Skip to content

DandaAkhilReddy/Speakskiptype

BranchesTags

Repository files navigation

SpeakSkipType

🎤 SpeakSkipType

The ULTIMATE Speech-to-Text Tool for Developers
No API. No Cloud. 100% FREE. Works Locally.

Quick Start Features Install Claude Code

Python Platform License Tests 

  ____                   _    ____  _    _       _____
 / ___| _ __   ___  __ _| | _/ ___|| | _(_)_ __ |_   _|   _ _ __   ___
 \___ \| '_ \ / _ \/ _` | |/ \___ \| |/ / | '_ \  | || | | | '_ \ / _ \
  ___) | |_) |  __/ (_| |   < ___) |   <| | |_) | | || |_| | |_) |  __/
 |____/| .__/ \___|\__,_|_|\_\____/|_|\_\_| .__/  |_| \__, | .__/ \___|
       |_|                                |_|         |___/|_|

🗣️ Speak → 📝 Text appears at cursor → Done!

🤔 The Problem

Every developer has been stuck in this painful loop:

┌─────────────────────────────────────────────────────────────┐
│  1. 🌐 Open browser/ChatGPT                                 │
│  2. 🎤 Click voice button                                   │
│  3. 🗣️ Speak                                                │
│  4. ⏳ Wait for transcription                               │
│  5. 📋 Copy text                                            │
│  6. 🔄 Switch back to terminal                              │
│  7. 📥 Paste                                                │
└─────────────────────────────────────────────────────────────┘
          ⬇️  That's 7 STEPS just to avoid typing!  ⬇️

✅ The Solution

┌─────────────────────────────────────────────────────────────┐
│                                                             │
│   Press Alt+R  →  🗣️ Speak  →  Press Alt+S  →  ✅ Done!   │
│                                                             │
└─────────────────────────────────────────────────────────────┘
              ⬆️  Just 2 STEPS with SpeakSkipType!  ⬆️

🚀 Quick Start

One-Line Install & Run

git clone https://github.com/DandaAkhilReddy/Speakskiptype.git && cd Speakskiptype && pip install -r requirements.txt && python speakskiptype.py

Note: On first run, wait 2-15 minutes for the speech model to download (~40MB). Time depends on your internet speed. This only happens once!

That's it! Now:

  1. Press Alt+R to start recording
  2. Speak your text
  3. Press Alt+S to stop & paste

⌨️ Keyboard Controls

Shortcut Action Description
Alt + R 🔴 Start Record Start recording (no terminal conflict!)
Alt + S ✅ Stop & Paste Stop recording and paste text
Alt + Q 🚪 Quit Exit application
Ctrl + D 🐛 Debug Mode Toggle debug output
Ctrl + H 📜 History Show transcription history

✨ Features

🏆 Superior to ALL Competitors

Feature SpeakSkipType Handy OpenWhispr voice_typing
100% Free ✅ ✅ ✅ ✅
100% Offline ✅ ❌ ✅ ✅
Multi-Engine (Vosk + Whisper) ✅ ❌ ❌ ❌
Filler Word Removal ✅ ❌ ❌ ❌
Continuous Listening ✅ ❌ ❌ ❌
Voice Commands ✅ ✅ ❌ ❌
Custom Vocabulary ✅ ❌ ❌ ❌
Code Dictation Mode ✅ ❌ ❌ ❌
Hold-to-Record ✅ ❌ ❌ ❌
Auto-Punctuation ✅ ❌ ❌ ❌
Export History ✅ ❌ ❌ ❌

