x264 Encoding Issues (Needing Better Software - Suggestions Please)

[Nexin]

Have had a problem with H264 encoding since x264 has been available to public.

The problem been that no profile is any good and neither so far any settings people have posted been good. The video always seems blurred or more so on writing car number plates or road signs not visible correctly.

I have tried many times using virtualdub with x264vfw but have tried a few others very small encoders all which have been the same very poor.

This is with scene 720 or 1080 i/p reducing size to 720x404 or 704x396. Even tried with all profiles and highest other settings 4:4:4 and 6.2 with any CRF or CQP at 1 or ABR at 10000. Extreme settings but still nothing has made any videos crisp. This is also with using intermediate lags, huffy or magic (uncompressed) avi.

End result always seems blurred or otherwise need to double take vision (readjust my vision) for each scene change. The source are crisp and nice not the encodes I do.

So looking for another solution that is also a so easy to change settings. Though mainly those would be to change resolution size and sometimes frame rate.

Not looking for any huge bloated GUI something small that can do resize and frame rate but mainly able to encode to encoders best ability for crisp video each time. Prefer to use video source than have need for intermediate. Simple minimal small ui is all that I need if it can show the settings used the better so get used to correct x264 settings been used so able to continue with those with other software's if one GUI stops development etc.

x264 Preset I would prefer to use would be 1 pass Veryfast with CRF 19 or 20 since that I have read people get very good results with and the encode time is then far less.

Need UI's only that are proven and able to encode x264 avc1 videos so they have crisp vision videos each time.

Crisp videos all pixels in detail present that are able to be seen easily and instantly for each video frame. All writing car number plates and road signs instantly readable all times all other details to visible. But not produce blurred videos that need constant eye readjustments to watch the videos.

:thanks:

[Sparktank]

Veryfast


If you want better looking results, ie more transparent to the source, then you would have to use settings that resemble closer to VerySlow/Placebo.

Which would mean you have to take into account encoding will be slower than you want it to be.

That's pretty much your only option right now.
If you don't want slower settings, you get what you get.

[Nexin]

I've tried many x264 settings such as Extreme max x264 can go as below using latest x264vfw or some previous versions with virtualdub and other software's. Tested also with x264gui and x264cli.

x264 settings including Max setting I have tested with.

Preset veryslow to veryfast / all tuning modes Film to SSIM / Profile High to High 4:4:4 / Level 3.1 to 6.2

The above tested with
CRF various 1 to 19
CQP various 1 to 19
ABR various 1000 to 10000 kbits/s


Max Settings
Preset Veryslow / Tuning All tested / Profile High and High 4:4:4 / Level 4.1 and 6.2
CRF 1
CQP 1
ABR 10000 kbits/s


Even some suggestions I have from other people posts on the net.

I know it is the settings at fault but what is best when reduce resolution size to maintain crisp crystal clear video pixels.

If I encode the source video the output is near identical as the source crystal clear. Only when I reduce the video to 720x404 or 704x396 (latter mainly tested with has 1.778 AR). But never yet output a video file with clear crystal pixel vision as the source has.

Adjust many setting even others I have read about AQ, deblock, subme, partitions etc etc. With the above setting all at max even the default profiles should produce crystal pixel clear videos.

The problem I see is mainly with road signs and car number plates but evident on many things in the video. The video also seem to be hazy (like sun haze or light mist) besides not having crystal clear pixel vision.

Reason for thinking and asking of other softwares that may have known better settings that are able to encode any video to any size as crystal clear pixel vision x264 videos each time. If such software has been coded to do so as yet without user needing to know all complex settings to do the same.

That said there is an inherent problem with x264 the same with other encoders. That if the user cannot produce good results what is the point of encoder or softwares to start with. That each encoder need to be the best at what they can do. Including preset for best clear crystal pixel vision videos. But also as they know the setting should give out or have available for all resolutions not a few best encoder settings to be used when encode standard profile resolution videos 1280x720, 1920x1080 etc to any resolution and support for all valid resolutions at various main Aspect Ratios 16x9, 4x3 etc.

Guarding encode settings info is lame. I have better looking video than you can do la-la-la-la childish. Everyone should be able to encode to same standard which is the best that the x264 encoder (or any has ability to do).

There has never been any progress of anything without information of how to do something. Information is missing from all encoders some know but keep to themselves which progresses nothing.

Again until that changes any software or tool that can take good source video and reduce the resolution while maintaining the clear pixel vision quality of the video easily.

[manolito]

I think you just want to ignore the real reason for the lack of crispness of your encodes. It has absolutely nothing to do with the encoder. If you use highest quality settings in X264 like you already tried, then the encode will be visually transparent.

When you convert a 1080p source to 396p you will end up with a number of pixels per frame which is about 1/7.5 of the source (2073600 vs. 278784). This means your target will only have 13.45% of the information of your source. How in the world can you expect that the encoding result will have the same "crispness" as the source?

If you expect this then you are just ignoring the laws of physics. There are a number of ways to mimick a higher sharpness and/or contrast in the encoding result (experiment with different resizers, sharpeners and contrast enhancement), but nothing in the world can give you back the lost details which result from using such a low target resolution.

Bottom line:
If you want to retain the "crispness" of the source then do not resize to much lower resolutions.


Cheers
manolito

[StainlessS]

For everything Manolito said ++

If you want better looking results, ie more transparent to the source, then you would have to use settings that resemble closer to VerySlow/Placebo.


I was under the impression that speed parameters (eg VerySlow/Placebo) just give a little extra compression (when using CRF [EDIT: At the expense of more CPU muscle required during decompress]), although for bitrate mode above would perhaps be accurate.

Yep, just checked in MeGUI
https://s20.postimg.cc/qcamvo03h/megui.jpg (https://postimages.cc/)

EDIT: You would get a better result encoding at higher rez and similar bitrate (higher CRF) as higher rez video encodes more efficiently,
although what the magic formula is, I dont know [a certain amount of denoising would also help].