quran-srt-generator / src /audio_processor.py
aladhefafalquran
Initial commit
df2fada
"""
Audio Processor
Extracts audio from video and transcribes using Whisper
Optimized for Arabic Quran recitation
"""
import os
import subprocess
import tempfile
from pathlib import Path
from typing import Optional, List, Dict
import json
# Try to import whisper
try:
import whisper
WHISPER_AVAILABLE = True
except ImportError:
WHISPER_AVAILABLE = False
print("Warning: Whisper not installed. Install with: pip install openai-whisper")
class AudioProcessor:
def __init__(self, model_size: str = "medium"):
"""
Initialize the audio processor
Args:
model_size: Whisper model size
- "tiny": Fastest, least accurate
- "base": Fast, basic accuracy
- "small": Good balance
- "medium": Recommended for Arabic (best balance)
- "large": Most accurate, slowest (requires more VRAM)
"""
self.model_size = model_size
self.model = None
self.temp_dir = Path(tempfile.gettempdir()) / "quran_srt"
self.temp_dir.mkdir(exist_ok=True)
def load_model(self):
"""Load Whisper model (lazy loading)"""
if not WHISPER_AVAILABLE:
raise RuntimeError("Whisper is not installed. Run: pip install openai-whisper")
if self.model is None:
print(f"Loading Whisper {self.model_size} model...")
self.model = whisper.load_model(self.model_size)
print("Model loaded successfully!")
return self.model
def extract_audio(self, video_path: str, output_path: Optional[str] = None) -> str:
"""
Extract audio from video file using FFmpeg
Args:
video_path: Path to the video file
output_path: Optional output path for audio file
Returns:
Path to the extracted audio file
"""
video_path = Path(video_path)
if not video_path.exists():
raise FileNotFoundError(f"Video file not found: {video_path}")
if output_path is None:
output_path = self.temp_dir / f"{video_path.stem}_audio.wav"
else:
output_path = Path(output_path)
# FFmpeg command to extract audio as WAV (16kHz for Whisper)
cmd = [
"ffmpeg",
"-i", str(video_path),
"-vn", # No video
"-acodec", "pcm_s16le", # PCM format
"-ar", "16000", # 16kHz sample rate (Whisper optimal)
"-ac", "1", # Mono
"-y", # Overwrite output
str(output_path)
]
try:
result = subprocess.run(
cmd,
capture_output=True,
text=True,
check=True
)
print(f"Audio extracted to: {output_path}")
return str(output_path)
except subprocess.CalledProcessError as e:
raise RuntimeError(f"FFmpeg error: {e.stderr}")
except FileNotFoundError:
raise RuntimeError("FFmpeg not found. Please install FFmpeg.")
def transcribe(
self,
audio_path: str,
language: str = "ar",
task: str = "transcribe"
) -> Dict:
"""
Transcribe audio using Whisper
Args:
audio_path: Path to audio file
language: Language code ("ar" for Arabic)
task: "transcribe" for same language, "translate" for English
Returns:
Transcription result with segments and timestamps
"""
model = self.load_model()
print(f"Transcribing audio: {audio_path}")
print("This may take a few minutes depending on the video length...")
result = model.transcribe(
audio_path,
language=language,
task=task,
word_timestamps=True, # Get word-level timestamps
verbose=False,
initial_prompt="ุจุณู… ุงู„ู„ู‡ ุงู„ุฑุญู…ู† ุงู„ุฑุญูŠู…", # Help with Quran context
)
return result
def transcribe_video(
self,
video_path: str,
language: str = "ar"
) -> Dict:
"""
Full pipeline: extract audio and transcribe
Args:
video_path: Path to video file
language: Language code
Returns:
Transcription result with segments
"""
# Extract audio
audio_path = self.extract_audio(video_path)
# Transcribe
result = self.transcribe(audio_path, language=language)
# Clean up temp audio file
try:
os.remove(audio_path)
except:
pass
return result
def get_segments_with_timing(self, transcription: Dict) -> List[Dict]:
"""
Extract segments with precise timing from transcription
Args:
transcription: Whisper transcription result
Returns:
List of segments with start, end, and text
"""
segments = []
for segment in transcription.get("segments", []):
segments.append({
"id": segment.get("id", len(segments)),
"start": segment.get("start", 0),
"end": segment.get("end", 0),
"text": segment.get("text", "").strip(),
"words": segment.get("words", []),
"confidence": segment.get("avg_logprob", 0)
})
return segments
class MockAudioProcessor:
"""
Mock processor for testing without Whisper installed
"""
def __init__(self, model_size: str = "medium"):
self.model_size = model_size
def transcribe_video(self, video_path: str, language: str = "ar") -> Dict:
"""Return mock transcription for testing"""
return {
"text": "ุจุณู… ุงู„ู„ู‡ ุงู„ุฑุญู…ู† ุงู„ุฑุญูŠู… ุงู„ุญู…ุฏ ู„ู„ู‡ ุฑุจ ุงู„ุนุงู„ู…ูŠู†",
"segments": [
{
"id": 0,
"start": 0.0,
"end": 3.5,
"text": "ุจุณู… ุงู„ู„ู‡ ุงู„ุฑุญู…ู† ุงู„ุฑุญูŠู…",
"words": []
},
{
"id": 1,
"start": 3.5,
"end": 6.0,
"text": "ุงู„ุญู…ุฏ ู„ู„ู‡ ุฑุจ ุงู„ุนุงู„ู…ูŠู†",
"words": []
}
],
"language": "ar"
}
def get_segments_with_timing(self, transcription: Dict) -> List[Dict]:
return transcription.get("segments", [])
def get_processor(model_size: str = "medium") -> AudioProcessor:
"""
Get appropriate processor based on Whisper availability
"""
if WHISPER_AVAILABLE:
return AudioProcessor(model_size)
else:
print("Using mock processor (Whisper not installed)")
return MockAudioProcessor(model_size)
# For testing
if __name__ == "__main__":
processor = get_processor()
print(f"Whisper available: {WHISPER_AVAILABLE}")
print(f"Processor type: {type(processor).__name__}")
# Test with mock data
if not WHISPER_AVAILABLE:
result = processor.transcribe_video("test.mp4")
segments = processor.get_segments_with_timing(result)
print("\nMock transcription result:")
for seg in segments:
print(f"[{seg['start']:.2f} - {seg['end']:.2f}] {seg['text']}")