View Full Version : StaxRip
dejong12
20th August 2015, 00:06
Stax, another "problem". I've imported yadif x64 into the latest StaxRip version, but it's not deinterlacing correctly. I can see some color (de)blending happening. Neither deinterlacer works 100% correctly. When using the last x86 version of StaxRip (1.2.2.2) and using the same encoding settings with yadif, the deinterlacing works as it should, deinterlacing the source file 100% correctly.
videoh
20th August 2015, 03:28
I'm not seeing what you are seeing.
Downloaded episode.ts, used "mkvmerge -o episode.mkv episode.ts" (mkvmerge 8.3.0).
Opened using lwlibavvideosource(), both files show exactly 34500 frames. Mkv timecodes are 100% regular. I used 7.9.0.
videoh
20th August 2015, 03:29
It only happens when using DGSource as a source filter. Nah. I simply used mmg on your TS. DGSource didn't enter at all. And you said you have no issue when using dsmux to create the mkv.
videoh
20th August 2015, 03:30
What does DGSource do when a stream does not start on a recovery point? Any frames before the first decodable one are replaced with copies of the first decodable frame.
dejong12
20th August 2015, 08:56
Nah. I simply used mmg on your TS. DGSource didn't enter at all. And you said you have no issue when using dsmux to create the mkv.
I'm just saying that when you convert the uploaded TS to MKV (not with DSMux) and import that MKV into StaxRip the problem only shows when using DGSource. The other source filters handle the MKV fine.
stax76
20th August 2015, 09:32
muxing ts to mkv using mkvmerge is probably not a good idea if you want to encode afterwards, it should get async with other source filters too, not necessarily always but probably often, I experimented a while with it with little success, in this time I found out about dsmux.
Groucho2004
20th August 2015, 09:36
muxing ts to mkv using mkvmerge is probably not a good idea if you want to encode afterwards
I was wondering about that - why re-mux the .ts to mkv?
stax76
20th August 2015, 09:47
I was wondering about that - why re-mux the .ts to mkv?
because most tools fail on ts producing async output, dsmux at least sometimes helps to get around this, people without DGDecNV license or other payware like TS-Doctor or VideoRedo don't have great choice. At one point I wanted to code a TS demuxer myself, I would probably done it already, the problem is I practically never have TS sources personally so my motivation working on it is rather limited.
Ajvar
20th August 2015, 16:07
stax76, can you please tell me, does this 64 bit version provides video comparison feature (aka Codec comparison) and if you ever make a separate app for this? Last version I used was 1.2.2.2 beta and you wrote that you updated that coparison afterwards however I never could successfully download your Alphas of x86.
Thank you.
EDIT: It is working fine after installing Avisynth+.
dejong12
20th August 2015, 19:48
Videoh, could you take a look at what the creator of mkvmerge had to say about the issue?
https://github.com/mbunkus/mkvtoolnix/issues/1409
videoh
21st August 2015, 10:55
I'll look into it later today.
videoh
21st August 2015, 14:45
You gave me the full episode file episode.ts. It is fine in every way. It starts with SPS/PPS as it should. If you index it with DGIndexNV you get all the frames.
But Moritz is talking about a file test.ts. I don't know what that is and so cannot comment on Moritz's findings. You should give him the same file you gave me, or give me the same file you gave him.
Fadeout
22nd August 2015, 10:58
I'm getting crazy trying to figure this out, so someone please help...
I'm trying to convert a 1080i 50fps file into something my TV player can read, so a 1080p at 30fps (or 25 to better match the reduction?).
I have a slow CPU, but a 750 Ti videocard, so I'm guessing the best option I have is to use NVENC, right? I'm using StaxRip x64 and the first problem I had is that by default the program doesn't come with Yadif, and so I couldn't properly deinterlace the image.
Now, the thing is, since then I managed to find a 64 bit version of Yadif and so deinterlace the image, but consider these two conversions:
1- did not use Yadif, so the original source, but with SelectEven. so I get a "progressive" 1080p 25fps, because I take only 1 field.
2- used Yadif, getting a progressive 1080p 25fps.
The problem is that the first video, with deinterlacing FORCED on a PC media player produces a result that is IMMENSELY better and smoother, even if the two files are a similar size.
But of course I can't do that on a normal player I need to use, so with the first file I only get an interlaced image.