🎯 All Features at a Glance

┌────────────────────────────────────────────────────────────────────┐
│                        🎤 SPEAKSKIPTYPE                            │
├────────────────────────────────────────────────────────────────────┤
│                                                                    │
│  🔧 ENGINES              │  🎛️ MODES                               │
│  ├─ Vosk (fast/light)    │  ├─ Normal recording                   │
│  └─ Whisper (accurate)   │  ├─ Hold-to-record                     │
│                          │  ├─ Continuous listening               │
│  🧹 TEXT PROCESSING      │  └─ Background/tray mode               │
│  ├─ Filler word removal  │                                        │
│  ├─ Auto-punctuation     │  🗣️ VOICE COMMANDS                     │
│  ├─ Custom vocabulary    │  ├─ "new line" → ↵                     │
│  └─ Code dictation mode  │  ├─ "period" → .                       │
│                          │  ├─ "delete that" → undo               │
│  📊 EXTRAS               │  └─ "question mark" → ?                │
│  ├─ Statistics (WPM)     │                                        │
│  ├─ History export       │  🌍 MULTI-PLATFORM                     │
│  └─ Real-time display    │  ├─ Windows ✅                          │
│                          │  ├─ macOS ✅                             │
│                          │  └─ Linux ✅                             │
└────────────────────────────────────────────────────────────────────┘

📦 Installation

Prerequisites

  • Python 3.7+
  • Microphone
  • ~40MB disk space

Step-by-Step

# 1. Clone the repo
git clone https://github.com/DandaAkhilReddy/Speakskiptype.git
cd Speakskiptype

# 2. Install dependencies
pip install -r requirements.txt

# 3. Run!
python speakskiptype.py

First Run - IMPORTANT!

┌─────────────────────────────────────────────────────────────────────────┐
│  ⚠️  FIRST RUN: Please wait for model download!                         │
│                                                                         │
│  The speech recognition model (~40MB) downloads automatically.          │
│  You'll see: "[*] Downloading speech model (first time only)..."       │
│                                                                         │
│  ⏱️  Download time: 2-15 minutes (depends on internet speed)            │
│      - Fast internet (50+ Mbps): ~2-3 minutes                           │
│      - Average internet (10-50 Mbps): ~5-8 minutes                      │
│      - Slow internet (<10 Mbps): ~10-15 minutes                         │
│                                                                         │
│  ✓ This only happens ONCE                                               │
│  ✓ After download, the app starts instantly                             │
│  ✓ Model is saved to ~/.speakskiptype/                                  │
└─────────────────────────────────────────────────────────────────────────┘

What to expect on first run:

  1. App starts and shows "Downloading speech model..."
  2. Wait 2-15 minutes (depending on your internet speed)
  3. You'll see "Model downloaded!" when complete
  4. The main interface appears - you're ready to go!

If it seems stuck:

  • The download IS happening in the background
  • No progress bar is shown (this is normal)
  • Just wait patiently - it WILL complete
  • On slow connections, it can take up to 15 minutes
  • Check ~/.speakskiptype/ folder to see if model is downloading

🧪 How to Test (Step-by-Step)

📋 Before You Start

Make sure you have:

  • ✅ Python 3.7+ installed (python --version to check)
  • ✅ A working microphone
  • ✅ Speakers/headphones (to hear beep feedback)

🪟 Windows Testing

Step 1: Open Command Prompt

Win + R → type "cmd" → Enter

Step 2: Navigate to the folder

cd C:\Users\YourUsername\Speakskiptype

Step 3: Install dependencies

pip install vosk sounddevice pynput pyperclip

Step 4: Run the app

python speakskiptype.py

Step 5: Test it!

┌────────────────────────────────────────────────────────────┐
│  1. Open Notepad (or any text editor)                      │
│  2. Click inside Notepad so cursor is there                │
│  3. Press Alt+R (you'll hear a beep - recording started)  │
│  4. Say: "Hello world this is a test"                      │
│  5. Press Alt+S (you'll hear a beep - recording stopped)  │
│  6. Watch the text appear in Notepad! ✨                   │
└────────────────────────────────────────────────────────────┘

🍎 macOS Testing

Step 1: Open Terminal

Cmd + Space → type "Terminal" → Enter

Step 2: Navigate and install

cd ~/Speakskiptype
pip3 install vosk sounddevice pynput pyperclip

Step 3: Grant permissions

  • Go to System Preferences → Security & Privacy → Privacy
  • Enable Microphone access for Terminal
  • Enable Accessibility access for Terminal (for hotkeys)

Step 4: Run and test

python3 speakskiptype.py

🐧 Linux Testing

Step 1: Open Terminal

Ctrl + Alt + T

Step 2: Install system dependencies

# Ubuntu/Debian
sudo apt install python3-pip portaudio19-dev

# Fedora
sudo dnf install python3-pip portaudio-devel

Step 3: Install Python packages

cd ~/Speakskiptype
pip3 install vosk sounddevice pynput pyperclip

Step 4: Run and test

python3 speakskiptype.py

✅ Test Checklist

Use this checklist to verify everything works:

