Updated with code

utils/audio_processing.py

@@ -0,0 +1,12 @@
+from moviepy.editor import AudioFileClip
+from pathlib import Path
+
+def extract_audio(video_path: Path):
+    """Extract audio from a video file."""
+    try:
+        audio = AudioFileClip(str(video_path))
+        audio_path = video_path.parent / f"{video_path.stem}_audio.wav"
+        audio.write_audiofile(str(audio_path), verbose=False, logger=None)
+        return audio_path
+    except Exception as e:
+        raise RuntimeError(f"Audio extraction failed: {e}")

utils/summarization.py

@@ -0,0 +1,8 @@
+from transformers import pipeline
+
+SUMMARY_MODEL = "Falconsai/text_summarization"
+
+def summarize_text(text):
+    """Summarize text using a Hugging Face pipeline."""
+    summarizer = pipeline("summarization", model=SUMMARY_MODEL)
+    return summarizer(text, max_length=150, min_length=30, do_sample=False)[0]["summary_text"]

utils/transcription.py

@@ -0,0 +1,60 @@
+import whisper
+from pathlib import Path
+from transformers import pipeline, AutoTokenizer
+
+WHISPER_MODEL = "base"
+SUMMARIZATION_MODEL = "t5-base"
+
+def transcribe_audio(audio_path: Path):
+    """Transcribe audio using Whisper."""
+    model = whisper.load_model(WHISPER_MODEL)
+    result = model.transcribe(str(audio_path))
+    transcript = result["text"]
+    summary = summarize_text(transcript)
+    return transcript, summary
+
+def summarize_text(text):
+    """Summarize text using a pre-trained T5 transformer model with chunking."""
+    summarization_pipeline = pipeline("summarization", model=SUMMARIZATION_MODEL)
+    tokenizer = AutoTokenizer.from_pretrained(SUMMARIZATION_MODEL)
+    
+    max_tokens = 512
+    
+    tokens = tokenizer(text, return_tensors='pt')
+    num_tokens = len(tokens['input_ids'][0])
+    
+    if num_tokens > max_tokens:
+        chunks = chunk_text(text, max_tokens)
+        summaries = []
+        for chunk in chunks:
+            summary_output = summarization_pipeline("summarize: " + chunk, max_length=150, min_length=30, do_sample=False)
+            summaries.append(summary_output[0]['summary_text'])
+        overall_summary = " ".join(summaries)
+    else:
+        overall_summary = summarization_pipeline("summarize: " + text, max_length=150, min_length=30, do_sample=False)[0]['summary_text']
+    
+    return overall_summary
+
+def chunk_text(text, max_tokens):
+    """Splits the text into a list of chunks based on token limits."""
+    tokenizer = AutoTokenizer.from_pretrained(SUMMARIZATION_MODEL)
+    words = text.split()
+    
+    chunks = []
+    current_chunk = []
+    current_length = 0
+    
+    for word in words:
+        hypothetical_length = current_length + len(tokenizer(word, return_tensors='pt')['input_ids'][0]) - 2
+        if hypothetical_length <= max_tokens:
+            current_chunk.append(word)
+            current_length = hypothetical_length
+        else:
+            chunks.append(' '.join(current_chunk))
+            current_chunk = [word]
+            current_length = len(tokenizer(word, return_tensors='pt')['input_ids'][0]) - 2
+    
+    if current_chunk:
+        chunks.append(' '.join(current_chunk))
+    
+    return chunks

utils/validation.py

@@ -0,0 +1,8 @@
+from pathlib import Path
+
+def validate_environment(obs_path: Path):
+    """Validate environment and prerequisites."""
+    errors = []
+    if not obs_path.exists():
+        errors.append(f"OBS directory not found: {obs_path}")
+    return errors