jason-ai-twin

Jason AI Twin (Offline Digital Twin & Cloned Voice)

An offline, privacy-first AI Digital Twin architecture featuring quantized local LLM reasoning, custom knowledge bases, GPU-accelerated speech recognition, and neural voice cloning.

🎙️ Cloned Voice & Audio Assets

The AI Twin features neural voice cloning sourced from /local-tts:

• cloned_output.wav: High-fidelity cloned voice reference sample synthesized from Jason's voice profile. Used as the primary speaker reference for zero-shot Coqui XTTS v2 voice synthesis.
• cloned_sample2.wav: Second verified neural synthesis sample generated with Coqui XTTS v2 on CUDA using Jason's vocal profile ("The voice cloning matrix is fully operational...").
• my_voice_clean.wav: Cleaned, noise-reduced studio vocal sample of Jason's voice (22.05 kHz PCM) for fine-tuning and voice cloning alignment.
• amy.onnx / amy.onnx.json: Ultra-fast, low-latency ONNX voice model for Piper TTS fallback.

---

🧠 Brain & Models

• digital_twin_q4.gguf: Q4_K_M quantized LLM brain (jason_twin) trained and adapted on Jason's communication style, knowledge bases, and preferences.
• Modelfile: Ollama configuration for launching jason_twin locally with optimized system prompt, temperature, and context length.

---

🚀 Interactive CLI Client (chat_twin.py)

The repository includes chat_twin.py, an interactive voice and text terminal client:

• Two-Way Voice Interaction: Press [ENTER] on an empty prompt to record your voice; local Whisper (openai/whisper-tiny.en) transcribes it with GPU acceleration.
• Neural Voice Synthesis: Speaks responses back in Jason's cloned voice using local XTTS v2 (--speaker_wav cloned_output.wav).
• Multi-Engine Fallback: Gracefully falls back to Piper ONNX, spd-say, or espeak-ng if GPU voice cloning is unavailable.

In-Chat Commands:

---

🛠️ Quickstart

1. Run via Ollama

ollama create jason_twin -f Modelfile
ollama run jason_twin

2. Run Interactive Voice Twin CLI

# Ensure dependencies are installed
pip install TTS transformers torch torchaudio

# Launch voice chat client
python3 chat_twin.py

3. Desktop Shortcut

On Ubuntu systems, launch directly via the included AI_Twin.desktop or:

./launch_twin.sh