feat: add support for processing uploaded WAV files instead of downloading YouTube audio .

- Replaced song title input with a file upload interface to accept WAV files.
- Implemented function to process uploaded audio files.
- Updated the `process_audio` function to handle uploaded WAV files.
- Removed YouTube download functionality

app.py

@@ -23,42 +23,18 @@ def delete_input_files(input_dir):
         wav_file.unlink()
         print(f"Deleted {wav_file}")
 
-def download_youtube_audio_by_title(query, state=True):
+def process_uploaded_audio(file, state=True):
     if state:
         delete_input_files(INPUT_FOLDER)
 
-    ydl_opts = {
-        'quiet': True,
-        'default_search': 'ytsearch',
-        'noplaylist': True,
-        'format': 'bestaudio/best',
-        'outtmpl': './input/wav/%(title)s.%(ext)s',
-        'postprocessors': [{
-            'key': 'FFmpegExtractAudio',
-            'preferredcodec': 'wav',
-        }],
-        'cookiefile': './cookies.txt',
-    }
+    sanitized_filename = sanitize_filename(file.name)
+    input_path = Path(INPUT_FOLDER) / "wav" / sanitized_filename
+    input_path.parent.mkdir(parents=True, exist_ok=True)
 
-    with yt_dlp.YoutubeDL(ydl_opts) as ydl:
-        search_results = ydl.extract_info(query, download=False)
-        video_info = search_results['entries'][0]
-        video_url = video_info['webpage_url']
-        video_title = video_info['title']
-
-    match = re.match(r'^(.*? - .*?)(?: \[.*\]|\(.*\))?$', video_title)
-    formatted_title = match.group(1) if match else video_title
-
-    formatted_title = sanitize_filename(formatted_title.strip())
-
-    ydl_opts['outtmpl'] = f'./input/wav/{formatted_title}.%(ext)s'
-
-    if state:
-        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
-            ydl.download([video_url])
-        return f'./input/wav/{formatted_title}.wav'
-
-    return formatted_title
+    with open(input_path, 'wb') as f:
+        f.write(file.read())
+    
+    return str(input_path)
 
 def run_inference(model_type, config_path, start_check_point, input_dir, output_dir, device_ids="0"):
     command = [
@@ -114,8 +90,6 @@ def move_stems_to_parent(input_dir):
                 new_instrumental_path = os.path.join(parent_dir, "instrumental.wav")
                 print(f"Moving {instrumental_path} to {new_instrumental_path}")
                 shutil.move(instrumental_path, new_instrumental_path)
-            else:
-                print(f"Instrumental file not found: {instrumental_path}")
 
 def combine_stems_for_all(input_dir):
     for subdir, _, _ in os.walk(input_dir):
@@ -172,29 +146,31 @@ def delete_folders_and_files(input_dir):
 
     print("Cleanup completed.")
 
-def process_audio(song_title):
+def process_audio(uploaded_file):
     try:
-        yield "Finding audio...", None
-        if title_input == "":
-            raise ValueError("Please enter a song title.")
+        yield "Processing audio file...", None
         
-        formatted_title = download_youtube_audio_by_title(song_title, False)
+        if uploaded_file is None:
+            raise ValueError("Please upload a WAV file.")
+        
+        file_path = process_uploaded_audio(uploaded_file, False)
 
         yield "Starting SCNet inference...", None
         proc_folder_direct("scnet", "configs/config_scnet_other.yaml", "results/model_scnet_other.ckpt", f"{INPUT_FOLDER}/wav", OUTPUT_FOLDER)
 
         yield "Starting Mel Band Roformer inference...", None
         proc_folder_direct("mel_band_roformer", "configs/config_mel_band_roformer_vocals.yaml", "results/model_mel_band_roformer_vocals.ckpt", f"{INPUT_FOLDER}/wav", OUTPUT_FOLDER, extract_instrumental=True)
+        
         yield "Starting HTDemucs inference...", None
         proc_folder_direct("htdemucs", "configs/config_htdemucs_bass.yaml", "results/model_htdemucs_bass.th", f"{INPUT_FOLDER}/wav", OUTPUT_FOLDER)
 
-        source_path = f'{OUTPUT_FOLDER}{formatted_title}/mel_band_roformer/{formatted_title}_instrumental.wav'
-        destination_path = f'{OUTPUT_FOLDER}{formatted_title}/mel_band_roformer/{formatted_title}.wav'
+        source_path = f'{OUTPUT_FOLDER}/{file_path.stem}/mel_band_roformer/{file_path.stem}_instrumental.wav'
+        destination_path = f'{OUTPUT_FOLDER}/{file_path.stem}/mel_band_roformer/{file_path.stem}.wav'
     
         os.rename(source_path, destination_path)
         
         yield "Starting BS Roformer inference...", None
-        proc_folder_direct("bs_roformer", "configs/config_bs_roformer_instrumental.yaml", "results/model_bs_roformer_instrumental.ckpt", f'{OUTPUT_FOLDER}{formatted_title}/mel_band_roformer', OUTPUT_FOLDER)
+        proc_folder_direct("bs_roformer", "configs/config_bs_roformer_instrumental.yaml", "results/model_bs_roformer_instrumental.ckpt", f'{OUTPUT_FOLDER}/{file_path.stem}/mel_band_roformer', OUTPUT_FOLDER)
 
         yield "Moving input files...", None
         delete_input_files(INPUT_FOLDER)
@@ -208,8 +184,7 @@ def process_audio(song_title):
         yield "Cleaning up...", None
         delete_folders_and_files(OUTPUT_FOLDER)
 
-        
-        yield f"Audio processing completed successfully.", f'{OUTPUT_FOLDER}{formatted_title}/{formatted_title}.MDS.wav'
+        yield f"Audio processing completed successfully.", f'{OUTPUT_FOLDER}/{file_path.stem}/{file_path.stem}.MDS.wav'
     except Exception as e:
         error_msg = f"An error occurred: {str(e)}\n{traceback.format_exc()}"
         logging.error(error_msg)
@@ -219,26 +194,17 @@ with gr.Blocks() as demo:
     gr.Markdown("# Music Player and Processor")
 
     with gr.Row():
-        title_input = gr.Textbox(label="Enter Song Title")
-        play_button = gr.Button("Play")
-
-    audio_output = gr.Audio(label="Audio Player")
-
-    process_button = gr.Button("Process Audio")
+        file_input = gr.File(label="Upload WAV File", file_types=['wav'])
+        process_button = gr.Button("Process Audio")
 
     log_output = gr.Textbox(label="Processing Log", interactive=False)
     processed_audio_output = gr.Audio(label="Processed Audio")
 
-    play_button.click(
-        fn=download_youtube_audio_by_title,
-        inputs=title_input,
-        outputs=audio_output
-    )
-
     process_button.click(
         fn=process_audio,
-        inputs=title_input,
+        inputs=file_input,
         outputs=[log_output, processed_audio_output],
         show_progress=True
     )
+
 demo.launch()