app.py

import gradio as gr
from transformers import pipeline
import torch

# Initialize the speech recognition pipeline
print("Loading Whisper Lozi model...")
try:
    # Use the specific Lozi model
    transcriber = pipeline(
        "automatic-speech-recognition",
        model="simzacademy/whisper-small-lozi1",
        device=0 if torch.cuda.is_available() else -1  # Use GPU if available
    )
    print("Model loaded successfully!")
except Exception as e:
    print(f"Error loading model: {e}")
    transcriber = None

def transcribe_audio(audio):
    """
    Transcribe audio to text using the Whisper Lozi model
    
    Args:
        audio: Audio file path or tuple (sample_rate, audio_data)
    
    Returns:
        Transcribed text
    """
    if transcriber is None:
        return "Error: Model failed to load. Please check your installation."
    
    if audio is None:
        return "Please provide an audio file or recording."
    
    try:
        # Transcribe the audio
        result = transcriber(audio)
        return result["text"]
    except Exception as e:
        return f"Error during transcription: {str(e)}"

# Create the Gradio interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown(
        """
        # 🎤 Lozi Speech-to-Text Interface
        ### Powered by Whisper Small Lozi Model
        
        This interface uses the `simzacademy/whisper-small-lozi1` model to transcribe 
        Lozi language speech to text.
        """
    )
    
    with gr.Row():
        with gr.Column():
            # Audio input - supports both recording and file upload
            audio_input = gr.Audio(
                sources=["microphone", "upload"],
                type="filepath",
                label="Record or Upload Audio"
            )
            
            transcribe_btn = gr.Button("🔄 Transcribe", variant="primary", size="lg")
        
        with gr.Column():
            output_text = gr.Textbox(
                label="Transcription",
                placeholder="Your transcription will appear here...",
                lines=10
            )
    
    gr.Markdown(
        """
        ### 📋 Instructions:
        1. **Record**: Click the microphone icon to record audio directly
        2. **Upload**: Or click to upload an audio file (MP3, WAV, etc.)
        3. **Transcribe**: Click the "Transcribe" button to convert speech to text
        4. **View**: The transcribed text will appear on the right
        
        ### ℹ️ Notes:
        - Speak clearly in Lozi for best results
        - The model works best with clear audio and minimal background noise
        - First transcription may take longer as the model loads
        """
    )
    
    # Set up the transcription action
    transcribe_btn.click(
        fn=transcribe_audio,
        inputs=audio_input,
        outputs=output_text
    )
    
    # Also allow Enter key to trigger transcription
    audio_input.change(
        fn=lambda: gr.update(interactive=True),
        outputs=transcribe_btn
    )

# Launch the interface
if __name__ == "__main__":
    demo.launch(
        share=False,  # Set to True to create a public link
        server_name="0.0.0.0",  # Allow access from network
        server_port=7860
    )