Unicorn Execution Engine

Multi-Platform Hardware-Accelerated AI Execution Framework

🚀 Overview

The Unicorn Execution Engine is a high-performance runtime for deploying AI models on specialized hardware accelerators including Intel iGPUs, AMD NPUs, and more. Developed by Magic Unicorn Inc., this engine achieves unprecedented performance through hardware-specific optimizations.

✨ Key Features

Intel iGPU (OpenVINO)

• 3-5x Speedup for Kokoro TTS vs CPU
• 15W Power Efficiency for laptops
• 50+ Professional Voices for TTS
• Zero-Copy Shared Memory architecture

AMD NPU (MLIR-AIE2)

• 220x Speedup for WhisperX vs CPU
• Custom MLIR-AIE2 Kernels for optimal utilization
• INT8/INT4 Quantization with minimal accuracy loss
• 16 TOPS INT8 performance on Phoenix NPU

📊 Performance Benchmarks

Kokoro TTS v0.19 (Intel iGPU)

Platform	Latency	Power	Speedup
Intel Iris Xe	150ms	15W	3.0x
Intel UHD	250ms	12W	1.8x
CPU (i7)	450ms	35W	Baseline

WhisperX Speech Recognition (AMD NPU)

Model	CPU Time	NPU Time	Speedup	Accuracy
Large-v3	59.4 min	16.2 sec	220x	99%
Large-v2	54.0 min	18.0 sec	180x	98%
Medium	27.0 min	14.4 sec	112x	95%

🛠️ Quick Start

Intel iGPU - Kokoro TTS

# Install with OpenVINO support
pip install unicorn-execution-engine[intel-igpu]

# Or use Docker
docker pull magicunicorn/unicorn-execution-engine:kokoro-intel-igpu

from tts.kokoro_intel_igpu import KokoroIntelTTS

# Initialize with Intel iGPU
tts = KokoroIntelTTS(device="igpu")

# Synthesize with 50+ voices
audio = tts.synthesize("Hello world!", voice="af_bella")

AMD NPU - WhisperX

# Install with NPU support
pip install unicorn-execution-engine[amd-npu]

from unicorn_engine import NPUWhisperX

# Load quantized model
model = NPUWhisperX.from_pretrained("magicunicorn/whisperx-large-v3-npu")

# Transcribe with 220x speedup
result = model.transcribe("meeting.wav")

🏗️ Architecture

Unicorn Execution Engine
├── TTS Module
│   ├── Kokoro v0.19 (Intel iGPU) ✅
│   ├── Whisper (AMD NPU) ✅
│   └── Bark (Planned)
├── LLM Module
│   ├── Llama (AMD NPU) 🚧
│   └── Mistral (NVIDIA) 📋
└── Vision Module
    ├── CLIP (Apple ANE) 📋
    └── SAM (Qualcomm) 📋

📦 Pre-built Models & Packages

HuggingFace Models

• kokoro-tts-intel - Kokoro TTS for Intel iGPU
• whisperx-large-v3-npu - WhisperX for AMD NPU

Docker Images

# Intel iGPU with Kokoro TTS
docker run --device /dev/dri -p 8880:8880 \
    magicunicorn/unicorn-execution-engine:kokoro-intel-igpu

# AMD NPU with WhisperX  
docker run --device /dev/accel -p 8881:8881 \
    magicunicorn/unicorn-execution-engine:whisperx-amd-npu

🔧 Platform-Specific Setup

Intel iGPU Setup

# Install OpenVINO and drivers
sudo apt-get install intel-opencl-icd intel-level-zero-gpu level-zero
pip install openvino==2024.0.0 onnxruntime-openvino==1.17.0

# Verify Intel GPU
lspci | grep -i intel | grep -i vga

AMD NPU Setup

# Install XRT and drivers
sudo apt-get install xrt amd-npu-driver
pip install pyxrt>=2.0.0

# Verify NPU
ls /dev/accel/accel0

💾 Model Files (Git LFS)

Kokoro TTS

• models/kokoro-v0_19.onnx (311MB) - TTS model
• models/voices-v1.0.bin (25MB) - 50+ voice embeddings

WhisperX

• models/whisperx-large-v3.npumodel (1.5GB) - Quantized INT8 model
• models/whisperx-kernels.xclbin (50MB) - Custom MLIR kernels

🚀 Advanced Features

Multi-Hardware Pipeline

from unicorn_engine import MultiPlatformEngine

engine = MultiPlatformEngine()

# Automatically selects best hardware
# Intel iGPU for TTS, AMD NPU for STT
pipeline = engine.create_pipeline([
    ("speech_recognition", "amd-npu"),
    ("text_synthesis", "intel-igpu")
])

result = await pipeline.process(audio_input)

Custom Optimization

# Intel iGPU optimization
from intel_igpu_module import IntelIGPUExecutor
executor = IntelIGPUExecutor()
executor.optimize_for_latency()

# AMD NPU quantization
from unicorn_engine import Quantizer
quantizer = Quantizer(target="npu", precision="int8")

📈 Roadmap

• Intel iGPU support (OpenVINO)
• AMD NPU support (MLIR-AIE2)
• NVIDIA GPU support (TensorRT)
• Apple Neural Engine support
• Qualcomm Hexagon DSP support
• Multi-device distribution

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Development

# Clone repo
git clone https://github.com/Unicorn-Commander/Unicorn-Execution-Engine
cd Unicorn-Execution-Engine

# Install dev dependencies
pip install -e .[dev]

# Run tests
pytest tests/

📚 Documentation

• Intel iGPU Guide - Kokoro TTS optimization
• AMD NPU Guide - WhisperX acceleration
• API Reference - Complete API documentation
• Benchmarks - Performance analysis

🏢 About Magic Unicorn Inc.

Magic Unicorn Inc. develops enterprise AI solutions optimized for edge deployment. The Unicorn Commander Suite provides complete AI infrastructure for on-premise deployments.

Related Projects

• Unicorn-Orator - Full TTS platform
• Meeting-Ops - AI meeting recorder
• Unicorn Models - Pre-optimized models

📄 License

MIT License - see LICENSE file for details.

🙏 Acknowledgments

• Intel for OpenVINO and iGPU support
• AMD for NPU hardware and MLIR-AIE2
• OpenAI for original Whisper models
• The open-source community

📞 Contact

• GitHub Issues: Report bugs
• HuggingFace: Discussion forum
• Email: [email protected]

---

© 2025 Magic Unicorn Inc. | Part of the Unicorn Commander Suite

⭐ Star us on GitHub if you find this useful!