Update app.py

app.py

@@ -1,188 +1,246 @@
-import gradio as gr
 import os
+import spaces
+import torch
+import torchaudio
+import gradio as gr
+import logging
+from whosper import WhosperTranscriber
+
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+
+if torch.cuda.is_available():
+    device = "cuda"
+    logger.info("Using CUDA for inference.")
+elif torch.backends.mps.is_available():
+    device = "mps"
+    logger.info("Using MPS for inference.")
+else:
+    device = "cpu"
+    logger.info("Using CPU for inference.")
+
+
+model_id = "sudoping01/maliba-asr-v1"
+transcriber = WhosperTranscriber(model_id=model_id)
+logger.info(f"Transcriber initialized with model: {model_id}")
+
+def resample_audio(audio_path, target_sample_rate=16000):
 
-def build_interface():
     """
-    Builds an enhanced Gradio interface for Bambara speech recognition.
+    Converts the audio file to the target sampling rate (16000 Hz).
+    
+    Args:
+        audio_path (str): Path to the audio file.
+        target_sample_rate (int): The desired sample rate.
+    Returns:
+        A tensor containing the resampled audio data and the target sample rate.
     """
-    example_files = get_example_files()
+    try:
+        waveform, original_sample_rate = torchaudio.load(audio_path)
+        
+        if original_sample_rate != target_sample_rate:
+            resampler = torchaudio.transforms.Resample(
+                orig_freq=original_sample_rate, 
+                new_freq=target_sample_rate
+            )
+            waveform = resampler(waveform)
+        
+        return waveform, target_sample_rate
+    except Exception as e:
+        logger.error(f"Error resampling audio: {e}")
+        raise e
+
+@spaces.GPU()
+def transcribe_audio(audio_file):
+
+    """
+    Transcribes the provided audio file into Bambara text using Whosper.
     
-    custom_css = """
-    .gr-button-primary {
-        background-color: #2c5282 !important;
-        color: white !important;
-        border-radius: 8px !important;
-        font-weight: bold !important;
-    }
-    .gr-button-secondary {
-        background-color: #e2e8f0 !important;
-        color: #2d3748 !important;
-        border-radius: 8px !important;
-    }
-    .example-container {
-        background-color: #f7fafc;
-        padding: 16px;
-        border-radius: 8px;
-        margin-top: 16px;
-    }
-    .gr-textbox {
-        border-radius: 8px !important;
-        border: 1px solid #cbd5e0 !important;
-    }
-    .gr-audio {
-        border-radius: 8px !important;
-    }
-    .header {
-        text-align: center;
-        color: #2d3748;
-    }
-    .info-section {
-        background-color: #edf2f7;
-        padding: 16px;
-        border-radius: 8px;
-        margin-top: 16px;
-    }
+    Args:
+        audio_file: The path to the audio file to transcribe.
+    Returns:
+        A string representing the transcribed Bambara text.
     """
 