This doesn't make a lot of sense to me. There should be a way to code the file as a "real" progressive by retaining the quality of the first file. How is it possible that a 1080p 25fps forced to deinterlace during playback gives substantially better results than a 1080p 25fps that was instead properly deinterlaced during the encoding, and with same file size?
Also, is it normal I barely get 1.3x encoding speed on a E8400 CPU with a 750 Ti? Are there better options than FFVideoSource? I can manage to have better speed (around 1.5) with MediaCoder because it offers different deinterlacers that seems to perform better (including Yadif, that is immensely slow on StaxRip. So I'm looking for suggestions on what the best options are.
Fadeout
22nd August 2015, 11:17
Also, a conversion from 1080i 50fps to 1080p 50fps, using yadif reduces the smoothness of the movement.
Whereas the 1080i 50fps converted without previously deinterlacing to 1080p 25fps, when deinterlacing forced in the player is MUCH smoother than 1080p 50fps.
So that doesn't make any sense, there's something wrong that loses that information in the process and there must a way to make a better conversion.
EDIT:
Deinterlacing with MPC-HC and MadVR = super smooth
Deinterlacing with MPC-HC and without MadVR = super smooth
Deinterlacing with default Windows Media Player = super smooth
Deinterlacing with VLC and Yadif = loses smoothness (tried all other deinterlacers, same result)
The fourth case is what I get with my encodes. So what the hell are WMP and MPC-HC using to deinterlace that produces so much better results?
Is there a way to edit the file so to turn it from progressive to interlaced, so to force the deinterlacing normally? Because that's my whole problem: I have a perfect file that says it progressive but that would look perfect when forcing interlacing, but of course my TV player can't deinterlace unless the video is properly flagged as interlaced. So I'd need the option to change that flag in the file and turn on deinterlacing.
NikosD
22nd August 2015, 17:34
A new QSVEncC v2.10 is out mainly for Skylake.
It uses a new mode of HW encoding enabled only in Skylake which leverages the fixed-function HW during encoding, without the use of GPU EUs.
This is the only mode supported by Nvidia from the beginning of their HW encoders and this is the mode of all HW decoders (AMD, Nvidia, Intel)
The old mode is PG-mode (Processor Graphics) and the new one is FF-mode (fixed function)
It can be used for H.264 only, as I read.
For the first time in HW encoders, it uses a programmable adaptive rate control in order to trigger a second pass if needed to achieve the targeted bitrate, without using any CPU or GPU processing, only fixed function HW.
The reason Intel added this mode is for "power saving", because the GPU load is extremely low, but it is possible that performance maybe drops that way and perhaps quality too.
We need a Skylake to test it.
Intel calls this mode as a low latency, real time encoding.
Actually, I had asked Intel about that a few months ago:
https://software.intel.com/en-us/forums/topic/542414
Yups
22nd August 2015, 17:39
Isn't QSVEncC japanese or chinese?
NikosD
22nd August 2015, 17:50
Japanese
Yups
22nd August 2015, 18:27
Without japanese reading skills how do do think could someone try it out?
NikosD
22nd August 2015, 18:30
The help file which lists all possible commands is in English and you can print it out in your screen via the CLI.
All the CLI switches are in English of course.
No need to know Japanese.
If you need to browse his site, just use Chrome with the auto translation.
Yups
22nd August 2015, 19:40
Staxrip is using QSVEnc 2.00 beta12 or can I manually copy a new QSVEnc version into the folder?
NikosD
22nd August 2015, 19:43
Yes you can do that and copy an updated version of QSVEncC, but without GUI support of the new functionality of the updated version.
We have to "push" Frank (stax76) to update the QSVEncC GUI to include the latest functionality of QSVEncC CLI.
Yups
23rd August 2015, 17:17
My browser blocks QSVEncC 2.10 download because of a virus (possibly a false alarm). I better wait for an updated Staxrip.
videoh
23rd August 2015, 19:31
Videoh, could you take a look at what the creator of mkvmerge had to say about the issue?
https://github.com/mbunkus/mkvtoolnix/issues/1409 Did you lose interest in this?
stax76
24th August 2015, 14:03
Yes you can do that and copy an updated version of QSVEncC, but without GUI support of the new functionality of the updated version.
We have to "push" Frank (stax76) to update the QSVEncC GUI to include the latest functionality of QSVEncC CLI.
I'm working on it. ;)
NikosD
24th August 2015, 14:05
Nice!
