or a local AI setup (Ollama). EchoCraft does not provide or include AI services. You are responsible for all API costs charged directly by the provider.

💬

Natural Language Commands

Use conversational language to get things done

"echo action, translate my selection to Spanish"
🌐

Web Research

Fetch and summarize web content instantly

"echo action, fetch this URL and summarize it"
📄

File Analysis

Read and analyze local files with context

"echo action, read my notes.txt and create bullet points"

Text Selection Tools

Work with highlighted text without copying

"echo action, elaborate on my selection"
💻

Code Generation

Generate working code in any language

"echo action, create a TypeScript debounce function"
🔍

History Search

Find past transcriptions with smart search

"echo action, search history for Python examples"
🚀

Application Launching

Open applications by voice command

"echo action, open Safari" or "echo action, launch VS Code"

Supported Third-Party AI Providers

You must create your own account and obtain API keys directly from these providers

OpenAI (GPT-4)
Anthropic (Claude)
Azure OpenAI
Ollama (Local - Free)

Note: Cloud AI providers (OpenAI, Anthropic, Azure) charge per API request. Typical cost: $0.005-$0.01 per command. Local models via Ollama are completely free but require local setup.

Choose Your Model

Select the perfect model based on your hardware and needs

💻 Choose Based on RAM

  • ≤2 GB RAM: Tiny or Base (English)
  • 2-4 GB RAM: Small (English) or Large v3 Turbo
  • 4-8 GB RAM: Medium (English) or Large v3 Turbo
  • >8 GB RAM: Large v3 (multilingual, best quality)

🎯 Choose Based on Use Case

  • General use: Base (English) - Recommended
  • Speed priority: Tiny or Base (English)
  • Maximum accuracy: Large v3 (multilingual)
  • Multilingual: Large v3 Turbo - Recommended

📊 Understanding Quantization

Each model is available in two quantization variants:

Standard

Full precision - best accuracy

Q8_0 (8-bit)

~50% smaller file size, slightly lower accuracy

💡 Recommended for most users

🇬🇧 English-Only Models

Whisper Tiny

78 MB 44 MB (Q8_0)
Speed: Ultra Fast 15-20x realtime
Accuracy: Fair
RAM Usage: 500 MB - 1 GB
  • ✅ Smallest download size
  • ✅ Minimal RAM usage
  • ✅ Perfect for testing
  • ⚠️ Lower accuracy

Whisper Small

488 MB 274 MB (Q8_0)
Speed: Fast 8-12x realtime
Accuracy: Very Good
RAM Usage: 2-3 GB
  • ✅ Very good accuracy
  • ✅ Reasonable speed
  • ✅ Good for production use
  • ⚠️ Larger download size

Whisper Medium

1.5 GB 844 MB (Q8_0)
Speed: Moderate 5-8x realtime
Accuracy: Excellent
RAM Usage: 3-4 GB
  • ✅ Excellent accuracy
  • ✅ Great for professional use
  • ✅ Handles complex audio well
  • ⚠️ Large download size

🌍 Multilingual Models

Most Accurate

Whisper Large v3

3.1 GB 1.7 GB (Q8_0)
Speed: Slow 3-5x realtime
Accuracy: Best
RAM Usage: 5-6 GB
Languages: 99 languages
  • ✅ Best accuracy available
  • ✅ Supports 99 languages
  • ✅ Handles accents and dialects well
  • ⚠️ Very large download
  • ⚠️ Requires powerful hardware

🔧 How Models Work

📥 Download Models

Models are downloaded on-demand from HuggingFace when you select them in Settings. They're cached locally for future use.

💾 Storage Location

macOS/Linux: ~/.cache/whisper.cpp/models/
Windows: %LOCALAPPDATA%\whisper.cpp\models\

🔄 Switch Models

You can download multiple models and switch between them anytime in Settings → Model tab. The active model loads into memory when you start recording.

📦 Bundled Models

Full builds include the Base model pre-bundled (~230 MB), so you can start using EchoCraft immediately without downloading models.

Works Everywhere

Native support for all major platforms

🍎

macOS

Native Swift UI for Big Sur and later

Intel & Apple Silicon

Download DMG →
🪟

Windows

Windows 10 version 1903 or later

32-bit & 64-bit

Download Installer →
🐧

Linux

Ubuntu, Fedora, and compatible distros

AppImage, DEB, RPM

Download Package →

Simple, Honest Pricing

One-time purchase. No subscriptions.

EchoCraft

$9.99 /forever
  • ✅ Unlimited transcriptions
  • ✅ All features included
  • ✅ No subscriptions
  • ✅ No cloud costs
  • ✅ 100% private
  • ✅ Free version 1.x updates
  • ✅ Professional support
  • ✅ Cross-platform compatibility

AI Assistant (Optional): AI features are completely optional and require your own third-party API key from OpenAI, Anthropic, or Azure. You create an account directly with these providers and are billed by them separately. EchoCraft does not provide, resell, or charge for AI services. Alternatively, use free local AI models via Ollama with no external accounts required.

* One-time purchase includes all version 1.x updates

Download EchoCraft

Get started in minutes

📦 Slim Build (Recommended)

Smaller download (~150-200 MB)

  • No bundled AI models - download models on first use
  • Faster initial download and installation
  • Choose models based on your needs
  • Perfect for users with limited bandwidth

📚 Full Build

Larger download (~250-300 MB)

  • Includes Base model pre-bundled (~230 MB)
  • Works immediately without internet connection
  • No model download required on first launch
  • Perfect for offline use or slower connections

🍎 macOS

For macOS Big Sur (11.0) or later

🪟 Windows

For Windows 10/11

🐧 Linux

For Ubuntu, Fedora, and compatible distributions

$9.99 - Buy License

Need help with installation?

Contact Support

Perfect For

💼

Professionals

Write emails, documents, and reports faster than typing

👨‍💻

Developers

Code comments, documentation, and commit messages in seconds

✍️

Writers

Capture ideas and draft content at the speed of thought

🎓

Students

Take notes, write essays, and complete assignments efficiently

🌍

Multilingual Users

Switch between 99 languages seamlessly

Accessibility

Alternative input method for those with typing difficulties

By the Numbers

99
Languages Supported
15x
Faster Than Real-Time
100%
Privacy Guaranteed
$9.99
One-Time Purchase

Frequently Asked Questions

Is my data private?

Yes! All transcription happens locally on your device using Apple's MLX Whisper. Nothing is sent to the cloud, and there's no telemetry or tracking.

Do I need an internet connection?

No! Voice transcription works completely offline. Internet is only needed for initial model downloads and optional AI assistant features.

How accurate is the transcription?

EchoCraft uses Whisper, one of the most accurate transcription models available. Accuracy depends on the model you choose and audio quality, but is typically 95%+ for clear speech.

Can I use it in any application?

Yes! EchoCraft injects text into any active text field - Chrome, Slack, VS Code, TextEdit, and more.

What about the AI features?

AI assistant features are completely optional. To use them, you must create your own account with a third-party provider (OpenAI, Anthropic, or Azure) and obtain an API key. You will be charged directly by that provider for API usage (typically $0.005-$0.01 per command). EchoCraft does not provide, include, or charge for AI services. Alternatively, you can use completely free local AI models through Ollama with no external accounts or costs.

What are the system requirements?

macOS 11.0+, Windows 10+, or Linux (Ubuntu 20.04+). Minimum 4GB RAM, 8GB recommended. 2GB disk space for installation and models.

Ready to Transform Your Workflow?

Get EchoCraft now for just $9.99 - one-time purchase, no subscriptions