Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Hacking YouTube with a MP4

realkeyboardwarrior.github.io

Post reply on HN
Print view View original

61–70 of 94 posts

Re: Hacking YouTube with a MP4

#61

Post by jchw »

dleslie wrote: post #57
deathanatos wrote: post #48

It's almost like we didn't learn from the days of MP3. I have several MP3s that, in certain players, are like a half hour long, despite being only 2 minutes long. My best guess was that they were assumed to be CBR, despite nothing about MP3 implying CBR… (there's not a flag or anything that says "this is a VBR" file, CBR files are just special…) Nowadays it's mostly moot since MP3 is obsolete.

> MP3 is obsolete What should we be using instead for lossy audio?

The current state of the art is Opus, but HE-AAC is also superior, and then there’s always the appeal of lossless which is a lot more practical than it once was.

Re: Hacking YouTube with a MP4

#63

Post by foepys »

ronsor wrote: post #58
LinuxBender wrote: post #16

Earlier quoted context omitted.

Not discord, but the default player is vulnerable to many different crash shenanigans. I get them sent to me all the time to look into and its usually just people using bogus timestamps, bogus seek times or concatenating multiple videos of different resolutions/rates that the player can't handle. If there was a way to get discord to spawn VLC for playing videos by default this would be less of a problem.

> get discord to spawn VLC So rather than loading the bogus videos in a sandboxed Chromium instance, you want to load them in an unsandboxed VLC instance? I smell eventual RCE.

Aren't quite a few Android security fixes every month related to the media framework? Are those not severe in a browser context because it's sandboxed?

Re: Hacking YouTube with a MP4

#64

Post by hulitu »

rozab wrote: post #39

I've seen many strange mp4s and webms floating around various discord communities. Some crash your client at a fitting moment in the video, some appear to be thousands of hours long, some appear to be seconds long but are actually hours long, some even loop! somehow.

This is the beauty of webm. You only need one HD frame to declare the stream HD. /s My child asks why we don't have nerflix or spotify. "Because it's crap, son". As Amazon says "up to ...". Up your's Amazon !

Re: Hacking YouTube with a MP4

#65

Post by rendall »

nightpool wrote: post #41
swyx wrote: post #33

i thought this as well explained. title a bit clickbaity, but it got me to click. i'm interested in learning more about the mp4 format. where can I read more? is there a canonical read that everyone but me knows about? OP seems like he has some kind of file explorer UI for it - also interested in that

"This is clickbait-y enough that I fell for it" is uh. Not exactly an endorsement? It seems like kind of the opposite of what you'd want to encourage?

I didn't like the title either. On the other hand, I doubt the fellow expected much of an audience but hit HN's front page. Also, the article is pretty great

Re: Hacking YouTube with a MP4

#66

Post by hulitu »

jchw wrote: post #61
dleslie wrote: post #57

Earlier quoted context omitted.

> MP3 is obsolete What should we be using instead for lossy audio?

The current state of the art is Opus, but HE-AAC is also superior, and then there’s always the appeal of lossless which is a lot more practical than it once was.

A format which cannot deliver quality is not state of the art. Opus is the Internet Explorer 6 of musical and video formats.

Re: Hacking YouTube with a MP4

#67

Post by oefrha »

Video timestamps are weird. Years ago I routinely pulled event VODs from an HLS source and re-uploaded to YouTube. To speed up downloading, I downloaded the MPEG-TS segments in parallel and assembled them with FFmpeg. Initially I used the basic and familiar concat demuxer during assembly. The results were fine locally. Months in a visitor told me that all my VODs had subtle yet frequent stutters. Turned out the videos played perfectly fine in any libavcodec-based (i.e. FFmpeg-based) video player, and still played fine even after libavcodec re-encoding, yet once they went through YouTube’s encoder, which AFIAK was also libavcodec-derived, subtle stutters appeared at segment boundaries. I then switched to the hls demuxer during assembly and the YouTube problem went away. I never got to the bottom of this, so to this day it’s still a mystery to me.

Re: Hacking YouTube with a MP4

#69

Post by jart »

rozab wrote: post #39

I've seen many strange mp4s and webms floating around various discord communities. Some crash your client at a fitting moment in the video, some appear to be thousands of hours long, some appear to be seconds long but are actually hours long, some even loop! somehow.

Do you still have copies of them? Could you send them to jtunney@gmail.com? I'd like to setup a web page hosting MPEG torture tests, since there doesn't appear to be one already. This is actually a very common practice for things like RFCs written for text-based protocols. We should ideally have more accessible information online that helps video software authors to harden their implementations against these sorts of busy beaver attacks.

Re: Hacking YouTube with a MP4

#70

Post by Thorrez »

hulitu wrote: post #66
jchw wrote: post #61

Earlier quoted context omitted.

The current state of the art is Opus, but HE-AAC is also superior, and then there’s always the appeal of lossless which is a lot more practical than it once was.

A format which cannot deliver quality is not state of the art. Opus is the Internet Explorer 6 of musical and video formats.

Opus is an audio format, not a video format. Opus is better than MP3. Wouldn't MP3 actually be the Internet Explorer 6 of audio formats?

https://sound.stackexchange.com/questions/26167/opus-vs-mp3-...

Post reply on HN