-    with gr.Blocks(title="Bambara Speech Recognition", css=custom_css) as demo:
-        # Header
+    if audio_file is None:
+        return "Please provide an audio file for transcription."
+    
+    try:
+        logger.info(f"Transcribing audio file: {audio_file}")
+        
+
+        result = transcriber.transcribe_audio(audio_file)
+        
+        logger.info("Transcription successful.")
+        return result.get("text", "")
+        
+    except Exception as e:
+        logger.error(f"Transcription failed: {e}")
+        return f"Error during transcription: {str(e)}"
+
+def get_example_files(directory="./examples"):
+
+    """
+    Returns a list of audio files from the examples directory.
+    
+    Args:
+        directory (str): The directory to search for audio files.
+    Returns:
+        list: A list of paths to the audio files.
+    """
+
+    if not os.path.exists(directory):
+        logger.warning(f"Examples directory {directory} not found.")
+        return []
+    
+
+    audio_extensions = ['.wav', '.mp3', '.m4a', '.flac', '.ogg']
+    audio_files = []
+    
+    try:
+        files = os.listdir(directory)
+        for file in files:
+            if any(file.lower().endswith(ext) for ext in audio_extensions):
+                full_path = os.path.abspath(os.path.join(directory, file))
+                audio_files.append(full_path)
+        
+        logger.info(f"Found {len(audio_files)} example audio files.")
+        return audio_files[:5]  
+        
+    except Exception as e:
+        logger.error(f"Error reading examples directory: {e}")
+        return []
+
+def build_interface():
+    """
+    Builds the Gradio interface for Bambara speech recognition.
+    """
+
+    example_files = get_example_files()
+    
+    with gr.Blocks(title="Bambara Speech Recognition") as demo:
         gr.Markdown(
             """
-            <h1 class="header">🎤 Bambara Speech Recognition</h1>
-            <p style="text-align: center; color: #4a5568;">
-                Powered by <b>MALIBA-AI</b> | Convert Bambara speech to text effortlessly
-            </p>
+            # 🎤 Bambara Automatic Speech Recognition
+            
+            **Powered by MALIBA-AI**
+            
+            Convert Bambara speech to text using our state-of-the-art ASR model. You can either:
+            - 🎙️ **Record** your voice directly
+            - 📁 **Upload** an audio file
+            - 🎵 **Try** our example audio files
+            
+            ## Supported Audio Formats
+            WAV, MP3, M4A, FLAC, OGG
             """
         )
-
-        # Main interaction section
+        
         with gr.Row():
-            with gr.Column(scale=1):
-                gr.Markdown("### 🎙️ Record or Upload Audio")
+            with gr.Column():
+
                 audio_input = gr.Audio(
-                    label="Record or Upload Audio",
+                    label="🎤 Record or Upload Audio",
                     type="filepath",
-                    sources=["microphone", "upload"],
-                    show_label=False
+                    sources=["microphone", "upload"]
                 )
-                audio_preview = gr.Audio(
-                    label="Preview Your Audio",
-                    interactive=False,
-                    visible=False
+                
+                transcribe_btn = gr.Button(
+                    "🔄 Transcribe Audio", 
+                    variant="primary",
+                    size="lg"
                 )
+                
 
-                with gr.Row():
-                    transcribe_btn = gr.Button(
-                        "🔄 Transcribe Audio",
-                        variant="primary",
-                        size="lg"
-                    )
-                    clear_btn = gr.Button(
-                        "🗑️ Clear",
-                        variant="secondary",
-                        size="lg"
-                    )
-
-            with gr.Column(scale=1):
-                gr.Markdown("### 📝 Transcription Output")
+                clear_btn = gr.Button("🗑️ Clear", variant="secondary")
+            
+            with gr.Column():
                 output_text = gr.Textbox(
-                    label="Transcribed Text (Bambara)",
-                    lines=6,
+                    label="📝 Transcribed Text (Bambara)",
+                    lines=8,
                     placeholder="Your transcribed Bambara text will appear here...",
-                    interactive=False,
-                    show_copy_button=True
-                )
-                status_message = gr.Markdown(
-                    value="",
-                    visible=False
+                    interactive=False
                 )
+        
 
-        # Example audio section
         if example_files:
-            gr.Markdown("## 🎵 Try Example Audio Files")
-            with gr.Group(elem_classes="example-container"):
-                gr.Markdown(
-                    """
-                    Listen to these sample Bambara audio files and transcribe them with one click.
-                    """
-                )
-                for idx, file in enumerate(example_files):
-                    with gr.Row():
-                        gr.Audio(
-                            value=file,
-                            label=f"Example {idx + 1}: {os.path.basename(file)}",
-                            interactive=False,
-                            show_label=True
-                        )
-                        gr.Button(
-                            f"Transcribe Example {idx + 1}",
-                            variant="primary",
-                            size="sm"
-                        ).click(
-                            fn=transcribe_audio,
-                            inputs=gr.State(value=file),
-                            outputs=[output_text, status_message],
-                            show_progress=True,
-                            _js="() => {return {show_progress: true}}"
-                        )
+            gr.Markdown("## 🎵 Try These Examples")
+            gr.Examples(
+                examples=[[f] for f in example_files],
+                inputs=[audio_input],
+                outputs=output_text,
+                fn=transcribe_audio,
+                cache_examples=False,
+                label="Example Audio Files"
+            )
+        
 
