📚 The doc issue

tl;dr options parameter under torchvision.io.write_video is inadequately documented, challenging for end-users to tweak videos to their liking. same goes for audio_options

In torchvision.io.write_video, options and audio_options parameters are not sufficiently elaborated upon.

At first glance, this is okay as the function is ultimately just a wrapper around PyAV's video recording capabilities.

In practice, however, it is nontrivial to find relevant documentation to finetune the output.
This is especially so if the end-user, like me, is not familiar with PyAV or video recording in general.

Here is an example of the potential issues this may cause.

Suggest a potential alternative/fix

This is an issue we are working on as well in TorchRL.

As suggested by @vmoens, it would be nice if a link could be provided to the FFmpeg documentation, which is ultimately behind the options exposed by PyAV.

A potential extra step could then be to translate the wiki information into specific code examples that end-users can easily read and apply. This will probably benefit more people than if such information were only being documented in downstream libraries. Here's what a "specific code example" could look like,