Update README.md
Browse files
README.md
CHANGED
|
@@ -152,7 +152,9 @@ Benchmarks based on standard text generation ("Bonjour le monde") using the C++
|
|
| 152 |
|
| 153 |
> **Note:** The Vulkan backend is significantly faster for the HiFiGAN part of the pipeline compared to CUDA due to optimized command buffers and reduced kernel launch overhead for small convolutions.
|
| 154 |
>
|
| 155 |
-
|
|
|
|
|
|
|
| 156 |
|
| 157 |
* Split Model Architecture (GPT2 + HiFiGAN):
|
| 158 |
|
|
@@ -164,12 +166,21 @@ Benchmarks based on standard text generation ("Bonjour le monde") using the C++
|
|
| 164 |
|
| 165 |
* Advanced Text Preprocessing:
|
| 166 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 167 |
* **MeCab** Integration: Native Japanese tokenization and segmentation for natural prosody.
|
| 168 |
|
| 169 |
* Romanization: Automatic romanization for Chinese (Pinyin), Japanese (Romaji), and Korean (Revised Romanization).
|
| 170 |
|
| 171 |
* Num2Words: Converts numbers (e.g., "123") into words ("one hundred twenty-three") across all 17
|
| 172 |
|
|
|
|
|
|
|
| 173 |
## 🛠️ Usage
|
| 174 |
|
| 175 |
This repository contains the converted `.zip` and `.gguf` weights required by the ATTS1HG1 software.
|
|
@@ -224,10 +235,6 @@ Test the quality with these pre-generated samples covering various speakers and
|
|
| 224 |
## 📜 License
|
| 225 |
|
| 226 |
This project uses the weights from **Coqui XTTS-v2**, which is licensed under the **Coqui Public Model License (CPML)**.
|
| 227 |
-
* **Non-commercial use:** You can use this model for personal, educational, and non-commercial projects.
|
| 228 |
-
* **Commercial use:** Requires a license from Coqui (check their repository for details).
|
| 229 |
-
|
| 230 |
-
The C++ code (inference engine) is available under the MIT License (see GitHub).
|
| 231 |
|
| 232 |
---
|
| 233 |
*Credits: Based on the excellent work by Coqui.ai and the GGML library by ggerganov.*
|
|
|
|
| 152 |
|
| 153 |
> **Note:** The Vulkan backend is significantly faster for the HiFiGAN part of the pipeline compared to CUDA due to optimized command buffers and reduced kernel launch overhead for small convolutions.
|
| 154 |
>
|
| 155 |
+
|
| 156 |
+
|
| 157 |
+
## 🔧 Key Technical Upgrades
|
| 158 |
|
| 159 |
* Split Model Architecture (GPT2 + HiFiGAN):
|
| 160 |
|
|
|
|
| 166 |
|
| 167 |
* Advanced Text Preprocessing:
|
| 168 |
|
| 169 |
+
* Automatic Language Detection (New!):
|
| 170 |
+
|
| 171 |
+
ATTS now features a robust LanguageDetector class that analyzes input text.
|
| 172 |
+
Uses Script Range (Unicode blocks), Dictionary (common words), and N-Grams (suffixes/patterns).
|
| 173 |
+
Supports 17 languages with confidence scoring and fallback logic.
|
| 174 |
+
In "Auto Mode" (LangDirIndex == 0), the system automatically selects the language for synthesis.
|
| 175 |
+
|
| 176 |
* **MeCab** Integration: Native Japanese tokenization and segmentation for natural prosody.
|
| 177 |
|
| 178 |
* Romanization: Automatic romanization for Chinese (Pinyin), Japanese (Romaji), and Korean (Revised Romanization).
|
| 179 |
|
| 180 |
* Num2Words: Converts numbers (e.g., "123") into words ("one hundred twenty-three") across all 17
|
| 181 |
|
| 182 |
+
|
| 183 |
+
|
| 184 |
## 🛠️ Usage
|
| 185 |
|
| 186 |
This repository contains the converted `.zip` and `.gguf` weights required by the ATTS1HG1 software.
|
|
|
|
| 235 |
## 📜 License
|
| 236 |
|
| 237 |
This project uses the weights from **Coqui XTTS-v2**, which is licensed under the **Coqui Public Model License (CPML)**.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 238 |
|
| 239 |
---
|
| 240 |
*Credits: Based on the excellent work by Coqui.ai and the GGML library by ggerganov.*
|