About This Software

MobMate WhispTalk is a zero-cloud voice assistance toolkit designed for gaming voice chat.

It's not for those who "can't speak" — it's for those who choose not to speak at this moment, respecting the atmosphere and distance of the situation.

Stay connected to voice chat without using your real voice. Only when needed, briefly, naturally.

Unlike cloud-based services, all voice processing is completed on your PC (Zero-Cloud). Your voice and text are never sent externally, prioritizing privacy.

---

## 🧠 Key Features

### Real-time Speech Recognition & Anonymous Voice Output

Transcribe your speech locally and play it back as an anonymous voice. Leave your reactions and responses in the conversation without using your real voice.

**Use Cases:**

* Late-night gaming: Participate in VC without waking family

* While streaming: Communicate without exposing your real voice to viewers

* Poor health: Continue voice chat when your throat hurts

* Privacy-focused: Participate while protecting your voiceprint and personal information

---

### 📻 Radio Chat

Send preset phrases and frequently used expressions instantly via global hotkey or overlay.

**Features:**

* **Global Hotkey Support:** Speak without Alt+Tab during gameplay

* **Overlay Display:** Overlay on game screen, send with mouse click

* **Easy Registration from History:** Right-click past phrases → Add to Radio Chat

* **Customizable:** Register frequently used phrases freely

**Use Cases:**

* Tactical Calls: Send "Enemy incoming," "Need heal," "Regroup" instantly

* Standard Greetings: One-touch "GG," "Thanks," "Roger"

* During Gameplay: Quick communication without typing

* Foreign Language VC: Send pre-prepared English phrases instantly

**How to Use:**

1. Check past phrases in History window

2. Right-click frequently used phrases → "Add to Radio"

3. Press hotkey or use overlay during game

4. Instantly TTS reads aloud → Send to VC

---

### 🎧 Hearing Feature (Experimental - WIP)

Real-time transcription of desktop audio (system output) from Discord, VRChat, in-game VC, etc. Display other players' speech on screen.

**Features:**

* Overlay Display: Overlay on game screen

* Multi-language Support: Auto-detect 100+ languages

* Customizable: Font size, background color, opacity, position, history size, display selection

**Use Cases:**

* Communication in Foreign Language VC: Check speech content in real-time

* Prevent Mishearing: Review rapid instructions and tactical calls later

* As Subtitles: Understand content in hard-to-hear environments

* As Record: Keep important conversations in history

**Note:** This feature is experimental. Recognition accuracy and stability are scheduled for future improvement. Features like auto-translation are currently under development.

---

### Speaker Verification (Recognize Only Your Voice)

Using MFCC (Mel-Frequency Cepstral Coefficients) speaker verification, recognize only your voice and ignore others' voices or environmental sounds. Prevent false recognition and send only intended speech.

**Use Cases:**

* Ignore Cohabitant Voices: Don't pick up family conversations

* Ignore Game Audio: Don't pick up in-game NPCs or BGM

* Ignore Stream Audio: Don't pick up streaming software audio

* Improved Accuracy: Reliably recognize only your own speech

**Initial Setup:**

Simply speak a few times at first launch to register your voiceprint. After that, it automatically recognizes only your voice.

---

### Designed for Gaming VC Tempo

Not for long sentences, but for short reactions and callouts. Use without breaking the "flow" of multiplayer — laughter, acknowledgments, tactical phrases.

**Preset Features:**

* Laughter: Auto-convert "lol," "haha" to laughter

* Acknowledgments: Natural acknowledgments like "yeah," "right"

* Tactical Phrases: Register in Radio Chat for instant sending

---

### Lightweight & Offline Operation

Local recognition via Whisper.cpp operates with low latency without relying on the internet. (*Experience varies by environment)

**CPU Version:**

* Included standard, no additional cost

* Works on Intel/AMD CPUs

* Low latency, stable operation

**GPU Version (DLC):**

* Accelerated with Vulkan-compatible GPU

* Even lower latency, comfortable operation

* AMD, NVIDIA, Intel Arc compatible

---

### 3 TTS Engine Support

Supporting 3 TTS engines for clearer, more natural voice output.