         gr.Markdown(
             """
-            <div class="info-section">
-            ## ℹ️ How to Use
+            ---
+            
+            ## ℹ️ About This Model
+            
+            - **Model:** [sudoping01/maliba-asr-v1](https://huggingface.co/sudoping01/maliba-asr-v1)
+            - **Developer:** MALIBA-AI
+            - **Language:** Bambara (bm)
+            - **Task:** Automatic Speech Recognition (ASR)
+            - **Sample Rate:** 16kHz (automatically resampled)
+            
+            ## 🚀 How to Use
             
-            1. **Record**: Click the microphone to speak in Bambara.
-            2. **Upload**: Select an audio file (WAV, MP3, M4A, FLAC, OGG).
-            3. **Transcribe**: Click "Transcribe Audio" or try an example.
-            4. **View**: See the transcribed text in Bambara.
+            1. **Record Audio:** Click the microphone button and speak in Bambara
+            2. **Upload File:** Click the upload button to select an audio file
+            3. **Transcribe:** Click the "Transcribe Audio" button
+            4. **View Results:** See your transcribed text in Bambara
             
-            ## 📊 Model Details
+            ## 📊 Performance Notes
             
-            - **Model**: [sudoping01/maliba-asr-v1](https://huggingface.co/sudoping01/maliba-asr-v1)
-            - **Language**: Bambara (bm)
-            - **Sample Rate**: 16kHz (auto-resampled)
-            - **Best for**: Clear speech with minimal background noise
-            </div>
+            - Best results with clear speech and minimal background noise
+            - Supports various audio formats and durations
+            - Optimized for Bambara language patterns and phonetics
             """
         )
+        
 
-
-        def update_audio_preview(audio_file):
-            return gr.update(value=audio_file, visible=True), ""
-
-        def clear_inputs():
-            return None, "", gr.update(visible=False), ""
-
-        def transcribe_with_status(audio_file):
-            if not audio_file:
-                return "", "**Error**: Please provide an audio file."
-            result = transcribe_audio(audio_file)
-            if "Error" in result:
-                return result, f"**Error**: {result}"
-            return result, "**Success**: Transcription completed!"
-
-
-        audio_input.change(
-            fn=update_audio_preview,
-            inputs=audio_input,
-            outputs=[audio_preview, status_message]
-        ).then(
-            fn=transcribe_with_status,
-            inputs=audio_input,
-            outputs=[output_text, status_message],
+        transcribe_btn.click(
+            fn=transcribe_audio,
+            inputs=[audio_input],
+            outputs=output_text,
             show_progress=True
         )
+        
+        clear_btn.click(
+            fn=lambda: (None, ""),
+            outputs=[audio_input, output_text]
+        )
+        
 
-        transcribe_btn.click(
-            fn=transcribe_with_status,
-            inputs=audio_input,
-            outputs=[output_text, status_message],
+        audio_input.change(
+            fn=transcribe_audio,
+            inputs=[audio_input],
+            outputs=output_text,
             show_progress=True
         )
+    
+    return demo
 
-        clear_btn.click(
-            fn=clear_inputs,
-            outputs=[audio_input, output_text, audio_preview, status_message]
-        )
+def main():
+    """
+    Main function to launch the Gradio interface.
+    """
+    logger.info("Starting Bambara ASR Gradio interface.")
+    
+
+    interface = build_interface()
+    interface.launch(
+        share=False,
+        server_name="0.0.0.0",
+        server_port=7860
+    )
+    
+    logger.info("Gradio interface launched successfully.")
 
-    return demo
+if __name__ == "__main__":
+    main()