Upload all space files

README.md

@@ -1,12 +1,185 @@
----
-title: Speecht5 Darija
-emoji: 📈
-colorFrom: yellow
-colorTo: purple
-sdk: gradio
-sdk_version: 5.27.0
-app_file: app.py
-pinned: false
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Moroccan Darija Text-to-Speech Model
+
+This project implements a Text-to-Speech (TTS) system for Moroccan Darija using the SpeechT5 architecture. It's fine-tuned on the DODa-audio-dataset to generate natural-sounding Darija speech from text input.
+
+## Table of Contents
+- [Dataset Overview](#dataset-overview)
+- [Project Structure](#project-structure)
+- [Installation](#installation)
+- [Model Training](#model-training)
+- [Inference](#inference)
+- [Gradio Demo](#gradio-demo)
+- [Project Features](#project-features)
+- [Potential Applications](#potential-applications)
+- [Limitations and Future Work](#limitations-and-future-work)
+- [License](#license)
+
+## Dataset Overview
+
+The **DODa audio dataset** contains 12,743 sentences recorded by 7 contributors (4 females, 3 males). Key characteristics:
+
+- Audio recordings standardized at 16kHz sample rate
+- Multiple text representations (Latin script, Arabic script, and English translations)
+- High-quality recordings with manual corrections
+
+### Dataset Structure
+| Column Name | Description |
+|-------------|-------------|
+| **audio** | Speech recordings for Darija sentences |
+| **darija_Ltn** | Darija sentences using Latin letters |
+| **darija_Arab_new** | Corrected Darija sentences using Arabic script |
+| **english** | English translation of Darija sentences |
+| **darija_Arab_old** | Original (uncorrected) Darija sentences in Arabic script |
+
+### Speaker Distribution
+The dataset includes recordings from 7 speakers (4 females, 3 males) with the following distribution:
+```
+Samples 0-999     -> Female 1
+Samples 1000-1999 -> Male 3
+Samples 2000-2730 -> Female 2
+Samples 2731-2800 -> Male 1
+Samples 2801-3999 -> Male 2
+Samples 4000-4999 -> Male 1
+Samples 5000-5999 -> Female 3
+Samples 6000-6999 -> Male 1
+Samples 7000-7999 -> Female 4
+Samples 8000-8999 -> Female 1
+Samples 9000-9999 -> Male 2
+Samples 10000-11999 -> Male 1
+Samples 12000-12350 -> Male 2
+Samples 12351-12742 -> Male 1
+```
+
+
+
+## Installation
+
+To set up the project environment:
+
+```bash
+# Clone the repository
+git clone https://github.com/yourusername/darija-tts.git
+cd darija-tts
+
+# Create a virtual environment (optional but recommended)
+python -m venv venv
+source venv/bin/activate  # On Windows: venv\Scriptsctivate
+
+# Install dependencies
+pip install -r requirements.txt
+```
+
+## Model Training
+
+The model training process involves:
+
+1. **Data Loading**: Loading the DODa audio dataset from Hugging Face
+2. **Data Preprocessing**: Normalizing text and extracting speaker embeddings
+3. **Model Setup**: Configuring the SpeechT5 model for Darija TTS
+4. **Training**: Fine-tuning the model using the prepared dataset
+
+To run the training:
+
+```bash
+# Open the Jupyter notebook
+jupyter notebook notebooks/train_darija_tts.ipynb
+```
+
+Key training parameters:
+- Learning rate: 1e-4
+- Batch size: 4 (with gradient accumulation: 8)
+- Training steps: 1000
+- Evaluation frequency: Every 100 steps
+
+## Inference
+
+To generate speech from text after training:
+
+```python
+from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
+import torch
+import soundfile as sf
+
+# Load models
+model_path = "./models/speecht5_finetuned_Darija"
+processor = SpeechT5Processor.from_pretrained(model_path)
+model = SpeechT5ForTextToSpeech.from_pretrained(model_path)
+vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
+
+# Load speaker embedding
+speaker_embedding = torch.load("./data/speaker_embeddings/female_embedding.pt")
+
+# Normalize and process input text
+text = "Salam, kifach nta lyoum?"
+inputs = processor(text=text, return_tensors="pt")
+
+# Generate speech
+speech = model.generate_speech(inputs["input_ids"], speaker_embedding, vocoder=vocoder)
+
+# Save audio file
+sf.write("output.wav", speech.numpy(), 16000)
+```
+
+## Gradio Demo
+
+The project includes a Gradio demo that provides a user-friendly interface for text-to-speech conversion:
+
+```bash
+# Run the demo locally
+cd demo
+python app.py
+```
+
+The demo features:
+- Text input field for Darija text (Latin script)
+- Voice selection (male/female)
+- Speech speed adjustment
+- Audio playback of generated speech
+
+### Deploying to Hugging Face Spaces
+
+To deploy the demo to Hugging Face Spaces:
+
+1. Push your model to the Hugging Face Hub
+2. Create a new Space with the Gradio SDK
+3. Upload the demo files to the Space
+
+See the notebook for detailed deployment instructions.
+
+## Project Features
+
+- **Multi-Speaker TTS**: Generate speech in both male and female voices
+- **Voice Cloning**: Utilizes speaker embeddings for voice characteristics preservation
+- **Speed Control**: Adjust the speech rate as needed
+- **Text Normalization**: Handles various text inputs through proper normalization
+
+## Potential Applications
+
+- **Voice Assistants**: Build voice assistants that speak Moroccan Darija
+- **Accessibility Tools**: Create tools for people with visual impairments
+- **Language Learning**: Develop applications for learning Darija pronunciation
+- **Content Creation**: Generate voiceovers for videos or audio content
+- **Public Announcements**: Create automated announcement systems in Darija
+
+## Limitations and Future Work
+
+Current limitations:
+- The model may struggle with code-switching between Darija and other languages
+- Pronunciation of certain loanwords might be inconsistent
+- Limited emotional range in the generated speech
+
+Future improvements:
+- Fine-tune with more diverse speech data
+- Implement emotion control for expressive speech
+- Add support for Arabic script input
+- Develop a multilingual version supporting Darija, Arabic, and French
+
+## License
+
+This project is released under the MIT License. The DODa audio dataset is also available under the MIT License.
+
+## Acknowledgments
+
+- The [DODa audio dataset](https://huggingface.co/datasets/atlasia/DODa-audio-dataset) creators
+- Hugging Face for the Transformers library and model hosting
+- Microsoft Research for the SpeechT5 model architecture