# Test Expected Result Status
1 Run python speakskiptype.py Banner appears with controls ⬜
2 Press Alt+R See [REC] Recording... + beep ⬜
3 Speak "hello world" See Hearing: hello world... ⬜
4 Press Alt+S See [STOP] + [DONE] + beep ⬜
5 Check Notepad/editor Text "hello world" appeared ⬜
6 Press Alt+Q App exits cleanly ⬜

🎤 Test Filler Word Removal

Say this:

"Um I would like to you know test this application"

Expected output:

"I would like to test this application"

The filler words (um, you know) are automatically removed! ✨

🔊 Test Voice Commands

Say This Expected Output
"hello period" hello.
"new line test" hello
test
"open paren test close paren" (test)

🐛 Common Issues & Fixes

❌ "No module named 'vosk'" 
pip install vosk
❌ "No speech detected" every time
  1. Check microphone is not muted
  2. Check microphone permissions in system settings
  3. Try speaking louder/closer to mic
  4. Run: python -c "import sounddevice; print(sounddevice.query_devices())" to see devices
❌ Text not appearing in editor
  1. Make sure the editor window is focused/active
  2. Try clicking in the editor right before pressing Alt+S
  3. Check if clipboard is working: python -c "import pyperclip; pyperclip.copy('test'); print(pyperclip.paste())"
❌ Hotkeys not working (Windows)

Run Command Prompt as Administrator:

  1. Search "cmd" in Start Menu
  2. Right-click → "Run as administrator"
  3. Try again

🎮 Usage Examples

Basic Usage

python speakskiptype.py

With Whisper (More Accurate)

python speakskiptype.py --whisper

Continuous Listening (Like Speechnotes)

python speakskiptype.py --continuous

Code Dictation Mode

python speakskiptype.py --code

Background Mode (System Tray)

python speakskiptype.py --bg

All Options Combined

python speakskiptype.py --whisper --continuous --no-filler --code

🤖 Claude Code Integration

NEW! Use voice input directly with Claude Code!

Setup

# Terminal 1: Start voice input helper
python claude_voice.py

# Terminal 2: Open Claude Code
claude

Usage

Key Action
F8 Toggle recording (start/stop + auto-paste)
ESC Exit voice helper

Workflow

┌─────────────────────────────────────────────────────────────┐
│  1. Run claude_voice.py (keep open)                         │
│  2. Run claude in another terminal                          │
│  3. Press F8 → Speak → Press F8                             │
│  4. Your speech appears in Claude Code! 🎉                  │
└─────────────────────────────────────────────────────────────┘

🧠 Anthropic API with Extended Thinking

NEW! Direct integration with Claude's API including proper support for Extended Thinking.

The Problem

When using Claude's extended thinking feature via the Anthropic API, you might encounter this error:

Error: thinking blocks must be preserved exactly as received

This happens when:

  • Thinking blocks are modified or removed from conversation history
  • The order of content blocks is changed
  • Follow-up messages don't include the original thinking blocks

The Solution

We provide anthropic_client.py - a client that automatically handles extended thinking correctly.

Installation

pip install anthropic

Quick Start

from anthropic_client import ClaudeClient

# Create client (uses ANTHROPIC_API_KEY env var)
client = ClaudeClient()

# First message - thinking blocks are generated internally
response = client.send_message("Explain quantum entanglement")
print(response)

# Follow-up - thinking blocks are automatically preserved!
response = client.send_message("Can you give a simpler analogy?")
print(response)

# Start fresh conversation (clears all history including thinking)
client.clear_conversation()

Configuration Options

from anthropic_client import ClaudeClient

client = ClaudeClient(
    api_key="your-api-key",           # Or set ANTHROPIC_API_KEY env var
    model="claude-sonnet-4-20250514",    # Model to use
    max_tokens=16000,                  # Max response tokens
    thinking_enabled=True,             # Enable extended thinking
    thinking_budget=10000              # Token budget for thinking
)

