File size: 4,994 Bytes
7b3c770 f0617ea 7b3c770 c95222f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 |
import gradio as gr
import subprocess
import whisper
from transformers import pipeline , T5ForConditionalGeneration, T5Tokenizer
import os
import torch
import spacy
# Load models once
whisper_model = whisper.load_model("base")
summarizer = pipeline("summarization", model="facebook/bart-large-cnn", device=-1)
# Load model and tokenizer
model_name = "valhalla/t5-base-qg-hl"
tokenizer = T5Tokenizer.from_pretrained(model_name)
model = T5ForConditionalGeneration.from_pretrained(model_name)
import spacy
try:
nlp = spacy.load("en_core_web_sm")
except OSError:
from spacy.cli import download
download("en_core_web_sm")
nlp = spacy.load("en_core_web_sm")
# Load QA pipeline
qa_pipeline = pipeline("question-answering", model="deepset/roberta-base-squad2")
def extract_audio(video_path, audio_output_path):
command = ['ffmpeg', '-i', video_path, '-vn', '-acodec', 'pcm_s16le', '-ar', '44100', '-ac', '2', audio_output_path]
subprocess.run(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
return audio_output_path
def process_video(video_file):
try:
import whisper
from transformers import pipeline
import subprocess
import os
audio_path = "extracted_audio.wav"
# Extract audio from video using FFmpeg
command = ['ffmpeg', '-i', video_file, '-vn', '-acodec', 'pcm_s16le', '-ar', '44100', '-ac', '2', audio_path]
subprocess.run(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
if not os.path.exists(audio_path):
return "Audio extraction failed.", "No summary generated."
# Load Whisper model
model = whisper.load_model("base")
result = model.transcribe(audio_path)
transcript_text = result['text']
# Load summarizer
summarizer = pipeline("summarization", model="facebook/bart-large-cnn", device=-1)
# Chunk text if needed
chunks = [transcript_text[i:i + 1024] for i in range(0, len(transcript_text), 1024)]
summaries = [summarizer(chunk, max_length=100, min_length=30, do_sample=False)[0]['summary_text'] for chunk in chunks]
final_summary = ' '.join(summaries)
return transcript_text, final_summary
except Exception as e:
return f"Error: {str(e)}", f"Error: {str(e)}"
# Extract top named entities for highlighting
def select_top_entities(text, max_entities=3):
doc = nlp(text)
candidates = [ent.text for ent in doc.ents if 2 <= len(ent.text) <= 30 and len(ent.text.split()) <= 5]
seen = set()
top_entities = []
for entity in candidates:
if entity not in seen:
seen.add(entity)
top_entities.append(entity)
if len(top_entities) >= max_entities:
break
return top_entities
# Generate questions for each highlighted entity
def generate_questions(context):
entities = select_top_entities(context, max_entities=3)
questions = []
for ent in entities:
highlighted = context.replace(ent, f"<hl> {ent} <hl>", 1)
input_text = f"generate question: {highlighted}"
input_ids = tokenizer.encode(input_text, return_tensors="pt", truncation=True)
outputs = model.generate(
input_ids=input_ids,
max_length=64,
num_beams=4,
num_return_sequences=1,
no_repeat_ngram_size=2,
early_stopping=True
)
question = tokenizer.decode(outputs[0], skip_special_tokens=True)
questions.append(question)
return "\n".join(f"Q{i+1}: {q}" for i, q in enumerate(questions))
def generate_answers(context, questions):
"""
context: str — typically the summary
questions: list[str] or str — can be multiline string or list
returns: str — formatted answers
"""
if isinstance(questions, str):
questions = questions.strip().split('\n')
answers = []
for q in questions:
if q.strip():
result = qa_pipeline(question=q.strip(), context=context)
answers.append(f"Q: {q.strip()}\nA: {result['answer']}")
return "\n\n".join(answers)
import gradio as gr
# Dummy processing functions — replace these with your actual logic
def process_video_(video_path):
# Step 1: Transcribe the video
transcript , summary = process_video(video_path)
questions = generate_questions(summary)
answers = generate_answers(summary, questions)
return transcript, summary, questions , answers
# Gradio Interface
iface = gr.Interface(
fn=process_video_,
inputs=gr.Video(label="Upload a video"),
outputs=[
gr.Textbox(label="Transcript"),
gr.Textbox(label="Summary"),
gr.Textbox(label="Generated Questions"),
gr.Textbox(label="Generated Answers")
],
title="Vision to Insight",
description="Upload a video to extract a transcript, generate a summary, and get 2–3 meaningful questions based on the summary."
)
iface.launch(share=True)
|