Hybrid(Windows/Linux/Mac): Input -> x264/x265/Xvid/VP8/VP9

[Leo99]

@Selur, I have a flv 500x281 @30.000fps, 16:9 (in/out par 1x1) and want to convert using x264 into mp4 container, but the Hybrid says:

"Problem Width and Height need both to be dividable through 2 without rest. Reason: Yv12-colorspace"

What's up and how to correct this..?

[dicto]

Selur, Is there any command line in x265 that haven't you include in Hybrid?

[Selur]

Strange somehow I didn't get any email notifications for the thread.

What's up and how to correct this..?

You resize to a resolution where width and height are both multiple of two.

Selur, Is there any command line in x265 that haven't you include in Hybrid?

Yes, Hybrid does not offer every command line option of x265.

Here's a list of the options Hybrid does not use :

Output Options:
   --log-level <string>          Logging level: none error warning info debug full. Default info
   --no-progress                 Disable CLI progress reports
   --[no-]cu-stats               Enable logging stats about distribution of cu across all modes. Default disabled
   --csv <filename>              Comma separated log file, log level >= 3 frame log, else one line per run

Input Options:
   -f/--frames <integer>            Maximum number of frames to encode. Default all
   --seek <integer>              First frame to encode 
   --dither                      Enable dither if downscaling to 8 bit pixels. Default disabled

Threading, performance:
  --[no-]asm <bool|int|string>  Override CPU detection. Default: auto

Temporal / motion search options:
--[no-]temporal-mvp           Enable temporal MV predictors. Default enabled
(this one I'll probably add over the weekend for the next release, but I doubt anyone will use it)

Rate control, Adaptive Quantization:
 --scaling-list <string>       Specify a file containing HM style quant scaling lists or 'default' or 'off'. Default: off
 --lambda-file <string>        Specify a file containing replacement values for the lambda tables
                                 MAX_MAX_QP+1 floats for lambda table, then again for lambda2 table
                                 Blank lines and lines starting with hash(#) are ignored
                                 Comma is considered to be white-space
Bitstream options:
 --hash <integer>              Decoded Picture Hash SEI 0: disabled, 1: MD5, 2: CRC, 3: Checksum. Default 0

Reconstructed video options (debugging):
-r/--recon <filename>            Reconstructed raw image YUV or Y4M output file name
   --recon-depth <integer>       Bit-depth of reconstructed raw image file. Defaults to input bit depth, or 8 if Y4M

[MS-HD]

@Selur
Hi,
I'm doing a video installation, Video Coming license.
Video installation by someone else; Copyright passed
Video help pass obstacles

https://www.youtube.com/watch?v=O_S7B86XrI8


Paid, make free video rendering programs.
please help

[Selur]

Linked video looks like there was a problem with the color space alignment.
If this was created with Hybrid I would need a debug output (+may be a small source sample) to see the source of the problem.
see: http://www.selur.de/support
If the file send to youtube wasn't send created with Hybrid, the problem seems to be a color space a alignment issue, which normally happens if a decoder and encoder (or filter) combination assume different input and output color samplings.

-> atm. not enough information to help

Cu Selur

[MS-HD]

skype adress fullcom.org add



youtube copyright: to pass

make video rendering

[Selur]

skype adress fullcom.org add

skype: rarely use it and only for family calls (and when calling a few friends in china)

youtube copyright: to pass

Doesn't YouTube reencode everything no matter what nowadays?
Last time I checked converting avc to vp9 didn't really make sense for youtube uploaders since youtube would still reencode another time.

make video rendering

you didn't read the infos that would be needed for me to reproduce a problem, did you?
(tried with different samples and got no problems converting AVC content to vp9 here)

[MS-HD]

to teach the world to pass obstacles videos

PayPal to pay fees

To use educational video program

[Selur]

Sorry, but it seems like some of the meanings you wanted to translate got lost during translation.

Cu Selur

[MS-HD]

Matched third party content
blocked in some of the world

youtube videos to pass the obstacles on.

to actively make money

video rendering programs to teach

[Selur]

@all: can someone make heads or tails out of what MS-HD is writing? it look more and more to spam to me.
@MS-HD: if you need Hybrid related help please read the support page and try to better describe your problem.
If you need help with something not related to Hybrid, best create a separate thread.

[MS-HD]

please

education videos

teaching programs

YouTube videos money

Payment: Paypal

[gonca]

Looks like he wants to use YouTube to spread videos which are banned in parts of the world, and make money at it.

blocked in some of the world

youtube videos to pass the obstacles on.

to actively make money

[Selur]

But do you get how Hybrid is related to it?

[gonca]

No

[Selur]

@all: Do we have any native French speaker here who could try to contact MS-HD and ask him?

[aymanalz]

@selur: I'm new to using your program, and I have a couple of suggestions:

1) Please add --allow-non-conformance as a tickbox, and preferrably make it default. I forgot to add it, and the encode failed after first pass. It would be a reminder to users to add that argument when necessary.

2) If possible, make the program pause the encode upon encountering a rectifiable error, instead of terminating. After a very long X265 encode, and just before mixing the audio, video and subtitle streams, my temp directory ran out of space, which caused the whole encode to fail. If it had paused with an error message, I could have made space on the drive.

[Selur]

regarding 1.: What did you do that you need non-profile conform clips? Normally Hybrid shouldn't allow such settings,... -> I might add it as an option, but will not enable it as default. (you can do that yourself, by saving it as a default)

regarding 2.: Did Hybrid abort with a 'Not enough free space on drive, still available .. MB'-message?