app.py

@@ -0,0 +1,255 @@
+
+import torch
+import soundfile as sf
+import os
+import re
+from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
+from speechbrain.pretrained import EncoderClassifier
+
+# Define paths and device
+model_path = "HAMMALE/speecht5-darija"  # Path to your model on HF Hub
+device = "cuda" if torch.cuda.is_available() else "cpu"
+print(f"Using device: {device}")
+
+# Load models
+processor = SpeechT5Processor.from_pretrained(model_path)
+model = SpeechT5ForTextToSpeech.from_pretrained(model_path).to(device)
+vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
+
+# Load speaker embedding model
+speaker_model = EncoderClassifier.from_hparams(
+    source="speechbrain/spkrec-xvect-voxceleb",
+    run_opts={"device": device},
+    savedir=os.path.join("/tmp", "spkrec-xvect-voxceleb"),
+)
+
+# Load pre-computed speaker embeddings
+male_embedding = torch.load("male_embedding.pt") if os.path.exists("male_embedding.pt") else torch.randn(1, 512)
+female_embedding = torch.load("female_embedding.pt") if os.path.exists("female_embedding.pt") else torch.randn(1, 512)
+
+# Text normalization function
+def normalize_text(text):
+    """Normalize text for TTS processing"""
+    text = text.lower()
+    # Keep letters, numbers, spaces and apostrophes - fixed regex
+    text = re.sub(r'[^\w\s\'\u0600-\u06FF]', '', text)
+    text = ' '.join(text.split())
+    return text
+
+# Function to synthesize speech
+def synthesize_speech(text, voice_type="male", speed=1.0):
+    """Generate speech from text using the specified voice type"""
+    try:
+        # Select speaker embedding based on voice type
+        if voice_type == "male":
+            speaker_embeddings = male_embedding.to(device)
+        else:
+            speaker_embeddings = female_embedding.to(device)
+
+        # Normalize and tokenize input text
+        normalized_text = normalize_text(text)
+        inputs = processor(text=normalized_text, return_tensors="pt").to(device)
+
+        # Generate speech
+        with torch.no_grad():
+            speech = model.generate_speech(
+                inputs["input_ids"],
+                speaker_embeddings,
+                vocoder=vocoder
+            )
+
+        # Convert to numpy array and adjust speed if needed
+        speech_np = speech.cpu().numpy()
+
+        # Apply speed adjustment (simple resampling)
+        if speed != 1.0:
+            # This is a simple approach - for production use a proper resampling library
+            import numpy as np
+            from scipy import signal
+            sample_rate = 16000
+            new_length = int(len(speech_np) / speed)
+            speech_np = signal.resample(speech_np, new_length)
+
+        # Save temporary audio file
+        output_file = "output_speech.wav"
+        sf.write(output_file, speech_np, 16000)
+
+        return output_file, None
+
+    except Exception as e:
+        return None, f"Error generating speech: {str(e)}"
+
+# Gradio imports need to be added
+import gradio as gr
+
+# Custom CSS for better design
+custom_css = """
+.gradio-container {
+    font-family: 'Poppins', 'Arial', sans-serif;
+    max-width: 750px;
+    margin: auto;
+}
+
+.main-header {
+    background: linear-gradient(90deg, #c31432, #240b36);
+    color: white;
+    padding: 1.5em;
+    border-radius: 10px;
+    text-align: center;
+    margin-bottom: 1em;
+    box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
+}
+
+.main-header h1 {
+    font-size: 2.2em;
+    margin-bottom: 0.3em;
+}
+
+.main-header p {
+    font-size: 1.1em;
+    opacity: 0.9;
+}
+
+footer {
+    text-align: center;
+    margin-top: 2em;
+    color: #555;
+    font-size: 0.9em;
+}
+
+.flag-icon {
+    width: 24px;
+    height: 24px;
+    vertical-align: middle;
+    margin-right: 8px;
+}
+
+.example-header {
+    font-weight: bold;
+    color: #c31432;
+    margin-top: 1em;
+}
+
+.info-box {
+    background-color: #f9f9f9;
