Introducing Vireo: A Lightweight and Versatile Video Processing Library | Hacker News Reader

Introducing Vireo: A Lightweight and Versatile Video Processing Library

blog.twitter.com

61 pointsf2n8 comments

Here's the Git log on the GitHub repo: https://github.com/twitter/vireo/commits/master . As of this writing, it consists of a single commit with the commit message "first commit."

I'll give the devs here the benefit of the doubt that the original project repo has more history than that, but what might be the reason it wasn't preserved when the GitHub repo was created? Copyright reasons? Or someone just couldn't be bothered?

The project is relatively mature at Twitter and has indeed a much longer commit history. Per our internal policies, the repo had to be cleaned before we could release it to public, and thus we erased the history before putting it on github. An exciting fresh new page in the project history.

Okay, I figured this was probably a policy thing, but why?

Not trying to be critical here. Just curious.

It's mostly related to properly complying with OSS licenses of other projects and we also had some pieces of code that were for Twitter specific use cases.

Commit messages are often leaky about IP and internal processes/personnel.

So a lightweight version of ffmpeg/mencoder, but what is meant by "lightweight" in terms of stream video processing? Does it perform conversions from one codec to another using less CPU cycles.

Lightweight in terms of new code on top of existing codecs: we kept it compact, meaning small binary, fast execution, and easier to add to a mobile app for instance. (ffmpeg gets big very quickly). Also, there aren't any extra memory management system, buffer pools etc: it's all done through C++11. We basically thinly rearchitectured around the media technologies used today in 99% of cases: mp4, mpeg ts, jpg, GIFs, VPX/WebM, WebP, ...

It still uses x264, vpx etc to do decode & encode. It's lightweight in the sense that it's a shallow wrapper around these libraries and it doesn't consume unnecessary memory or CPU cycles on it's own.

Reply on news.ycombinator.com