Skip to content

Instantly share code, notes, and snippets.

View kodawah's full-sized avatar

Vittorio Giovara kodawah

76 followers · 3 following
View GitHub Profile
@kodawah
kodawah / gist:9389e55ef839fc7aa12fc8330a14bb92
Last active November 17, 2016 17:47
/**
* @name Initial orientation
* @{
* There fields describe additional rotations applied to the sphere after
* the video frame is mapped onto it. The sphere is rotated around the
* viewer, who remains stationary.
*
* A positive yaw moves the left half of the frame forward.
* A positive pitch moves the bottom half of the frame upward.
* A positive roll moves the left half of the frame downward.
@kodawah
kodawah / notes on opus in mp4 in ffmpeg
Created September 26, 2016 21:11
[6:04pm] Paranoialmaniac: https://github.com/kinetiknz/FFmpeg/commit/6f874bd6347c8938b4d0f2b483ed037f7e9e291a#diff-44e5be1f9a8e0df35dd607e74773ce6eR1654
[6:04pm] Paranoialmaniac: opus can have variable pre-roll size
[6:04pm] Paranoialmaniac: since opus can vary its frame-size by frame by frame
[6:10pm] Paranoialmaniac: he misunderstands units of roll_distance
[6:13pm] Paranoialmaniac: roll_distance is in units of Opus samples not PCM samples
@kodawah
kodawah / multimedia_state_of_the_art.md
Last active December 22, 2018 04:45

Introduction

This table shows the current implementation status of several multimedia fomats by popular tools and browsers. All tests were run with the latest available version, or git HEAD for open source software.

Symbols mean

  • ✅ fully supported
  • ❗ unsupported
  • 💭 work in progress
  • 1️⃣️ see notes
@kodawah
kodawah / x264_preset_benchmark_2015.md
Last active May 22, 2017 17:55

##introduction

This benchmark evaluates the different presets options offered by x264, using medium as reference, and providing size and time differences (relative and not).

##benchmark Tests run with latest x264 (12139) and FFmpeg, configured to run with threading enabled. The slower and veryslow presets have been modified to keep the number of refs constant and compatible with most decoders. All results represent an average over 10 runs computed by perf.