+    border-left: 4px solid #c31432;
+    padding: 1em;
+    margin: 1em 0;
+    border-radius: 5px;
+}
+
+.voice-selector {
+    display: flex;
+    justify-content: center;
+    gap: 20px;
+    margin: 10px 0;
+}
+
+.voice-option {
+    border: 2px solid #ddd;
+    border-radius: 10px;
+    padding: 10px 15px;
+    transition: all 0.3s ease;
+    cursor: pointer;
+}
+
+.voice-option.selected {
+    border-color: #c31432;
+    background-color: #fff5f5;
+}
+
+.slider-container {
+    margin: 20px 0;
+}
+"""
+
+# Create Gradio interface with improved design
+with gr.Blocks(css=custom_css) as demo:
+    gr.HTML(
+        """
+        <div class="main-header">
+            <h1>🇲🇦 Moroccan Darija Text-to-Speech 🎧</h1>
+            <p>Convert Moroccan Arabic (Darija) text into natural-sounding speech</p>
+        </div>
+        """
+    )
+    
+    with gr.Row():
+        with gr.Column():
+            gr.HTML(
+                """
+                <div class="info-box">
+                    <p>This model was fine-tuned on the DODa audio dataset to produce high-quality
+                    Darija speech from text input. You can adjust the voice and speed below.</p>
+                </div>
+                """
+            )
+            
+            text_input = gr.Textbox(
+                label="Enter Darija Text",
+                placeholder="Kteb chi jomla b darija hna...",
+                lines=3
+            )
+            
+            with gr.Row():
+                voice_type = gr.Radio(
+                    ["male", "female"],
+                    label="Voice Type",
+                    value="male"
+                )
+                
+            speed = gr.Slider(
+                minimum=0.5,
+                maximum=2.0,
+                value=1.0,
+                step=0.1,
+                label="Speech Speed"
+            )
+            
+            generate_btn = gr.Button("Generate Speech", variant="primary")
+            
+            gr.HTML(
+                """
+                <div class="example-header">Example phrases:</div>
+                <ul>
+                    <li>"Ana Nadi Bezzaaf hhh"</li>
+                    <li>"Lyoum ajwaa zwina bezzaf."</li>
+                    <li>"lmaghrib ahssan blad fi l3alam "</li>
+                </ul>
+                """
+            )
+            
+        with gr.Column():
+            audio_output = gr.Audio(label="Generated Speech")
+            error_output = gr.Textbox(label="Error (if any)", visible=False)
+            
+            gr.Examples(
+                examples=[
+                    ["Ana Nadi Bezzaaf hhh", "male", 1.0],
+                    ["Lyoum ajwaa zwina bezzaf.", "female", 1.0],
+                    ["lmaghrib ahssan blad fi l3alam", "male", 1.0],
+                    ["Filistine hora mina lbar ila lbahr", "female", 0.8],
+                ],
+                inputs=[text_input, voice_type, speed],
+                outputs=[audio_output, error_output],
+                fn=synthesize_speech
+            )
+    
+    gr.HTML(
+        """
+        <footer>
+            <p>Developed by HAMMALE | Powered by Microsoft SpeechT5 | Data: DODa</p>
+        </footer>
+        """
+    )
+    
+    # Set button click action
+    generate_btn.click(
+        fn=synthesize_speech,
+        inputs=[text_input, voice_type, speed],
+        outputs=[audio_output, error_output]
+    )
+
+# Launch the demo
+if __name__ == "__main__":
+    demo.launch()

female_embedding.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f87744b22b910652a6cc2645e99e9c2d6d63d4d8cb5e803dd0b8b520863ea0f
+size 3273

male_embedding.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:268bdfbc37c7bab90de174bc30105c509d8149b1b8be2ddd888353ab924d6c7c
+size 3263

requirements.txt

@@ -0,0 +1,6 @@
+gradio>=3.50.2
+torch>=2.0.0
+transformers>=4.32.0
+speechbrain==0.5.16
+soundfile>=0.12.1
+scipy>=1.11.1