google-labs-jules[bot]
7cf1b45545
- Implements `process_batched_images` generator in `nodes/video_node.py` to process video frames in batches (default 20), significantly reducing GPU-CPU synchronization overhead. - Optimizes `tensor_to_numpy_uint8` in `shared/media/image_processing.py` to use in-place operations (`.clamp_()`), saving memory allocations for large tensors. - Reduces performance bottlenecks in video encoding pipelines.