**VoiceVox (Japanese):**

* Free, high-quality Japanese TTS

* Select from multiple character voices

* Emotional expression available

**Voiceger (5 Languages):**

* **Supported Languages:** Japanese, English, Chinese (Simplified), Korean, Cantonese

* Next-generation TTS engine

* More natural pronunciation

* Low latency, high quality

* Ideal for global multiplayer VC

**XTTS (Multi-language, Experimental):**

* Supports 100+ languages

* Custom voice reactions

* External API integration (requires separate setup)

* *XTTS not included. See GitHub for details

---

### Customization (Ignore List / Dictionary Filter / Good List)

Remove filler words and adjust phrasing to your preference. Adjustment features let you continue using it "your way."

**Ignore List:**

* Remove filler words like "uh," "um"

* Automatically exclude unwanted recognition results

**Dictionary Filter:**

* Shorten conversions like "roger" → "rog"

* Auto-correct game-specific terminology

**Good List:**

* Register frequently used phrases

* Send quickly via Radio Chat integration

---

## 🎮 How It Works

### Normal Speech Recognition:

1. Speak into microphone

2. WhispTalk recognizes voice locally (speaker verification for your voice only)

3. Play back recognition result as anonymous voice

4. Connect to Discord, VRChat, various multiplayer games via virtual audio

### Radio Chat:

1. Register preset phrases from History or settings

2. Press hotkey during game (e.g., Ctrl+1)

3. Instantly TTS reads aloud → Send to VC

---

## 🔒 Privacy-Focused (Zero-Cloud)

This software does not send voice or text data externally. All processing is completed on your PC. No dependence on cloud services, complete privacy protection.

**What is Zero-Cloud:**

* Never send to cloud

* Voice data stays on PC

* 100% privacy protection

* Internet connection unnecessary (offline operation possible)

---

## ⚡ GPU Acceleration Pack (DLC)

The base works on CPU, but introducing DLC offloads heavy processing to Vulkan-compatible GPU. Recommended for those who want low latency and more comfortable use.

**Compatible GPUs:**

* AMD Radeon (Vulkan-compatible)

* NVIDIA GeForce (Vulkan-compatible)

* Intel Arc (Vulkan-compatible)

**Effects:**

* Recognition speed up to 3x faster

* Lower latency for improved real-time performance

* Reduced CPU load

---

## ⚠️ Notes & Limitations

* Recognition accuracy varies with ambient noise and speech patterns

* This tool is for short reactions and simple callouts, not for long transcriptions or meeting minutes

* Speaker verification accuracy may decrease with high environmental noise

* Hearing feature is experimental and recognition accuracy/stability scheduled for future improvement

---

## 💡 Who MobMate WhispTalk Is For

In multiplayer VC, it's not just "can I speak" but also "should I speak now."

* Want to participate without speaking now

* Hesitant to join VC but want to maintain presence

* Want to play without sharing too much personal information

* Want to use voice chat while prioritizing privacy

* Situations where speaking is difficult: late-night gaming, living with family, streaming, poor health

* Want to communicate in foreign language VC (Voiceger 5-language support)

* Want to send tactical calls and preset phrases quickly (Radio Chat)

MobMate WhispTalk is a tool for naturally continuing to engage in voice chat without using your real voice.

**Quiet players can be loud.** Even quiet participation deserves a clear voice.

---

## 🌐 Compatible Platforms

Compatible with Discord, VRChat, and other multiplayer games in general. Combined with virtual audio devices (VB-Audio VoiceMeeter, etc.), usable with any voice chat app.

**Confirmed Working:**

* Discord

* VRChat

* Other games supporting virtual audio devices

---

## 🌍 Multi-language UI Support

UI supports Japanese, English, Chinese (Simplified/Traditional), Korean.

---

## 📚 Support / Technical Information

Source code and technical information published on GitHub:

**Discord Community:**

This product includes whisper.cpp (MIT License)

AI Generated Content Disclosure

The developers describe how their game uses AI Generated Content like this:

This application uses AI technologies for real-time speech recognition, optional translation, and voice synthesis during gameplay.

User voice input is processed locally or via optional AI models to generate text and synthesized speech. No personal data is stored or shared without user consent.

AI is also used for some pre-generated assets such as icons or promotional images.