decent settings = bad quality :(

[erove]

Hi guys ,

i'm doing a gaming video and even though i use very good megui settings for encoding , i get terrible results

example of one test (http://erove2.wippiespace.com/test.mkv)

these settings i used , but result is always +- same
example1 (http://img705.imageshack.us/img705/8109/maxmegui.png)
example2 (http://www.upload.ee/files/293014/New_Megui_Settings.txt.html)
example3 (http://beta-gaming.de/et/ag0n/x264_dp_%20_st_scratchpad_st_.xml)

could anyone of you please help me ?

[poisondeathray]

Your sample has a lot of motion, and is difficult to encode. Use a higher bitrate, 5.8Mbps is not a lot for 720p50 with that type of source complexity

BTW, satd and 16 ref frames is overkill, the quality gains will be negligible , but encoding speed will be 5-10x slower. 3pass is pretty much useless as well.

[nurbs]

Can you post a sample from the source so we know what it is supposed to look like. The colours look a bit off, but maybe they are supposed to be like that and anyway that would be more of a PC/TV range or BT.601/709 issue. Impossible to tell without the source and the avisynth script you used. The final stats x264 spits out would also be interesting.

Apart from that video game footage isn't easily compressible in general because unlike film there is no motion blur. So for 720p50 at your bitrate you maybe can't get much better results.

Also you should probably stick to the presets and tunings and not set everything manually. Everything you set is very slow, but then you use --subme 9, --no-mbtree and don't use b-pyramid and potentially throw away a lot of quality (mainly mb-tree) for relatively little speed gain, especially compared to using --slow-firstpass and --me tesa.

[erove]

this is huffy example (http://img28.imageshack.us/img28/4448/huffy.jpg) , and well , i know more bitrate should be used but there are several people that has already created way more quality videos with these presets , probably could you recommend me some exact settings which i should change ? and btw the motion blur is just half amount what is usually used

[poisondeathray]

i know more bitrate should be used but there are several people that has already created way more quality videos with these presets

No they don't. Not with that source complexity.

If you had very little motion, with very little action & explosions (e.g. just camped out like a sniper) , then maybe with that bitrate, otherwise I seriously doubt it.

You also have some weird frame blends in your encoded sample. How did you do this capture? Did you feed it through an NLE?

[erove]

then probably i did something wrong in vegas or during capturing , but this first preset example also used a guy for 720p 50 fps vid who gained very decent quality with that setting

i dont know what is nle , but i was capturing it with image et tool , which records and also immediately adds motion blur into frames , merged them in virtual dub into huffy avi file and done some color corrections in vegas

[poisondeathray]

then probably i did something wrong in vegas or during capturing , but this first preset example also used a guy for 720p 50 fps vid who gained very decent quality with that setting

i dont know what is nle , but i was capturing it with image et tool , which records and also immediately adds motion blur into frames , merged them in virtual dub into huffy avi file and done some color corrections in vegas


But what was the source complexity for the other guy? The content matters. For example, if it's a still shot with low complexity you might be able to use 1Mb/s. If you're not fixated on bitrate or don't know what to use, try a crf encode. e.g. crf16-20

Your encoding settings are already very high, and most of them are absurd (diminishing returns) as discussed above. You won't get much better by adjusting them, except the bitrate.

NLE = non linear editor. e.g. vegas, premiere

Frame blending is the default action in most NLE's when your project/sequence settings DON'T match the actual footage. e.g you might have had a 60fps capture, but used 50fps project settings. It might have also been a product of your capture

If you're wondering what I'm talking about blends here is an example from your encode:
http://tinypic.com/r/i73gd0/6

What did you use to capture the original gameplay footage? e.g., FRAPS ? , Blackmagic intensity pro ? etc...

[erove]

i dont know what "source complexity" you mean, if the action in the video , then its equal or even higher in that video of the other guy

i was capturing it with image-et tool as i said , and i was capturing 150fps - because im planning to play a lot with velocity(the speed of video) and just 50 would result in terrible overall quality , dont know if that could cause those blends

sorry for my lack of informations about this encoding stuff , im pretty much a newbie to this :)

[poisondeathray]

i dont know what "source complexity" you mean, if the action in the video , then its equal or even higher in that video of the other guy


Yes , by source complexity, I mean things like action, scene changes, noise, all that requires more bitrate to encode for a certain level of quality. A static scene with no motion might only need 0.1Mb/s and look 10x better than your encode.

I'm sorry. I don't believe you. Please post an example from "that guy". If your claims are true, I am certain there will be differences ,like fps, dimensions, content.


i was capturing it with image-et tool as i said , and i was capturing 150fps - because im planning to play a lot with velocity(the speed of video) and just 50 would result in terrible overall quality , dont know if that could cause those blends


I'm not familiar with image-et tool , it might be screwing things up, or you might be getting frame drops @ 150fps (I suspect this is the case)

50fps should be easily obtained from 150fps, because it's evenly divisible. (there should be no blends or resampling)

In vegas, you can right click the clip and disable resample. When speed changes occur, it will duplicate frames instead of blending . This will give jerky playback. Both are bad IMO.

[erove]

I'm sorry. I don't believe you. Please post an example from "that guy". If your claims are true, I am certain there will be differences ,like fps, dimensions, content.


here is stream : http://www.own3d.tv/watch/21986 , but quality and colors are different then the actual file - here is link for download but i doubt you would like to do it , its over 500mb (http://myhannes.info/ccount11/click.php?id=48) so i cant show you anything else of that , but belive me , with that preset or very similar ones are working many people , tho many use just 30 fps


I'm not familiar with image-et tool , it might be screwing things up, or you might be getting frame drops @ 150fps

50fps should be easily obtained from 150fps, because it's evenly divisible. (there should be no blends or resampling)

In vegas, you can right click the clip and disable resample.

that tool is reliable in this case and yes , i did disable resampling , btw the blending was caused by luminance (i guess)

[poisondeathray]

... tho many use just 30 fps


I didn't check the sample or what fps he used yet, but that's a big difference, wouldn't you say...

The colors and luma values are an entirely different huge disussion


that tool is reliable in this case and yes , i did disable resampling , btw the blending was caused by luminance (i guess)

I doubt it.

If it was a straight encode and you had a CFR source (constant frame rate) at 150fps with no frame drops, and did no speed changes (ramps), and encoded a 50fps export, there should be no blends, because 50 is evenly divisible (just every 3rd frame). So either you screwed up the settings, the source had frame drops, or blends were in the source to begin with

[erove]

i mean the final clip is for example 33.33 fps , they capture 100 fps

ill try to get some changes in capturing and so , and post some results , so far thanks very much for your time

[poisondeathray]

I downloaded that video. He's using ~10.5Mbps. Of course it looks better.

Summary: Use a higher bitrate for 720p50 if that content is typical of what you're planning to encode

x264 is a great encoder, but it can only do so much at relatively low bitrates for that content complexity

Cheers