feat: Add support for uploading audio files to generate quizzes

README.md

@@ -22,6 +22,10 @@ Online learning has taken the front seat in the post-pandemic age. With the adve
 
 EduAid is one such project currently available in the form of a browser extension.
 
+## System Requirements
+
+- **ffmpeg** or **libav** installed and available on your system `PATH`. The backend's `extract_text_from_audio()` method (in `backend/Generator/main.py`) uses `pydub` to convert MP3 to WAV, and `pydub` depends on these tools for audio transcoding.
+
 ## Installation and Setup
 
 ### 1. Clone the Repository

backend/Generator/main.py

@@ -22,6 +22,9 @@
 import os
 import fitz 
 import mammoth
+import speech_recognition as sr
+from pydub import AudioSegment
+import uuid
 
 class MCQGenerator:
 
@@ -367,20 +370,70 @@ def extract_text_from_docx(self, file_path):
             result = mammoth.extract_raw_text(docx_file)
             return result.value
 
+    def extract_text_from_audio(self, file_path):
+        wav_path = file_path
+        request_id = uuid.uuid4().hex
+        if file_path.endswith('.mp3'):
+            audio = AudioSegment.from_file(file_path, format='mp3')
+            wav_path = os.path.join(self.upload_folder, f"{request_id}.wav")
+            audio.export(wav_path, format='wav')
+
+        r = sr.Recognizer()
+        audio = AudioSegment.from_wav(wav_path)
+
+        chunk_length_ms = 60 * 1000
+        chunks = [
+            audio[i: i + chunk_length_ms] for i in range(0, len(audio), chunk_length_ms)
+        ]
+
+        full_text = []
+        chunk_files = []
+
+        try:
+            for i, chunk in enumerate(chunks):
+                chunk_filename = os.path.join(self.upload_folder, f"{request_id}_chunk_{i}.wav")
+                chunk_files.append(chunk_filename)
+                chunk.export(chunk_filename, format='wav')
+
+                with sr.AudioFile(chunk_filename) as source:
+                    audio_data = r.record(source)
+                    try:
+                        text = r.recognize_google(audio_data)
+                        full_text.append(text)
+                    except sr.UnknownValueError:
+                        full_text.append("[Unintelligible]")
+                    except sr.RequestError as e:
+                        raise RuntimeError("Could not request results") from e
+        finally:
+            for chunk_file in chunk_files:
+                if os.path.exists(chunk_file):
+                    os.remove(chunk_file)
+            if wav_path != file_path and os.path.exists(wav_path):
+                os.remove(wav_path)
+
+        return "\n".join(full_text)
+
     def process_file(self, file):
         file_path = os.path.join(self.upload_folder, file.filename)
         file.save(file_path)
         content = ""
 
-        if file.filename.endswith('.txt'):
-            with open(file_path, 'r') as f:
-                content = f.read()
-        elif file.filename.endswith('.pdf'):
-            content = self.extract_text_from_pdf(file_path)
-        elif file.filename.endswith('.docx'):
-            content = self.extract_text_from_docx(file_path)
+        try:
+            if file.filename.endswith('.txt'):
+                with open(file_path, 'r') as f:
+                    content = f.read()
+            elif file.filename.endswith('.pdf'):
+                content = self.extract_text_from_pdf(file_path)
+            elif file.filename.endswith('.docx'):
+                content = self.extract_text_from_docx(file_path)
+            elif file.filename.endswith('.wav') or file.filename.endswith('.mp3'):
+                content = self.extract_text_from_audio(file_path)
+            else:
+                raise ValueError('Unsupported file format')
+        finally:
+            if os.path.exists(file_path):
+                os.remove(file_path)
 
-        os.remove(file_path)
         return content
 
 class QuestionGenerator:

eduaid_web/src/pages/Text_Input.jsx

@@ -185,7 +185,7 @@ const Text_Input = () => {
         {/* File Upload Section */}
         <div className="w-full max-w-2xl mx-auto border-[3px] rounded-2xl text-center px-6 py-6 border-dotted border-[#3E5063] mt-6">
           <img className="mx-auto mb-2" height={32} width={32} src={cloud} alt="cloud" />
-          <p className="text-white text-lg">Choose a file (PDF, MP3 supported)</p>
+          <p className="text-white text-lg">Choose a file (PDF, Audio supported)</p>
 
           <input type="file" ref={fileInputRef} onChange={handleFileUpload} style={{ display: "none" }} />
           <button

extension/src/pages/text_input/TextInput.jsx

@@ -234,7 +234,7 @@ function Second() {
           <img className="mx-auto" height={24} width={24} src={cloud} alt="cloud" />
           <div className="text-center text-white text-sm">Choose a file</div>
           <div className="text-center text-white text-sm">
-            PDF, MP3 supported
+            PDF, Audio supported
           </div>
           <div>
             <input

requirements.txt

@@ -31,5 +31,7 @@ tokenizers
 mammoth
 mediawikiapi
 PyMuPDF
-textblob
+textblob==0.19.0
+SpeechRecognition==3.14.5
+pydub==0.25.1
 llama-cpp-python