Update app.py

app.py

@@ -2,13 +2,13 @@ import os
 import torch
 import gradio as gr
 import logging
+import subprocess
 from pydub import AudioSegment
 from pydub.exceptions import CouldntDecodeError
 from transformers import pipeline, AutoModelForSpeechSeq2Seq, AutoProcessor
 from pathlib import Path
 from tempfile import NamedTemporaryFile
 from datetime import timedelta
-import time
 
 # Setup logging
 logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
@@ -21,6 +21,16 @@ DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 TORCH_DTYPE = torch.float16 if torch.cuda.is_available() else torch.float32
 SUPPORTED_FORMATS = {".wav", ".mp3", ".m4a"}
 
+# Check for ffmpeg availability
+def check_ffmpeg():
+    try:
+        subprocess.run(["ffmpeg", "-version"], capture_output=True, check=True)
+        logger.info("ffmpeg is installed and accessible.")
+        return True
+    except (subprocess.CalledProcessError, FileNotFoundError):
+        logger.error("ffmpeg is not installed or not found in PATH.")
+        return False
+
 # Initialize model and pipeline
 def initialize_pipeline():
     try:
@@ -46,38 +56,43 @@ def initialize_pipeline():
 # Convert audio if needed
 def convert_to_wav(audio_path: str) -> str:
     try:
+        if not check_ffmpeg():
+            raise RuntimeError("ffmpeg is required to process .m4a files. Please install ffmpeg and ensure it's in your PATH.")
         ext = str(Path(audio_path).suffix).lower()
         if ext not in SUPPORTED_FORMATS:
             raise ValueError(f"Unsupported audio format: {ext}. Supported formats: {', '.join(SUPPORTED_FORMATS)}")
         if ext != ".wav":
+            logger.info(f"Converting {ext} file to WAV: {audio_path}")
             audio = AudioSegment.from_file(audio_path)
             wav_path = str(Path(audio_path).with_suffix(".converted.wav"))
             audio.export(wav_path, format="wav")
+            logger.info(f"Conversion successful: {wav_path}")
             return wav_path
         return audio_path
     except CouldntDecodeError:
-        logger.error(f"Failed to decode audio file: {audio_path}")
-        raise ValueError("Audio file is corrupted or in an unsupported format.")
+        logger.error(f"Failed to decode .m4a file: {audio_path}")
+        raise ValueError("The .m4a file is corrupted or not supported. Ensure it's a valid iPhone recording and ffmpeg is installed.")
     except OSError as e:
         logger.error(f"OS error during audio conversion: {str(e)}")
-        raise ValueError("Failed to process audio file due to a system error.")
+        raise ValueError("Failed to process the .m4a file due to a system error. Check file permissions or disk space.")
     except Exception as e:
-        logger.error(f"Unexpected error during audio conversion: {str(e)}")
-        raise ValueError("An unexpected error occurred while converting the audio.")
+        logger.error(f"Unexpected error during .m4a conversion: {str(e)}")
+        raise ValueError(f"An unexpected error occurred while converting the .m4a file: {str(e)}")
 
 # Split audio into chunks
 def split_audio(audio_path: str) -> list:
     try:
         audio = AudioSegment.from_file(audio_path)
         if len(audio) == 0:
-            raise ValueError("Audio file is empty or invalid.")
+            raise ValueError("The .m4a file is empty or invalid.")
+        logger.info(f"Splitting audio into {CHUNK_DURATION_MS/1000}-second chunks: {audio_path}")
         return [audio[i:i + CHUNK_DURATION_MS] for i in range(0, len(audio), CHUNK_DURATION_MS)]
     except CouldntDecodeError:
         logger.error(f"Failed to decode audio for splitting: {audio_path}")
-        raise ValueError("Audio file is corrupted or in an unsupported format.")
+        raise ValueError("The .m4a file is corrupted or not supported. Ensure it's a valid iPhone recording.")
     except Exception as e:
         logger.error(f"Failed to split audio: {str(e)}")
-        raise ValueError(f"Failed to process audio: {str(e)}")
+        raise ValueError(f"Failed to process the .m4a file: {str(e)}")
 
 # Helper to compute chunk start time
 def get_chunk_time(index: int, chunk_duration_ms: int) -> str:
@@ -89,7 +104,7 @@ def transcribe(audio_path: str, include_timestamps: bool = False, progress=gr.Pr
     try:
         if not audio_path or not os.path.exists(audio_path):
             logger.warning("Invalid or missing audio file path.")
-            return "Please upload a valid audio file.", None
+            return "Please upload a valid .m4a file.", None
 
         # Convert to WAV if needed
         wav_path = convert_to_wav(audio_path)
@@ -110,7 +125,7 @@ def transcribe(audio_path: str, include_timestamps: bool = False, progress=gr.Pr
                     result = PIPELINE(temp_file.name, 
                                    generate_kwargs={"task": "transcribe", "language": "sv"})
                     text = result["text"].strip()
-                    if text:  # Only append non-empty transcriptions
+                    if text:
                         transcript.append(text)
                         if include_timestamps:
                             timestamp = get_chunk_time(i, CHUNK_DURATION_MS)
@@ -168,7 +183,7 @@ def transcribe(audio_path: str, include_timestamps: bool = False, progress=gr.Pr
         return str(e), None
     except Exception as e:
         logger.error(f"Unexpected error during transcription: {str(e)}")
-        return f"An unexpected error occurred: {str(e)}. Please try again or contact support.", None
+        return f"An unexpected error occurred while processing the .m4a file: {str(e)}. Please ensure the file is a valid iPhone recording and try again.", None
 
 # Initialize pipeline globally
 try:
@@ -181,11 +196,11 @@ except RuntimeError as e:
 def create_interface():
     with gr.Blocks(theme=gr.themes.Soft()) as demo:
         gr.Markdown("# Swedish Whisper Transcriber")
-        gr.Markdown("Upload audio (.wav, .mp3, .m4a) for real-time Swedish speech transcription.")
+        gr.Markdown("Upload an .m4a file from your iPhone for real-time Swedish speech transcription.")
         
         with gr.Row():
             with gr.Column():
-                audio_input = gr.Audio(type="filepath", label="Upload Audio")
+                audio_input = gr.Audio(type="filepath", label="Upload .m4a Audio")
                 timestamp_toggle = gr.Checkbox(label="Include Timestamps in Download", value=False)
                 transcribe_btn = gr.Button("Transcribe")
             
@@ -203,6 +218,9 @@ def create_interface():
 
 if __name__ == "__main__":
     try:
+        if not check_ffmpeg():
+            print("Error: ffmpeg is required to process .m4a files. Please install ffmpeg and ensure it's in your PATH.")
+            exit(1)
         create_interface().launch()
     except Exception as e:
         logger.critical(f"Failed to launch Gradio interface: {str(e)}")