# Disable thinking for simpler queries
client.disable_thinking()

# Re-enable with custom budget
client.enable_thinking(budget_tokens=5000)

How It Works

The key fix is preserving response.content exactly as returned:

# CORRECT - Store complete response including thinking blocks
self._messages.append({
    "role": "assistant",
    "content": response.content  # Includes ALL blocks unchanged
})

# WRONG - Don't filter or modify!
# content = [b for b in response.content if b.type == "text"]  # BAD!

Extended Thinking Rules

Rule Description
Never modify Keep thinking/redacted_thinking blocks exactly as received
Never remove Include all blocks in conversation history
Keep order Don't reorder content blocks
First position Thinking block must be first in assistant's content array

API Reference

Method Description
send_message(text) Send message, returns text response
clear_conversation() Start fresh conversation
get_conversation_history() Get full history with thinking blocks
disable_thinking() Turn off extended thinking
enable_thinking(budget) Turn on with token budget

Example: Multi-turn Conversation

from anthropic_client import ClaudeClient

client = ClaudeClient(thinking_budget=8000)

# Turn 1 - Claude thinks through the problem
r1 = client.send_message("What's 15 * 23?")
print(f"Answer: {r1}")

# Turn 2 - Previous thinking is preserved automatically
r2 = client.send_message("Now multiply that by 2")
print(f"Answer: {r2}")

# Turn 3 - Full context maintained
r3 = client.send_message("What was my first question?")
print(f"Answer: {r3}")

🗣️ Voice Commands

Speak these commands while recording:

Say This Get This
"new line" ↵ (line break)
"new paragraph" ↵↵ (double line break)
"period" .
"comma" ,
"question mark" ?
"exclamation mark" !
"open paren" (
"close paren" )
"delete that" (removes last phrase)

🧹 Filler Word Removal

SpeakSkipType automatically removes filler words:

You Say You Get
"I um want to uh test this" "I want to test this"
"So like you know it works" "So it works"
"I mean basically it's done" "it's done"

Filler words removed: um, uh, er, ah, like, you know, i mean, sort of, kind of, basically, actually, literally, so yeah, right, okay so, well

⚙️ Configuration

Command Line Flags

Flag Description
--whisper Use Whisper engine (more accurate)
--continuous Continuous listening mode
--code Code dictation mode
--vad Voice Activity Detection
--no-filler Remove filler words
--literal Literal punctuation mode
--bg Run in background
--stats Show statistics
--export Export history

Custom Vocabulary

Edit the config to add custom word replacements:

"custom_vocabulary": {
    "kubernetes": "K8s",
    "javascript": "JavaScript",
    "python": "Python"
}

📊 Performance

Metric Vosk Whisper
Model Size ~40MB ~150MB
RAM Usage ~200MB ~500MB
Speed ⚡ Fast 🐢 Slower
Accuracy 90% 98%
Offline ✅ Yes ✅ Yes

🔧 Troubleshooting

🔇 "No speech detected"
  • Check microphone is connected and not muted
  • Speak clearly at normal pace
  • Check system microphone permissions
🎤 "Audio Error"
  • Close other apps using microphone
  • Try different audio input device
  • Restart the application
🐧 Linux: Permission issues 
sudo usermod -a -G audio $USER
# Log out and log back in
🪟 Windows: Hotkeys not working
  • Run terminal as Administrator
  • Try different terminal (PowerShell, CMD)

🧪 Testing

# Run all tests
pytest tests/ -v

# Run with coverage
pytest tests/ --cov=speakskiptype

# 306 tests passing ✅

🤝 Contributing

Pull requests welcome! For major changes, open an issue first.

# Fork & clone
git clone https://github.com/YOUR_USERNAME/Speakskiptype.git

# Create feature branch
git checkout -b feature/amazing-feature

# Make changes & test
pytest tests/ -v

# Commit & push
git commit -m "Add amazing feature"
git push origin feature/amazing-feature

# Open Pull Request

📄 License

MIT License - Use it, modify it, share it freely.

⭐ Star This Repo!

If SpeakSkipType saves you time, please give it a ⭐!

It helps others discover this tool.

Built with ❤️ by developers, for developers.

Stop copy-pasting. Start speaking.

Star

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages
⚡