TMPGEnc Mastering Works - Very Slow vs Very Fast

[Anakin]

Hello guys

I am using 2 pass VBR for encoding H264 with an average of 8000. For years, I’ve always used the very slow preset. Bear in mind, my PC is old and slow. It averages around 15 hours for 1 hour.

So today, I used the very fast preset and the whole film was done in 12 hours. 2 hour film as well!

My question is, before I get carried aware, what do these presets actually do? The bitrate seems the same and I can’t tell the difference. I know the rule is “if you’re happy with it, that’s fine”, which is true, but I would like to know what it changes for a huge time difference.

Thanks

[LigH.de]

A video encoder knows several ways to improve a quality-per-bitrate ratio by taking more or less efforts to search for similarities in the video (redundancies) it is able to express in a shorter way, needing less bits for the same or a very similar result.

By an average of 8 Mbps, you seem to allow plenty of bitrate for SD resolution, possibly even enough for 720p HD material, so the quality loss is hardly noticeable for you even when the encoder doesn't take a lot of time and efforts to search for ways to need less bitrate to retain a good amount of quality.

[Anakin]

A video encoder knows several ways to improve a quality-per-bitrate ratio by taking more or less efforts to search for similarities in the video (redundancies) it is able to express in a shorter way, needing less bits for the same or a very similar result.

By an average of 8 Mbps, you seem to allow plenty of bitrate for SD resolution, possibly even enough for 720p HD material, so the quality loss is hardly noticeable for you even when the encoder doesn't take a lot of time and efforts to search for ways to need less bitrate to retain a good amount of quality.

Would you recommend 12?

[LigH.de]

12 ... what?!

[Anakin]

12 ... what?!

You said 8mbps is reasonable for SD and 720. Would 12mbps be ok for 1080?

[davexnet]

It's hard to assign an arbitrary bitrate, usually it's only done when you want to meet a specific target size.
In most cases, for most people, CRF encoding works better

[jagabo]

You said 8mbps is reasonable for SD and 720. Would 12mbps be ok for 1080?

It depends on the particular video, how you watch it, and your own personal expectations. As davexnet points out CRF encoding always give you the quality you specify with exactly the right bitrate to deliver that quality. And it's faster because it works in a single pass.

Regarding the slower presets: they work harder to find more ways to compress the video. And they use settings that give more accurate result. Some things to watch for: posterization in gradients (especially dark areas and blue skies), rough edges on moving objects. Those will be worse with fast presets.

[victoriabears]

I have been building my library of media since 2004, I have learnt 2 basic things,

1: you cannot make HD out of SD and all the many variants in that equation.

2: Create the best end result for YOU , experiment but not too much, it can drive you nuts, vary the settings a bit to suit the source. Focus on that you will want to watch the end result not create a PHd on the subject. In the hi-fi world it is said, be careful you do not end up listening to the equipment rather the music..

[rgr]

12 ... what?!

You said 8mbps is reasonable for SD and 720. Would 12mbps be ok for 1080?

Depends on who prefers. I compress with CRF 19 (ffmpeg) and the average bitrate is 25-35Mbps for 1080p. Two-pass encoding doesn't make much sense these days.

[sophisticles]

I have said this before and I will say it again, most speed/quality presets are snake oil and this applies to nearly every encoder.

In the cases where there is some "benefit", the cost far outweighs the benefit.

Case in point, I downloaded the touchdown pass 1080p clip from https://media.xiph.org/video/derf/ and using Handbrake on Gecko Rolling and created 2 encodes, both using CRF 25, the one with SF in the name was done with the super fast preset and no tune, the one with VS in the name was done with the very slow preset and tune film.

The super fast preset used a little more bit-rate but was 10x faster, encoding at over 80fps vs about 8dps for the very slow preset.

Also, the super fast preset did not max out my CPU whereas the very slow preset had the CPU pegged.

Attached are the final encodes.

Do yourself a favor, do a couple of test encodes with the super fast preset and see if you can tell the difference. If you can't, then save the CPU cycles and use that instead.