So, you got your Skylake ? ;)
stax76
24th August 2015, 16:57
I've still problems with the new PC, I received new RAM modules (http://www.crucial.com/usa/en/bls4k4g4d240fsa) today, unfortunately the symptoms are still there, blank screen and error beep codes. 4 long beeps and one short beep, after the short beep it restarts. I don't know exactly what the error codes mean, the info I found is somehow vague, according to the Gigabyte website (http://www.gigabyte.com/products/product-page.aspx?pid=5496#sp) the bios is AMI.
JohnLai
24th August 2015, 17:35
I've still problems with the new PC, I received new RAM modules (http://www.crucial.com/usa/en/bls4k4g4d240fsa) today, unfortunately the symptoms are still there, blank screen and error beep codes. 4 long beeps and one short beep, after the short beep it restarts. I don't know exactly what the error codes mean, the info I found is somehow vague, according to the Gigabyte website (http://www.gigabyte.com/products/product-page.aspx?pid=5496#sp) the bios is AMI.
You mean exactly like this https://www.youtube.com/watch?v=QMvy403xMyQ ?
The video is for older socket 775, is the beep sound the same?
Just asking......have you clear CMOS?
Using 8 pins power (ATX_12V_2X4) for the CPU?
Try different RAM socket? (one RAM at a time?)
If you didn't install any RAM, does it still beep 3 long 1 short?
Are pins on the cpu socket itself in good condition?
Try reseating your CPU yet?
In any case, you better stick with gigabyte Qualified Vendors List (QVL) for RAM.
Did your previous RAM is listed by gigabyte QVL?
Cause your current Crucial Module P/N. is not listed at QVL.
stax76
24th August 2015, 18:19
The long beeps are different, I think I found the problem, whenever I use RAM socket 1 or 3 the error occurs, socket 2 and 4 are OK, dual channel needs 1+2 or 3+4 so I absolutely have to return the motherboard.
JohnLai
24th August 2015, 18:21
The long beeps are different, I think I found the problem, whenever I use RAM socket 1 or 3 the error occurs, socket 2 and 4 are OK, dual channel needs 1+2 or 3+4 so I absolutely have to return the motherboard.
Before that....wanna try update the bios first?
There is an update for the board for RAM compatibility.......
stax76
24th August 2015, 18:28
Before that....wanna try update the bios first?
There is an update for the board for RAM compatibility.......
Sounds like a good idea.
edit:
Did not help. :(
JohnLai
25th August 2015, 04:00
Sounds like a good idea.
edit:
Did not help. :(
Hmm.....go RMA the board...
Still, it is weird running in single channel mode works while dual channel doesnt.
Either your CPU cooler shorting out something behind the board. (cpu cooler metal backplate for instance? Try using stock cooler from haswell,ivy bridge?)
Or the CPU doesnt have proper contact with the socket pins (integrated memory controller in the cpu)
stax76
25th August 2015, 09:42
First I tried the new motherboard/CPU/RAM in my current PC which uses a old Scythe CPU cooler with quick mounting system not using a backplate. After this failed I realized hardware problems happen way to often so I decided that from now on I don't change my current PC but always build a second PC whenever I change the motherboard, the second/new PC uses a Thermalright HR-02 Macho Rev. B cooler with mounting system using a backplate. I'll receive another board (GA-Z170X-UD3) in two days...
Yups
27th August 2015, 01:00
QSVEnc 2.11 is available with HEVC support for SKL.
NikosD
28th August 2015, 10:44
An interesting comparison of HW encoders (HEVC, H.264) vs SW encoders (x265, x264) in terms of performance and quality using QSVEncC v2.11
It looks like that on Skylake both QSV H.264 modes - PG and FF - have the same speed and QSV HEVC encoding has half speed of QSV H.264.
Compared to SW encoders, QSV HEVC is 2x faster than x264 and 12x (!) faster than x265.
Quality pictures and performance numbers here:
http://rigaya34589.blog135.fc2.com/blog-entry-673.html
Yups
28th August 2015, 23:57
His picture quality tests are nonsense nevertheless. 1080p is required at least with some proper video material and not some cartoon crap where low bitrate does fine.
stax76
29th August 2015, 01:58
I'm getting a mess of artifacts using QSV HEVC encoding, was anybody more successful?
NikosD
29th August 2015, 04:41
Anime is extremely popular to the East and makes a lot of sense there.
