Glossary

pts

Presentation Time Stamp. The time at which a frame should be played. In TorchCodec, pts are expressed in seconds.

best stream

The notion of “best” stream is determined by FFmpeg. Quoting the FFmpeg docs:

The best stream is determined according to various heuristics as the most likely to be what the user expects.

scan

A scan corresponds to an entire pass over a video file, with the purpose of retrieving metadata about the different streams and frames. It does not involve decoding, so it is a lot cheaper than decoding the file.

clips

A clip is a sequence of frames, usually in pts order. The frames may not necessarily be consecutive. A clip is represented as a 4D FrameBatch. A group of clips, which is what the samplers return, is represented as 5D FrameBatch.