Update app.py

app.py

@@ -7,7 +7,7 @@ import uuid
 from googletrans import Translator
 from TTS.api import TTS
 import ffmpeg
-from faster_whisper import WhisperModel
+import json
 from scipy.signal import wiener
 import soundfile as sf
 from pydub import AudioSegment
@@ -21,6 +21,7 @@ import torchvision
 from tqdm import tqdm
 from numba import jit
 from huggingface_hub import HfApi
+import moviepy.editor as mp
 
 HF_TOKEN = os.environ.get("HF_TOKEN")
 os.environ["COQUI_TOS_AGREED"] = "1"
@@ -30,9 +31,16 @@ ZipFile("ffmpeg.zip").extractall()
 st = os.stat('ffmpeg')
 os.chmod('ffmpeg', st.st_mode | stat.S_IEXEC)
 
-#Whisper
-model_size = "small"
-model = WhisperModel(model_size, device="cpu", compute_type="int8")
+print("Starting the program...")
+
+def generate_unique_filename(extension):
+    return f"{uuid.uuid4()}{extension}"
+
+def cleanup_files(*files):
+    for file in files:
+        if file and os.path.exists(file):
+            os.remove(file)
+            print(f"Removed file: {file}")
 
 def check_for_faces(video_path):
     face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
@@ -51,6 +59,68 @@ def check_for_faces(video_path):
 
     return False
 
+@spaces.GPU(duration=90)
+def transcribe_audio(file_path):
+    print(f"Starting transcription of file: {file_path}")
+    temp_audio = None
+    if file_path.endswith(('.mp4', '.avi', '.mov', '.flv')):
+        print("Video file detected. Extracting audio...")
+        try:
+            video = mp.VideoFileClip(file_path)
+            temp_audio = generate_unique_filename(".wav")
+            video.audio.write_audiofile(temp_audio)
+            file_path = temp_audio
+        except Exception as e:
+            print(f"Error extracting audio from video: {e}")
+            raise
+    
+    print(f"Does the file exist? {os.path.exists(file_path)}")
+    print(f"File size: {os.path.getsize(file_path) if os.path.exists(file_path) else 'N/A'} bytes")
+    
+    output_file = generate_unique_filename(".json")
+    command = [
+        "insanely-fast-whisper",
+        "--file-name", file_path,
+        "--device-id", "0",
+        "--model-name", "openai/whisper-large-v3",
+        "--task", "transcribe",
+        "--timestamp", "chunk",
+        "--transcript-path", output_file
+    ]
+    print(f"Executing command: {' '.join(command)}")
+    try:
+        result = subprocess.run(command, check=True, capture_output=True, text=True)
+        print(f"Standard output: {result.stdout}")
+        print(f"Error output: {result.stderr}")
+    except subprocess.CalledProcessError as e:
+        print(f"Error running insanely-fast-whisper: {e}")
+        print(f"Standard output: {e.stdout}")
+        print(f"Error output: {e.stderr}")
+        raise
+    
+    print(f"Reading transcription file: {output_file}")
+    try:
+        with open(output_file, "r") as f:
+            transcription = json.load(f)
+    except json.JSONDecodeError as e:
+        print(f"Error decoding JSON: {e}")
+        print(f"File content: {open(output_file, 'r').read()}")
+        raise
+    
+    if "text" in transcription:
+        result = transcription["text"]
+    else:
+        result = " ".join([chunk["text"] for chunk in transcription.get("chunks", [])])
+    
+    print("Transcription completed.")
+    
+    # Cleanup
+    cleanup_files(output_file)
+    if temp_audio:
+        cleanup_files(temp_audio)
+    
+    return result
+
 @spaces.GPU
 def process_video(radio, video, target_language, has_closeup_face):
     try:
@@ -80,20 +150,16 @@ def process_video(radio, video, target_language, has_closeup_face):
         
         print("Attempting to transcribe with Whisper...")
         try:
-            segments, info = model.transcribe(f"{run_uuid}_output_audio_final.wav", beam_size=5)
-            whisper_text = " ".join(segment.text for segment in segments)
-            whisper_language = info.language
+            whisper_text = transcribe_audio(f"{run_uuid}_output_audio_final.wav")
             print(f"Transcription successful: {whisper_text}")
-        except RuntimeError as e:
-            print(f"RuntimeError encountered: {str(e)}")
-            if "CUDA failed with error device-side assert triggered" in str(e):
-                gr.Warning("Error. Space need to restart. Please retry in a minute")
-                api.restart_space(repo_id=repo_id)
+        except Exception as e:
+            print(f"Error encountered during transcription: {str(e)}")
+            raise
                 
         language_mapping = {'English': 'en', 'Spanish': 'es', 'French': 'fr', 'German': 'de', 'Italian': 'it', 'Portuguese': 'pt', 'Polish': 'pl', 'Turkish': 'tr', 'Russian': 'ru', 'Dutch': 'nl', 'Czech': 'cs', 'Arabic': 'ar', 'Chinese (Simplified)': 'zh-cn'}
         target_language_code = language_mapping[target_language]
         translator = Translator()
-        translated_text = translator.translate(whisper_text, src=whisper_language, dest=target_language_code).text
+        translated_text = translator.translate(whisper_text, dest=target_language_code).text
         print(translated_text)
 
         tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2")
@@ -173,8 +239,8 @@ iface = gr.Interface(
     title="AI Video Dubbing",
     description="""This tool was developed by [@artificialguybr](https://twitter.com/artificialguybr) using entirely open-source tools. Special thanks to Hugging Face for the GPU support. Thanks [@yeswondwer](https://twitter.com/@yeswondwerr) for original code. Test the [Video Transcription and Translate](https://huggingface.co/spaces/artificialguybr/VIDEO-TRANSLATION-TRANSCRIPTION) space!""",
     allow_flagging=False
-
 )
+
 with gr.Blocks() as demo:
     iface.render()
     radio.change(swap, inputs=[radio], outputs=video)
@@ -188,5 +254,7 @@ with gr.Blocks() as demo:
     - If you need more than 1 minute, duplicate the Space and change the limit on app.py.
     - If you incorrectly mark the 'Video has a close-up face' checkbox, the dubbing may not work as expected.
     """)
+
+print("Launching Gradio interface...")
 demo.queue()
 demo.launch()