Question 1

How do I use Voconly?

Accepted Answer

It's simple - just three steps: Basic Flow: Single press shortcut - Start recording (default Right Alt key) Speak - Say what you want to input Single press again - Stop recording, if AI service is configured, text goes through AI processing first Double-tap to End (Direct Output): Double-tap the shortcut to end recording and output raw recognition results Skips AI processing even if configured Perfect for quick input when you don't need polishing Single-tap to End (AI Processing): Single-tap to end - if AI service is configured, text goes through AI processing first Auto-polish, add punctuation, and format before output Great for turning speech into polished text The app runs in the system tray and works across all your apps.

Question 2

Is Voconly free?

Accepted Answer

Yes. Voconly is completely free and open source (MIT License). All core features are permanently free: Offline speech recognition (local models) Global shortcuts Multi-language support AI services (requires your own API) No paid version, no feature limits, no usage limits.

Question 3

Is Voconly open source?

Accepted Answer

Yes. Source code is hosted on GitHub: Repository: https://github.com/xinkyle/Voconly License: MIT License Feel free to submit Issues, Pull Requests, or help with translations.

Question 4

Will my voice data be uploaded to the cloud?

Accepted Answer

No. Voconly currently only supports local speech model transcription, not cloud models. Offline Speech Recognition: Runs entirely locally, no network needed Local Models: Audio is processed locally, no data is uploaded AI Services (optional): If you configure an online AI API, text will be sent to your configured provider If you only need offline voice input, you can use Ollama for local AI deployment. All data stays local.

Question 5

Which platforms does Voconly support?

Accepted Answer

Platform Support Notes --------- ------- Windows ✅ Full Support Windows 10/11 macOS ✅ Full Support macOS 11+ (Apple Silicon)

Question 6

How to install Voconly?

Accepted Answer

Windows: Download the .exe installer from GitHub Releases or Gitee Releases Run the installer On first launch, you'll be prompted to download the speech recognition model macOS: Download the .dmg file from GitHub Releases or Gitee Releases Open the DMG file and drag Voconly to Applications On first launch, you may see "cannot verify developer" or "file is damaged" — this is macOS security protection See macOS Installation Guide for the solution You'll be prompted to download the speech recognition model on first launch.

Question 7

What should I do on first launch?

Accepted Answer

Download Model - Required for first-time use Grant Permissions - macOS needs microphone and accessibility permissions Set Shortcuts - You can customize shortcuts

Question 8

Which model should I use?

Accepted Answer

Choose based on your use case: Chinese Scenarios: Need Recommended Model Notes ------------------- ------- Pure Chinese SenseVoice Small Fast, low resource usage Chinese-English Mixed Qwen3-ASR 1.7B Best mixed accuracy Chinese Dialects Qwen3-ASR 1.7B Supports 22 Chinese dialects English Scenarios: Need Recommended Model Notes ------------------- ------- English-focused Parakeet Unified EN 0.6B English-optimized, streaming support General English Cohere Transcribe HuggingFace Open ASR 1 Other Scenarios: Need Recommended Model Notes ------------------- ------- Low latency/Real-time Nemotron 3.5 ASR Streaming Streaming, 80ms ultra-low latency Multi-language Whisper Large v3 Turbo 99 languages, built-in translation Low-spec PC SenseVoice Small / Nemotron 0.6B Low resource, pure CPU runnable Hardware Recommendations: With NVIDIA GPU: Use Qwen3-ASR or Cohere with GPU acceleration for faster performance No GPU: Use SenseVoice Small or Nemotron for smooth CPU-only operation

Question 9

What's the difference between local and cloud models?

Accepted Answer

Voconly currently only supports local models. All speech recognition runs entirely locally. Local Model Advantages: No internet needed, works offline No data uploads, privacy-safe No usage limits No extra costs Available Models: Qwen3-ASR, SenseVoice, Cohere, Parakeet, Nemotron, Whisper, etc. Once downloaded, models work offline

Question 10

Model download is slow or failing?

