🎙️ Magic Voice Chat

Magic Voice Chat lets you talk with AI characters in real time using your voice — like chatting with Einstein, a space wizard, or the OS from Her. It's fast, expressive, hilarious, and fully customizable. Run locally, use your favorite models (OpenAI, Anthropic, xAI, Ollama), and switch between multiple voices (OpenAI TTS, ElevenLabs, XTTS, or Kokoro TTS) — all from a slick Web UI.

---

🚀 Features

• Real-Time Conversations with OpenAI’s WebRTC Realtime API — interrupt mid-sentence!
• Expressive TTS: GPT-4o-mini-tts, ElevenLabs, XTTS, or Kokoro TTS with emotional voice prompts.
• Flexible Transcription: OpenAI or local Faster Whisper.
• Web UI + CLI: Talk, tweak, and test with your voice or keyboard.
• Huge Cast of Characters: 50+ AIs with wild personalities — therapists, pirates, vampires, and more.
• Game & Story Modes: Escape rooms, trivia games, noir mysteries, space adventures, and more.
• Docker & Local Support: Run it your way — Docker or native Python 3.10 on Windows, Linux, or macOS.
• Sentiment-Based Replies: AI adjusts its tone based on your mood.
• Fully Offline Capable: Use local models like Ollama + XTTS with zero internet.

---

🧪 Quick Start

🔧 Requirements

• Python 3.10
• ffmpeg
• A microphone
• (Optional) CUDA-enabled GPU for faster XTTS/Faster Whisper
• Docker (if running via container)

🖥️ Local Installation

git clone https://github.com/bigsk1/voice-chat-ai.git
cd voice-chat-ai
python -m venv venv
source venv/bin/activate      # Windows: venv\Scripts\activate
pip install -r requirements_cpu.txt  # or requirements.txt for GPU

Then run the app:

uvicorn app.main:app --host 0.0.0.0 --port 8000

Visit: http://localhost:8000

🐳 Docker Run (CPU)

docker pull bigsk1/voice-chat-ai:latest
docker run -d --env-file .env -p 8000:8000 --name magic-voice-chat bigsk1/voice-chat-ai:latest

Need CUDA? Use Dockerfile.cuda or native install.

---

🗣️ Voice, Model, and Transcription Options

Feature	Providers
Language Model	OpenAI, Anthropic, xAI, Ollama
TTS Voices	OpenAI, ElevenLabs, XTTS, Kokoro
Transcription	OpenAI Whisper, Faster Whisper
UI Control	Web browser or terminal

---

🎮 Game & Story Modes

• Games: Hangman, Escape Master, Trivia, Word Weavers, Logic Puzzles
• Stories: Noir Detective, Oregon Trail, Space Adventure, Haunted Mansion

All driven by in-character AI guides!

---

🧙 Create Your Own Characters

1. Create a new folder: characters/wizard/

2. Add:

   • wizard.txt (character prompt + voice instructions)
   • prompts.json (mood responses)
   • Optional: wizard.wav (custom XTTS voice)

3. Done! Select in the UI and start talking.

---

🔧 Configuration

Copy .env.sample → .env and adjust settings:

MODEL_PROVIDER=openai
CHARACTER_NAME=einstein
TTS_PROVIDER=elevenlabs
OPENAI_MODEL=gpt-4o
OPENAI_API_KEY=your_api_key
ELEVENLABS_API_KEY=your_api_key
...

---

🧠 Smart Features

• "What's on my screen?" — screen analysis with llava (Ollama/OpenAI)
• "Quit" / "Exit" — ends conversation
• Dynamic mood detection — TextBlob sentiment scores adjust AI tone
• ElevenLabs voice sync — load your voice list via:

curl -s -X GET https://api.elevenlabs.io/v1/voices \
  -H "xi-api-key: $ELEVENLABS_API_KEY" | jq ...

---

📺 Demos

Mode	Demo Video
OpenAI Realtime	whisperer.mp4
XTTS (local, GPU)	magic_xtts_gpu.mp4
OpenAI Enhanced TTS	expressive_voice.mp4
Escape Game Demo	ninja_assassin.mp4

---

💡 Tips

• Best combo: xAI + ElevenLabs + Faster Whisper (GPU)
• Want fast + local? Use Ollama + XTTS (slower on CPU)
• UI is recommended for switching voices/models on the fly

---

📜 License

MIT License

---

⭐ Like the project? Show some love with a star: Magic Voice Chat on GitHub