[aymanalz]

1: I'm not sure, but it is possible there were multiple non conformities. This was the settings I used:

x265 --pmode --pme --input - --input-res 1280x532 --fps 23.976 --profile main --high-tier --merange 114 --max-merge 3 --no-open-gop --bframes 6 --weightb --rc-lookahead 90 --pass 1 --bitrate 1000 --qpfile GENERATED_QP_FILE --psy-rd 1 --no-rdoq-level --no-psy-rdoq --cu-lossless --colormatrix bt709 --stats "R:\TEMP\g.stats" --output NUL

The trouble is that Hybrid gave an error only after doing the first pass, which took a couple of hours. It should alert the user about non conformance before starting the encode process at all.

2: Unfortunately, I did not save the error log. I don't remember the error verbatim, but I think it said "Disk I/O error". What happened was (I was observing it) that it finished the 2 pass encode, then began attaching the subtitle and audio, in the middle of which it abruptly stopped. I noticed that my ramdrive (where I have my temp directory) had become completely full. I should have allocated a larger drive. But if Hybrid had just paused instead of abandoned encoding, I could have cleared some space for the encode to complete. It stopped abruptly, the log file had an error message (disk i/o error), and it automatically deleted all temp files etc. The ramdrive was empty once again, before I did anything. So the encoding that was done was lost.

[Selur]

regarding 2: 'Disk I/O error' is not an error output Hybrid creates. If Hybrid isn't aware of the problem to begin with there is no way of thinking about how to create a workaround.
Also something is strange, since Hybrid normally doesn't delete temp files when running into a crash.
-> without a debug output I have no clue what happened.

regarding 1: don't see any problem with your settings and why they should create a non-conform output
(no clue why you would choose rc-lookahead 90 and merange 114, which seems to be simply bad choices)
-> if you can reproduce this please create a debug output

Cu Selur

[aymanalz]

^^
No problem, I'll save the debug output if it happens again. (Hopefully it won't, I should remember to set a large enough ramdrive.)

Could you tell me why the rc-lookahead and merange are bad choices? Because they are too high? The source was very high quality, so I thought a larger merange would help.

Another small tip, again for newbies like me - It is not enough to select a subtitle, you have to add it into the queue as well. Otherwise it will not be incorporated. (Another seemingly redundant step, especially if there is only on subtitle.)

Same thing for audio, it is not enough to select 're-encode' and set encoding parameters. You have to then add it to the audio queue.

[Selur]

Could you tell me why the rc-lookahead and merange are bad choices? Because they are too high? The source was very high quality, so I thought a larger merange would help.

Yes, for all normal content those two values are too high. Sure they might help, but the percentage of cases where such values create any visible perceivable change is really really small.

----

Yes, the queues have a meaning, which will not change. (there's also a chapter- and a tag-queue).
Most of the common cases can be automated by adjusting Hybrids default settings, but the normal defaults try to keep automation to a minimum.

Cu Selur

[aymanalz]

Yes, for all normal content those two values are too high. Sure they might help, but the percentage of cases where such values create any visible perceivable change is really really small.

Cu Selur

Is there a webpage or forum thread that gives a guide about X265 settings? I'm aware of the official X265 documentation, but what I'm looking for is a guide about which values are worth increasing. For example, does increasing the merange from default give enough of a quality boost to be worth the slowing down? Is me star worth choosing over umh? And so on.

I've been using x264 for a while now, and I have an idea about its parameters. But I'm new to x265.

BTW, is there a guide for hybrid's settings? Giving comparisons about the various denoisers and other filters available?

[Selur]

Don't know of any guide for Hybrid or for x265s settings.
So my guess is that, there isn't one atm.

[Leo99]

Selur, I have an video 1280x470 @23.976 with input PAR 1000x862 (DAR 2.35:1),
So I want to change the width to 1920. What an ideal Height size that doesn't messed up.. is it 1080, 800 or 816?

Please tutor me what the Input / Output PAR needs to be changed as well..?

[Selur]

simply set the output with to 1920 and Hybrid will adjust the rest, don't change the PAR,...

[Leo99]

OK thanks, I will try it.

[Leo99]

Selur, after trying, I get 1920x706 frame format with DAR 2.719:1... It's a rare aspect ratio.. oh no, I don't want it.
One question: What if I want the aspect ratio of 2.35:1 ?
It seems I should manually resize the height to 818 and Output PAR 1x1, but how about the Input PAR.. is it the same 1000x862 or 1x1..? Please correct me if I'm wrong.

[Selur]

if your input is:
1280x470 => SAR = 2,7234042553191489361702127659574
PAR of 1000x862 => Par = 0,862
SAR × PAR = DAR => DAR = 2,3475744680851063829787234042553


Since Hybrid, in it's default settings, will not change the PAR it's the DAR should stay the same if width&height are scaled by the same factor, because this way the SAR stay the same.
And with PAR&SAR not changing the DAR can only also stay the same.

1920/1280 = 1.5 -> 1920 x (470*1,5) = 705 (needs to be rounded to 706 since YV12/4:2:0 only allows resolutions with a height and width that is evenly dividable by 2)

-> seems like you changed some of the default settings, otherwise Hybrid wouldn't change the PAR and if Hybrid didn't change the PAR and the SAR staying the same, there is no way that the DAR changed.
read: https://en.wikipedia.org/wiki/Pixel_aspect_ratio

SAR = Sample Aspect Ratio = image width divided by the image height (without applying and resizing by the decoder)
DAR = Display Aspect Ratio = the image width divided by the image height and resized by the decoder according to the PAR
PAR = Pixel Aspect Ratio = width/height of an undistorted pixel

-> DAR = SAR * PAR
content is called anamorphic whenever the PAR is not 1 (=1x1), this is also called square pixel.

[Leo99]

I see.. But if I want to resize 1920x818, thus the Input PAR I should fill 1x1.. is it correct?