🦄 Upload NPU+iGPU kokoro-npu-quantized model

README.md

@@ -2,210 +2,47 @@
 language:
 - en
 license: apache-2.0
-library_name: onnxruntime
+library_name: onnx
 tags:
-- text-to-speech
+- tts
 - kokoro
 - npu
-- amd-ryzen-ai  
+- amd-ryzen-ai
 - quantized
-- onnx
-- magic-unicorn-tts
 pipeline_tag: text-to-speech
 ---
 
-# 🦄⚡ Kokoro TTS NPU-Optimized
+# 🎵 Kokoro NPU Quantized TTS Models
 
-**NPU-accelerated Kokoro TTS models specifically optimized for AMD Ryzen 9 8945HS NPU Phoenix (AIE-ML)**
+🎵 NPU-Optimized Text-to-Speech Models
 
-## Model Description
+NPU-optimized text-to-speech models for AMD Ryzen AI hardware.
 
-These models are NPU-optimized versions of Kokoro TTS, specifically quantized and optimized for AMD Ryzen AI NPU hardware. Developed by [Magic Unicorn Technologies](https://magicunicorn.tech) and [Unicorn Commander](https://unicorncommander.com).
+## Models Included
 
-### Key Features
-- 🚀 **30% Performance Improvement** on AMD NPU Phoenix in turbo mode (RTF 0.213)
-- ⚡ **Multiple Precision Options**: INT8, FP16, and full precision
-- 🎭 **54 Voice Support**: Complete voice library included
-- 🛠️ **Ready-to-Use**: Compatible with Magic Unicorn TTS interface
-
-## Model Variants
-
-| Model | Precision | Size | NPU Performance | Use Case |
-|-------|-----------|------|----------------|----------|
-| `kokoro-npu-quantized-int8.onnx` | INT8 | 128 MB | RTF 0.213 | Maximum speed with turbo |
-| `kokoro-npu-fp16.onnx` | FP16 | 178 MB | RTF 0.225 | Balanced quality/speed |
-
-*RTF = Real-Time Factor (lower is faster)*
+- **kokoro-npu-quantized-int8.onnx** (121.9 MB) - INT8 quantized for NPU
+- **kokoro-npu-fp16.onnx** (169.8 MB) - FP16 for quality
+- **voices-v1.0.bin** (26.9 MB) - Voice embeddings
 
 ## Hardware Requirements
 
-- **NPU**: AMD Ryzen 9 8945HS with NPU Phoenix (AIE-ML)
-- **iGPU**: AMD Radeon Graphics (RADV PHOENIX) gfx1103 (UI acceleration)
-- **RAM**: 96GB (16GB allocated to VRAM, heterogeneous memory architecture)
-- **OS**: Ubuntu 25.04 with KDE Plasma, Linux kernel 6.14.0+
+- **NPU**: NPU Phoenix
+- **Memory**: 1GB+ available
+- **Framework**: Unicorn Execution Engine
 
