Puffer is a research project in the computer science department at Stanford University. Please find more details in the FAQ and our research paper (USENIX NSDI '20 Community Award, IRTF Applied …

I am currently located in the United States.Log in

If you believe your copyright has been violated on a Stanford site, please notify puffer-copyright-notices [at] cs.stanford.edu and give notice as stated under Reporting of Alleged Copyright Infringement.

Puffer is a Stanford University research study about using machine learning to improve video-streaming algorithms: the kind of algorithms used by services such as YouTube, Netflix, and Twitch.

At a high level, each day's Puffer data comprises different "measurements" — each measurement contains a different set of time-series data collected on Puffer servers, and is dumped as a CSV file.

Puffer encodes each video chunk in ten different H.264 versions, using libx264 in veryfast mode. The encodings range from 240p60 video with constant rate factor (CRF) of 26 (about 200 kbps) to …

Nov 12, 2020 · Performance of both versions of BOLA-BASIC on Puffer (95% confidence intervals), for all stream speeds (top plot) and slow streams only (bottom plot). Both plots show data from 2020-07 …

Selected date (UTC): 2025-11-28T11_2025-11-29T11 (share a permalink) Full results: Storage bucket Retrained model: 20251128-1.tar.gz video_sent measurement: video ...