| from pathlib import Path |
| from typing import Dict, Iterable, List, Optional |
|
|
| from backend.types import RenderArtifact, RenderJob, VoiceConfig |
|
|
|
|
| class RenderPipeline: |
| def __init__(self, session_root: Path, synthesizer=None) -> None: |
| self.session_root = Path(session_root) |
| self.session_root.mkdir(parents=True, exist_ok=True) |
| self.synthesizer = synthesizer |
| self._jobs: Dict[str, RenderJob] = {} |
| self._cancelled: set[str] = set() |
| self._paused: set[str] = set() |
|
|
| def get_job(self, session_id: str) -> RenderJob: |
| return self._jobs.setdefault(session_id, RenderJob(session_id=session_id)) |
|
|
| def cancel(self, session_id: str) -> None: |
| self._cancelled.add(session_id) |
| self.get_job(session_id).status = "cancelled" |
|
|
| def pause(self, session_id: str) -> Dict[str, object]: |
| self._paused.add(session_id) |
| self.get_job(session_id).status = "paused" |
| return {"type": "paused", "session_id": session_id} |
|
|
| def resume(self, session_id: str) -> Dict[str, object]: |
| self._paused.discard(session_id) |
| self.get_job(session_id).status = "running" |
| return {"type": "resumed", "session_id": session_id} |
|
|
| def render( |
| self, |
| *, |
| session_id: str, |
| book: Dict[str, object], |
| chapters: List[Dict[str, object]], |
| voice_config: Dict[str, object], |
| diffusion_steps: int, |
| speed: float, |
| synthesizer=None, |
| ) -> Iterable[Dict[str, object]]: |
| selected = [chapter for chapter in chapters if chapter.get("included", True)] |
| job = self.get_job(session_id) |
| job.status = "running" |
| job.outputs.clear() |
| synth = synthesizer or self.synthesizer |
| if synth is None: |
| raise ValueError("No synthesizer provided for render pipeline") |
|
|
| yield { |
| "type": "started", |
| "session_id": session_id, |
| "total_chapters": len(selected), |
| "book_title": book.get("title"), |
| "backend": "local", |
| "model": VoiceConfig.from_dict(voice_config).model, |
| } |
|
|
| voice = VoiceConfig.from_dict(voice_config) |
| render_dir = self.session_root / session_id / "renders" |
| render_dir.mkdir(parents=True, exist_ok=True) |
|
|
| for index, chapter in enumerate(selected): |
| if session_id in self._cancelled: |
| yield {"type": "cancelled", "session_id": session_id, "backend": "local", "model": voice.model} |
| self._cancelled.discard(session_id) |
| job.status = "cancelled" |
| return |
|
|
| chapter_id = str(chapter["id"]) |
| job.current_chapter_id = chapter_id |
| yield { |
| "type": "chapter_started", |
| "session_id": session_id, |
| "chapter_id": chapter_id, |
| "chapter_title": chapter["title"], |
| "chapter_index": index, |
| "overall_progress": index / max(1, len(selected)), |
| "backend": "local", |
| "model": voice.model, |
| } |
|
|
| output_path = render_dir / f"{index + 1:03d}-{chapter_id}.wav" |
| result = synth.synthesize( |
| text=str(chapter["text"]), |
| output_path=output_path, |
| voice_config=voice, |
| diffusion_steps=diffusion_steps, |
| speed=speed, |
| ) |
|
|
| yield { |
| "type": "chapter_progress", |
| "session_id": session_id, |
| "chapter_id": chapter_id, |
| "percent": 100, |
| "backend": result.get("backend", "local"), |
| "model": result.get("model", voice.model), |
| } |
|
|
| artifact = RenderArtifact( |
| path=output_path, |
| duration_seconds=int(result.get("duration_seconds", 0)), |
| chapter_id=chapter_id, |
| ) |
| job.outputs.append(artifact) |
| yield { |
| "type": "chapter_done", |
| "session_id": session_id, |
| "chapter_id": chapter_id, |
| "duration_seconds": artifact.duration_seconds, |
| "overall_progress": (index + 1) / max(1, len(selected)), |
| "output_path": str(output_path), |
| "backend": result.get("backend", "local"), |
| "model": result.get("model", voice.model), |
| } |
|
|
| job.status = "completed" |
| yield { |
| "type": "completed", |
| "session_id": session_id, |
| "outputs": [str(artifact.path) for artifact in job.outputs], |
| "backend": "local", |
| "model": voice.model, |
| } |
|
|
