feat: Celery setup and transcription task — Whisper + FFmpeg pipeline

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/app/services/transcription.py

@@ -0,0 +1,68 @@
+import os
+import subprocess
+import time
+import uuid
+
+from app.config import settings
+
+_model = None
+
+
+def _get_model():
+    global _model
+    if _model is None:
+        from faster_whisper import WhisperModel
+        _model = WhisperModel(
+            settings.whisper_model,
+            device=settings.whisper_device,
+            compute_type=settings.whisper_compute_type,
+        )
+    return _model
+
+
+def extract_audio(video_path: str) -> str:
+    audio_path = os.path.join(
+        settings.upload_dir, f"{uuid.uuid4().hex}.wav"
+    )
+    subprocess.run(
+        [
+            "ffmpeg", "-i", video_path,
+            "-vn", "-acodec", "pcm_s16le",
+            "-ar", "16000", "-ac", "1",
+            audio_path, "-y",
+        ],
+        check=True,
+        capture_output=True,
+    )
+    return audio_path
+
+
+def transcribe_audio(audio_path: str) -> dict:
+    model = _get_model()
+    start_time = time.time()
+
+    segments_iter, info = model.transcribe(
+        audio_path,
+        beam_size=5,
+        language=None,
+        vad_filter=True,
+    )
+
+    segments = []
+    full_text_parts = []
+    for segment in segments_iter:
+        segments.append({
+            "start": round(segment.start, 2),
+            "end": round(segment.end, 2),
+            "text": segment.text.strip(),
+        })
+        full_text_parts.append(segment.text.strip())
+
+    processing_time = time.time() - start_time
+
+    return {
+        "text": " ".join(full_text_parts),
+        "segments": segments,
+        "language": info.language,
+        "processing_time": round(processing_time, 2),
+    }