Initial commit: Børnetime app scaffold

FastAPI + Celery + Svelte app for English→Danish video translation
with voice cloning. Includes diarization, STT, translation, TTS,
and audio mixing pipeline.

backend/app/tasks/process.py

@@ -0,0 +1,144 @@
+import os
+import logging
+from sqlalchemy import create_engine
+from sqlalchemy.orm import Session
+
+from .celery_app import celery_app
+from ..config import settings
+from ..models.tables import Video, Segment, Job, VideoStatus, JobStatus
+from ..services.diarization import diarize, unload_model as unload_diarization
+from ..services.stt import transcribe_segment, unload_model as unload_stt
+from ..services.translator import translate, unload_model as unload_translate
+from ..services.video_utils import extract_audio
+
+logger = logging.getLogger(__name__)
+
+engine = create_engine(settings.database_url.replace("+aiosqlite", ""))
+
+
+def _get_sync_session():
+    return Session(engine)
+
+
+@celery_app.task(bind=True, name="process_video")
+def process_video(self, video_id: int, job_id: int = None):
+    db = _get_sync_session()
+    temp_files = []
+    try:
+        video = db.query(Video).filter(Video.id == video_id).first()
+        if not video:
+            raise ValueError(f"Video {video_id} not found")
+
+        video.status = VideoStatus.processing.value
+        db.commit()
+
+        if job_id:
+            job = db.query(Job).filter(Job.id == job_id).first()
+        else:
+            job = Job(
+                video_id=video_id, job_type="process", status=JobStatus.running.value
+            )
+            db.add(job)
+        db.commit()
+
+        self.update_state(state="PROGRESS", meta={"progress": 0.0})
+        job.celery_task_id = self.request.id
+        job.status = JobStatus.running.value
+        db.commit()
+
+        audio_path = os.path.join(settings.temp_dir, f"audio_{video_id}.wav")
+        temp_files.append(audio_path)
+        extract_audio(video.filepath, audio_path)
+        logger.info("Extracted audio for video %d", video_id)
+
+        self.update_state(state="PROGRESS", meta={"progress": 0.10})
+        job.progress = 0.10
+        db.commit()
+
+        segments = diarize(audio_path)
+        if not segments:
+            raise RuntimeError("No speech segments detected in video")
+        logger.info("Diarized %d segments for video %d", len(segments), video_id)
+
+        unload_diarization()
+
+        self.update_state(state="PROGRESS", meta={"progress": 0.20})
+        job.progress = 0.20
+        db.commit()
+
+        total = len(segments)
+
+        logger.info("Starting STT for %d segments", total)
+        for i, seg in enumerate(segments):
+            en_text = transcribe_segment(audio_path, seg["start_time"], seg["end_time"])
+            seg["english_text"] = en_text
+            pct = 0.20 + (i + 1) / total * 0.35
+            self.update_state(state="PROGRESS", meta={"progress": pct})
+            job.progress = pct
+            db.commit()
+
+        unload_stt()
+        logger.info("STT complete, unloaded model")
+
+        logger.info("Starting translation for %d segments", total)
+        for i, seg in enumerate(segments):
+            en_text = seg.get("english_text", "")
+            da_text = translate(en_text) if en_text else ""
+            seg["danish_text"] = da_text
+
+            pct = 0.55 + (i + 1) / total * 0.35
+            self.update_state(state="PROGRESS", meta={"progress": pct})
+            job.progress = pct
+            db.commit()
+
+        unload_translate()
+        logger.info("Translation complete, unloaded model")
+
+        for seg in segments:
+            db_seg = Segment(
+                video_id=video_id,
+                speaker_label=seg["speaker_label"],
+                start_time=seg["start_time"],
+                end_time=seg["end_time"],
+                english_text=seg.get("english_text", ""),
+                danish_text=seg.get("danish_text", ""),
+                status="done",
+            )
+            db.add(db_seg)
+
+        video.status = VideoStatus.review.value
+        job.status = JobStatus.done.value
+        job.progress = 1.0
+        db.commit()
+
+        logger.info("Processing complete for video %d", video_id)
+        return {"video_id": video_id, "status": "review"}
+
+    except Exception as e:
+        logger.error("Processing failed for video %d: %s", video_id, e, exc_info=True)
+        video = db.query(Video).filter(Video.id == video_id).first()
+        if video:
+            video.status = VideoStatus.error.value
+            video.error_message = str(e)
+        if job_id:
+            job = db.query(Job).filter(Job.id == job_id).first()
+        else:
+            job = (
+                db.query(Job)
+                .filter(Job.video_id == video_id, Job.job_type == "process")
+                .first()
+            )
+        if job:
+            job.status = JobStatus.failed.value
+            job.error_message = str(e)
+        db.commit()
+        raise
+
+    finally:
+        for f in temp_files:
+            if os.path.exists(f):
+                try:
+                    os.remove(f)
+                except OSError:
+                    pass
+        db.close()