digital_video_introduction VS Av1an

There's a great introduction to video tech, including codecs, at https://github.com/leandromoreira/digital_video_introduction

A couple of AV centric scenarios I'd explore: - Set up two speakers fed with a mono signal, flip the phase on one and have students move around the area. From this introduce constructive and destructive interference and some basic acoustics. - Using an oscilloscope (or just prepped graphics) show a sine wave on an unbalanced line, how this is paired with an inverted copy of itself on a balanced line, and how this works to cancel transmission noise. - With an analogue signal now pictured, introduce analogue to digital conversion and what sample rate and bit depth mean. From this you can also start to introduce why gain staging is important in both pure analogue systems and those where you're bouncing between analogue and digital domains. - Introduce video: colour channels, different ways of representing colour and some (very surface level) compression basics like chroma subsampling. Really good resource for that here. - Introduce base patterns for getting any signals arounds: dedicated circuits (analogue audio, HDMI video, RS-232 control, logic IO) and network based (again for audio, video, and control). Dive intro signal flow, how to read a schematic and where reasoning about each of these modes differs.

If you want to keep digging from there, get familiar with the compression on NVX, JPEG-XS, and how it behaves. intoPIX are the upstream vendor that do some of the NVX internals and have some great resources on that here: https://www.intopix.com/jpeg-xs. For a guide on all things digital video I also highly recommend https://github.com/leandromoreira/digital_video_introduction.

Edit: After finishing writing the comment, I realized that it was more technical than I intended, and assumes that the reader has a basic to medium understanding of video compression. If not, I recommend this writeup/tutorial of digital video, especially the section about frame types.

A good high-level breakdown of H.264: https://sidbala.com/h-264-is-magic/

Associated HN post (although there have been a few): https://news.ycombinator.com/item?id=30710574

More technical: https://github.com/leandromoreira/digital_video_introduction...

For those interested in this topic, I highly recommend the approachable but more extensive technical introduction at https://github.com/leandromoreira/digital_video_introduction

Go look at this. https://github.com/leandromoreira/digital_video_introduction

The hardware encoders are very fast and generally better than x264 (but not by as much as you'd think with the x264 slow preset).

In addition, there are threaded AV1 encoders you may be overlooking, like SVT-AV1. For non-realtime, my favorite is av1an, which also yields better quality than is possible from aomenc: https://github.com/master-of-zen/Av1an

I specifically used https://github.com/master-of-zen/Av1an with --target-quality 100 to keep the compression as close to lossless as possible (you could probably shave even more size off if you didn’t use near-lossless compression)

There's also Av1an as suggested by someone else. It's going to produce more efficient encodes because it uses chunk-based encoding to reach the required VMAF for every single chunk, but this is also slower because every chunk will be re-encoded until the VMAF target is reached. It's also a PITA to compile, has a long list of bugs, and isn't very actively developed. One of the long-standing issues is that frames have a tendency to go missing during splitting, which IME is going to keep happening when the input files aren't standardized. And thus the recommendation is to transcode the input to lossless first, which is just completely unsustainable.

i'm still trying to figure out if av1an is the breakthrough to using large CPU counts without losing quality. but it's a ridiculously poorly documented program. i haven't gotten it to run effectively yet. https://github.com/master-of-zen/Av1an

I don't know about VVC, but I've never seen a recommendation of preset 0 for SVT-AV1. The docs say "presets 1-3 represent extremely high efficiency, for use when encode time is not important and quality/size of the resulting video file is critical". It seems like 0 is the equivalent placebo preset. But if you want AV1 with coding efficiency over all else, IIRC the typical approach is to use av1an to detect scene transitions and farm out parallel encodes with aomenc.

That might be because it doesn't work.

Have you seen Av1an? It's supposed to encode to a target VMAF.

Are you using the current version, 0.4.0, from https://github.com/master-of-zen/Av1an/releases