Create app.py

app.py

@@ -0,0 +1,94 @@
+import os
+
+import uuid
+import yt_dlp as youtube_dl
+from typing import Generator
+from faster_whisper import WhisperModel
+import pandas as pd
+from typing import Generator
+from faster_whisper import WhisperModel
+import pandas as pd
+import gradio as gr
+
+class YouTubeTranscriber:
+    def __init__(self, model_path: str):
+        self.model = WhisperModel(model_path)
+        
+    def download_audio(self, url: str, preferred_quality: str = "192") -> str:
+        file_name = f"{uuid.uuid4()}.mp3"
+        output_path = os.path.join("/tmp", file_name)  # Use /tmp directory for temporary storage
+
+        ydl_opts = {
+            'format': 'bestaudio/best',
+            'postprocessors': [{
+                'key': 'FFmpegExtractAudio',
+                'preferredcodec': 'mp3',
+                'preferredquality': preferred_quality,
+            }],
+            'outtmpl': output_path,  # Specify the output path and file name
+        }
+
+        try:
+            with youtube_dl.YoutubeDL(ydl_opts) as ydl:
+                info_dict = ydl.extract_info(url, download=False)
+                video_title = info_dict.get('title', 'Unknown title')
+                print(f"Downloading audio for: {video_title}")
+
+                ydl.download([url])
+                print(f"Audio file saved as: {output_path}")
+
+            return output_path
+
+        except youtube_dl.utils.DownloadError as e:
+            print(f"Error downloading audio: {e}")
+            return None
+
+    def transcribe_audio(self, path: str) -> Generator:
+        print(f"Reading {path}")
+        segments, _ = self.model.transcribe(path)
+        return segments
+
+    def process_segments(self, segments: Generator) -> pd.DataFrame:
+        result = []
+        for i, segment in enumerate(segments):
+            result.append({
+                'chunk_id': f"chunk_{i}",
+                'chunk_length': segment.end - segment.start,
+                'text': segment.text,
+                'start_time': segment.start,
+                'end_time': segment.end
+            })
+
+        df = pd.DataFrame(result)
+        return df
+
+# Function to be called by the Gradio interface
+def transcribe_youtube_video(url: str, model_path: str = "distil-large-v2") -> str:
+    yt_transcriber = YouTubeTranscriber(model_path)
+    audio_path = yt_transcriber.download_audio(url)
+    
+    if audio_path:
+        segments = yt_transcriber.transcribe_audio(audio_path)
+        df = yt_transcriber.process_segments(segments)
+        output_csv = os.path.join("/tmp", f"{uuid.uuid4()}.csv")
+        df.to_csv(output_csv, index=False)
+        return output_csv
+    else:
+        return "Failed to download audio."
+
+
+import gradio as gr
+
+interface = gr.Interface(
+    fn=transcribe_youtube_video,
+    inputs=[
+        gr.Textbox(lines=1, placeholder="Enter YouTube URL here...", label="YouTube URL"),
+        gr.Textbox(lines=1, label="Whisper Model Path")
+    ],
+    outputs=gr.File(label="Transcribed Segments CSV"),  # Use gr.File directly
+    title="YouTube Audio Transcriber",
+    description="Enter a YouTube URL to download the audio and transcribe it using Whisper."
+)
+
+# Launch the interface
+interface.launch()