57 lines
1.9 KiB
Python
57 lines
1.9 KiB
Python
from pathlib import Path
|
|
from typing import Any, Dict
|
|
|
|
from app.backends.base import BaseVideoBackend
|
|
from app.utils.ffmpeg_utils import extract_first_frame, frames_to_video
|
|
from app.utils.files import TASK_FIRST_FRAME_NAME, TASK_VIDEO_NAME
|
|
from app.utils.image_utils import make_dummy_frame
|
|
|
|
|
|
class LTXBackend(BaseVideoBackend):
|
|
backend_name = "ltx_backend"
|
|
model_name = "LTX-Video"
|
|
|
|
def __init__(self, model_dir: Path):
|
|
self.model_dir = model_dir
|
|
self._loaded = False
|
|
self._pipeline = None
|
|
|
|
def load(self) -> None:
|
|
if self._loaded:
|
|
return
|
|
# TODO: Replace with real LTX loading, e.g. DiffusionPipeline.from_pretrained(...)
|
|
self.model_dir.mkdir(parents=True, exist_ok=True)
|
|
self._pipeline = "ltx_pipeline_placeholder"
|
|
self._loaded = True
|
|
|
|
def is_loaded(self) -> bool:
|
|
return self._loaded
|
|
|
|
def generate(self, task_id: str, request_data: Dict[str, Any], output_dir: str) -> Dict[str, str]:
|
|
self.load()
|
|
output = Path(output_dir)
|
|
frames_dir = output / "frames"
|
|
frames_dir.mkdir(parents=True, exist_ok=True)
|
|
|
|
duration = int(request_data["duration_sec"])
|
|
fps = int(request_data["fps"])
|
|
width = int(request_data["width"])
|
|
height = int(request_data["height"])
|
|
prompt = request_data["prompt"]
|
|
|
|
total_frames = duration * fps
|
|
for i in range(total_frames):
|
|
frame_path = frames_dir / f"frame_{i:04d}.jpg"
|
|
make_dummy_frame(frame_path, width, height, f"LTX preview | {prompt[:60]}", i)
|
|
|
|
video_path = output / TASK_VIDEO_NAME
|
|
frames_to_video(str(frames_dir / "frame_%04d.jpg"), fps, video_path)
|
|
|
|
first_frame_path = output / TASK_FIRST_FRAME_NAME
|
|
extract_first_frame(video_path, first_frame_path)
|
|
|
|
return {
|
|
"video_path": str(video_path.resolve()),
|
|
"first_frame_path": str(first_frame_path.resolve()),
|
|
}
|