 ## Quick Start
 
-### Using Magic Unicorn TTS (Recommended)
-```bash
-# One-click installation
-curl -fsSL https://raw.githubusercontent.com/Unicorn-Commander/magic-unicorn-tts/main/install.sh | bash
-
-# Launch interface
-cd magic-unicorn-tts
-./launch_enhanced.sh
-```
-
-### Direct Usage with ONNX Runtime
-```python
-import onnxruntime as ort
-import numpy as np
-
-# Load NPU-optimized model
-session = ort.InferenceSession(
-    "kokoro-npu-quantized-int8.onnx",
-    providers=['VitisAIExecutionProvider', 'CPUExecutionProvider']
-)
-
-# Example usage (simplified)
-# For complete integration, see Magic Unicorn TTS repository
-```
-
-## Performance Benchmarks
-
-Tested on AMD Ryzen 9 8945HS with NPU Phoenix (AIE-ML) in **TURBO MODE** on NucBox K11:
-
-| Method | Generation Time | Audio Length | RTF | Speedup |
-|--------|-----------------|--------------|-----|---------|
-| CPU Baseline | 1.395s | 7.34s | 0.190 | 1.0x |
-| **NPU Phoenix Basic** | **1.262s** | 8.22s | **0.153** | **1.11x** |
-| NPU Phoenix MLIR-AIE | 1.532s | 8.22s | 0.186 | 0.91x |
-
-## Audio Quality
-
-- **Sample Rate**: 24kHz
-- **Format**: 16-bit PCM  
-- **Quality**: Identical to original Kokoro TTS
-- **Voices**: All 54 voices fully supported
-
-## Files in this Repository
-
-- `kokoro-npu-quantized-int8.onnx` - INT8 quantized model (maximum speed)
-- `kokoro-npu-fp16.onnx` - FP16 optimized model (balanced performance)
-- `voices-v1.0.bin` - Voice embeddings for all 54 voices
-
-## Usage Examples
-
-### Magic Unicorn TTS Integration
-```python
-from huggingface_hub import hf_hub_download
-
-# Download models
-int8_model = hf_hub_download(
-    repo_id="magicunicorn/kokoro-npu-quantized",
-    filename="kokoro-npu-quantized-int8.onnx"
-)
-
-voices = hf_hub_download(
-    repo_id="magicunicorn/kokoro-npu-quantized", 
-    filename="voices-v1.0.bin"
-)
-
-# Use with Magic Unicorn TTS
-# (See complete examples in Magic Unicorn TTS repository)
-```
-
-### Performance Monitoring
 ```python
-import time
-
-# Time the generation process
-start_time = time.time()
-# ... run TTS inference ...
-generation_time = time.time() - start_time
-
-# Calculate Real-Time Factor
-audio_duration = len(audio) / sample_rate
-rtf = generation_time / audio_duration
-
-print(f"Real-Time Factor: {rtf:.3f}")
-print(f"Performance: {'Real-time' if rtf < 1.0 else 'Slower than real-time'}")
-```
-
-## Technical Details
-
-### Quantization Process
-- **Method**: Post-training quantization with calibration dataset
-- **Calibration Data**: 500+ diverse text samples across all voices
-- **Target Hardware**: AMD NPU Phoenix (AIE-ML) architecture
-- **Optimization**: MLIR-AIE kernel compilation
-
-### NPU Architecture Support
-- **Phoenix (AIE-ML)**: Primary target, fully optimized on NucBox K11
-- **Strix Point (AIE2)**: Compatible with enhanced performance
-- **Future NPUs**: Forward compatibility planned
-
-## Installation & Setup
+from unicorn_execution_engine import UnicornTTS
 
-### Complete Magic Unicorn TTS Setup (Recommended)
-```bash
-# Install complete TTS application with NPU support
-curl -fsSL https://raw.githubusercontent.com/Unicorn-Commander/magic-unicorn-tts/main/install.sh | bash
-```
-
-### Manual Model Usage
-```bash
-# Install dependencies
-pip install onnxruntime huggingface_hub
+# Initialize NPU-accelerated TTS
+tts = UnicornTTS(model="kokoro-npu-quantized")
 
-# Download models programmatically
-python -c "
-from huggingface_hub import hf_hub_download
-hf_hub_download('Unicorn-Commander/kokoro-npu-quantized', 'kokoro-npu-quantized-int8.onnx')
-hf_hub_download('Unicorn-Commander/kokoro-npu-quantized', 'voices-v1.0.bin')
-"
+# Generate speech with NPU acceleration
+audio = tts.synthesize("Hello, this is NPU-accelerated speech!")
 ```
 
-## Related Projects
-
-- [Magic Unicorn TTS](https://github.com/Unicorn-Commander/magic-unicorn-tts) - Complete TTS application with web interface
-- [NPU Prebuilds](https://github.com/Unicorn-Commander/npu-prebuilds) - Pre-compiled NPU components  
-- [AMD NPU Utils](https://github.com/Unicorn-Commander/amd-npu-utils) - NPU development tools
-
-## Citation
-
-```bibtex
-@software{magic_unicorn_kokoro_npu,
-  title={Kokoro TTS NPU-Optimized Models},
-  author={Magic Unicorn Technologies},
-  organization={Unicorn Commander},
-  year={2025},
-  url={https://huggingface.co/Unicorn-Commander/kokoro-npu-quantized},
-  note={World's first NPU-accelerated TTS models}
-}
-```
-
-## License
-
-Based on original Kokoro TTS model with additional NPU optimizations.
-See individual component licenses for specific terms.
-
-## Acknowledgments
-
-- **Kokoro TTS**: Original high-quality text-to-speech model
-- **AMD**: Ryzen AI NPU platform and development tools  
-- **VitisAI**: Quantization and optimization framework
-- **MLIR-AIE**: NPU kernel compilation infrastructure
-
 ---
 
-<div align="center">
-  <p>
-    <strong>Powered by Magic Unicorn Technologies 🦄</strong><br>
-    <em>Where AI meets magic</em>
-  </p>
-  <p>
-    <a href="https://unicorncommander.com">Unicorn Commander</a> • 
-    <a href="https://magicunicorn.tech">Magic Unicorn Tech</a>
-  </p>
-</div>
+*🎵 NPU-Accelerated Text-to-Speech*  
+*⚡ Powered by Unicorn Execution Engine*

kokoro-npu-fp16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fa94b712f0894deb4267253ecf81d6af96ed82a4b4b1992885fc6d1487e1b75f
-size 178081100
+oid sha256:1773ebbe4e1ebca782320a6d5e334a03a47a4cd5f0c93283ce4a9e27943dabac
+size 134

kokoro-npu-quantized-int8.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e83b55dbf08d35f3aeb4ee23d5acce5c9601c058102e6f0919c4bad30eeb2d63
-size 127862875
+oid sha256:9e63389b9eaa002fb116f00e5b4d4798ab7d8d7b2edaf6c88b0796096a2d95b8
+size 134

voices-v1.0.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bca610b8308e8d99f32e6fe4197e7ec01679264efed0cac9140fe9c29f1fbf7d
-size 28214398
+oid sha256:c29abef6993ac4cc7a06c64ffdeec944fc4bf29ed2c687f7b7138011b83188bb
+size 133