Accepted Answer

Check your network connection Check disk space (models are 0.5GB-2GB) If interrupted, retry the download Manual Import: Download from ModelScope or HuggingFace Click "Import External Model" in the model list Select the model folder to import

Question 11

What are the shortcuts? Can I change them?

Accepted Answer

Default Shortcuts: Platform Start Recording Stop Recording ----------------- ---------------- Windows Right Alt Right Ctrl macOS Right Alt Right Ctrl You can customize shortcuts in settings.

Question 12

No response after pressing the shortcut?

Accepted Answer

Check in order: Check Microphone Permissions Windows: Settings > Privacy > Microphone Check if Model is Downloaded Open settings, confirm model status shows "Downloaded" Restart the App Try restarting Check Shortcut Conflicts May conflict with other apps, try changing the shortcut

Question 13

Why does "Thank you" appear when I didn't say it?

Accepted Answer

This is a known issue with Whisper models: when detecting silence, it sometimes "hallucinates" "Thank you". Solution: Start speaking immediately after pressing the shortcut, don't pause too long

Question 14

Why is the first or last word cut off?

Accepted Answer

This is a timing issue with audio capture. Solution: After pressing shortcut, pause half a second before speaking After finishing, pause half a second before releasing Give audio capture time to start and stop

Question 15

Why is recognition inaccurate when speaking fast?

Accepted Answer

Speaking too fast reduces accuracy. Solution: Slow down slightly Speak clearly, reduce connected speech Try a more accurate model (like Qwen3-ASR)

Question 16

Technical terminology not recognized accurately?

Accepted Answer

Solution: Speak clearly, slow down slightly Future versions will support custom dictionaries

Question 17

What are AI services?

Accepted Answer

AI services process text after speech recognition, including: Light Polish - Fix typos, add punctuation, format Professional Polish - Make text smoother and more formal Translation - Translate content to other languages You can configure different processing modes for different scenarios and switch with one click.

Question 18

Which AI services are supported?

Accepted Answer

Local Models: Ollama (currently supported local solution) Online Models: OpenAI (GPT) Anthropic (Claude) Other OpenAI API-compatible providers Recommend using Ollama for local deployment - privacy-safe with no extra cost.

Question 19

How to skip AI processing?

Accepted Answer

Once AI service is configured, it processes speech recognition results by default. To Skip: Double-tap the shortcut when ending recording to output raw recognition results Single-tap to end goes through AI processing This lets you quickly get raw text when needed.

Question 20

Where are transcription records stored?

Accepted Answer

Entirely locally: Platform Storage Path -------------- Windows %LOCALAPPDATA%\Local\Voconly\User Data\ macOS ~/Library/Application Support/Voconly/User Data/

Question 21

Can I export data?

Accepted Answer

Data export is not currently supported.

Question 22

How to completely clear data?

Accepted Answer

History is stored in history.db.

Question 23

How to check for updates?

Accepted Answer

Windows: Settings > About > Check for Updates App automatically checks for updates on startup

Question 24

How to report issues?

Accepted Answer

GitHub Issues: https://github.com/xinkyle/Voconly/issues Describe the problem and steps to reproduce Include system info and error logs (if available)

Question 25

What scenarios is Voconly suitable for?

Accepted Answer

Meeting Notes - Quickly record meeting points Document Writing - Emails, reports, articles, etc. Instant Messaging - Quick replies in WeChat, Slack, etc. Programming - Code comments, prompt writing Accessibility - Help users with hand mobility issues Content Creation - Video/podcast subtitles, blogs

Question 26

Can it be used for programming?

Accepted Answer

Yes.

FAQ

General

Installation & Configuration

Speech Recognition Models

Usage Issues

AI Services

Data & Privacy

Updates & Support

Use Cases

Save 30 minutes of typing every day