app.py

import gradio as gr
from transformers import pipeline
import os
import numpy as np
import torch

# Load the model
print("Loading model...")
model_id = "badrex/mms-300m-arabic-dialect-identifier"
classifier = pipeline("audio-classification", model=model_id)
print("Model loaded successfully")

# Define dialect mapping
dialect_mapping = {
    "MSA": "Modern Standard Arabic",
    "Egyptian": "Egyptian Arabic",
    "Gulf": "Gulf Arabic",
    "Levantine": "Levantine Arabic",
    "Maghrebi": "Maghrebi Arabic"
}

def predict_dialect(audio):
    if audio is None:
        return {"Error": 1.0}
    
    # The audio input from Gradio is a tuple of (sample_rate, audio_array)
    sr, audio_array = audio
    
    # Process the audio input
    if len(audio_array.shape) > 1:
        audio_array = audio_array.mean(axis=1)  # Convert stereo to mono

    # Convert audio to float32 if it's not already (fix for Chrome recording issue)
    if audio_array.dtype != np.float32:
        # Normalize to [-1, 1] range as expected by the model
        if audio_array.dtype == np.int16:
            audio_array = audio_array.astype(np.float32) / 32768.0
        else:
            audio_array = audio_array.astype(np.float32)
    
    print(f"Processing audio: sample rate={sr}, shape={audio_array.shape}")
    
    # Classify the dialect
    predictions = classifier({"sampling_rate": sr, "raw": audio_array})
    
    # Format results for display
    results = {}
    for pred in predictions:
        dialect_name = dialect_mapping.get(pred['label'], pred['label'])
        results[dialect_name] = float(pred['score'])
    
    return results

# Manually prepare example file paths without metadata
examples = []
examples_dir = "examples"
if os.path.exists(examples_dir):
    for filename in os.listdir(examples_dir):
        if filename.endswith((".wav", ".mp3", ".ogg")):
            examples.append([os.path.join(examples_dir, filename)])
    
    print(f"Found {len(examples)} example files")
else:
    print("Examples directory not found")

# Create the Gradio interface
demo = gr.Interface(
    fn=predict_dialect,
    inputs=gr.Audio(),
    outputs=gr.Label(num_top_classes=5, label="Predicted Dialect"),
    title="🎙️ Arabic Dialect Identification in Speech!",
    description="""
        Use this AI-powered tool to identify five major Arabic varieties from just a short audio clip:

        ✦ Modern Standard Arabic (MSA) - The formal language of media and education

        ✦ Egyptian Arabic - The dialect of Cairo, Alexandria, and popular Arabic cinema 

        ✦ Gulf Arabic - Spoken across Saudi Arabia, UAE, Kuwait, Qatar, Bahrain, and Oman

        ✦ Levantine Arabic - The dialect of Syria, Lebanon, Jordan, and Palestine

        ✦ Maghrebi Arabic - The distinctive varieties of Morocco, Algeria, Tunisia, and Libya

        Simply **upload an audio file** or **record yourself speaking** to see which dialect you match! Perfect for language learners, linguistics enthusiasts, or anyone curious about Arabic language variation.
        
        The demo is based on a Transformer model adapted for the ADI task [badrex/mms-300m-arabic-dialect-identifier](https://huggingface.co/badrex/mms-300m-arabic-dialect-identifier). 
        
        Developed with ❤️🤍💚 by [Badr Alabsi](https://badrex.github.io/)""", 
    examples=examples if examples else None,
    cache_examples=False,  # Disable caching to avoid issues
    flagging_mode=None
)

# Launch the app
demo.launch()