Schattenspieler
29th August 2015, 07:29
But the contentual popularity does not positivley affect (-> reduce) the general lack of detail in animes compared to real HD source material. And for that reason they could be used as an example (among others), but if you want to show the general quality of encondings, you would have to use different image sources, e.g. others with high details level. Otherwise such a comparison of image quality is not senseless but to some degree pointless. Just my 2 cent though …
Yups
29th August 2015, 15:41
I'm getting a mess of artifacts using QSV HEVC encoding, was anybody more successful?
How did you use HEVC and what driver do you have installed?
But the contentual popularity does not positivley affect (-> reduce) the general lack of detail in animes compared to real HD source material. And for that reason they could be used as an example (among others), but if you want to show the general quality of encondings, you would have to use different image sources, e.g. others with high details level. Otherwise such a comparison of image quality is not senseless but to some degree pointless. Just my 2 cent though …
Exactly this, plus for a better comparison something bigger than 720p.
stax76
29th August 2015, 16:09
How did you use HEVC and what driver do you have installed?
I did not use special settings and tried two different modes and two different sources with same results.
QSVEncC (x64) 2.11 by rigaya, build Aug 26 2015 20:34:15
based on Intel(R) Media SDK Encoding Sample 6,0,0,98947
avi reader: enabled
avs reader: enabled
vpy reader: enabled
avqsv reader: enabled [H.264/AVC, HEVC, MPEG2]
Environment Info
OS : Windows 10 (x64)
CPU: Intel Core i5-6600K @ 3.50GHz [TB: 3.89GHz] (4C/4T)
RAM: Used 2708 MB, Total 8069 MB
GPU: Intel HD Graphics 530 (24EU) 1150MHz (10.18.15.4274)
Media SDK Hardware API v1.16
Supported Enc features:
Supported Vpp features:
Media SDK Software unavailable.
I installed the driver from the Gigabyte site:
15.40.5.1.64.4274
2015/08/19
The Intel driver from the Intel site is older:
15.40.4.64.4256
8/12/2015
Yups
29th August 2015, 22:18
I will try once there is Intel H265 build into the GUI.
stax76
29th August 2015, 23:28
Hey everybody,
I uploaded a new release, it does not have big new features but many small improvements and updated tools of course, sorry if I missed some posts.
StaxRip x64 1.3.1.6 beta (2015-08-30)
New: HEVC and MPEG-2 encoding modes added to Intel encoder
New: KNLMeansCL plugin added
New: Added option for fixed bitrate even though using a fixed bitrate is not recommended. It was added because over the years it was requested dozens of times. StaxRip will show a warning telling to rather use quality mode and constrain the maximum data rate if necessary
New: Added option to process a audio profile using the context menu in the main dialog
Tweak: Play feature adds resize filter to VapourSynth play script if the source PAR is non 1:1
Tweak: Added clear feature to audio file context menu to easily remove a audio file
Tweak: Added screen bounds magnet docking feature
Tweak: When jobs are completed StaxRip activates now the main window again like it did before but now StaxRip checks if a player is in the foreground and prevents activation if true, it works with MPC, VLC and MediaMonkey, other players might be added on request
Tweak: ProjectX and dsmux are always enabled by default but StaxRip checks only if Java and Haali is installed in case of ProjectX and dsmux are actually executed
Tweak: Improved DGDecNV and DGDecIM integration
fix: filenames with characters like ä ü ö not working VapourSynth
fix: fullscreen mode in preview was not working in Windows 10 because the task bar was in front
Update: x264 r2597
Update: x265 1.7+433
Update: AVSMeter 2.1.0
Update: ffmpeg 2015-08-28
Update: qaac 2.52
Update: NVEncC 1.09
Update: QSVEncC 2.11
Update: MKVToolNix 8.3.0
https://github.com/stax76/staxrip/releases
I will try once there is Intel H265 build into the GUI.
The new release has a new profile 'Intel H.265' but I did not reset the profiles so the new profile is only available if the profiles are manually reset.
Yups
30th August 2015, 00:48
Just a quick test on one video, I can say no artifacts issue for me with HEVC.
stax76
30th August 2015, 01:22
Which CPU and driver do you use?
dejong12
30th August 2015, 01:32
NVEncC 1.10 was released 12 days ago by the way.
stax76
30th August 2015, 03:05
Seems the artifacts I get with the Intel HEVC encoder happen only in the first 10 seconds so I can make a codec comparison.
NVEncC 1.10 was released 12 days ago by the way.
I've updated it locally now. I like that it has a VapourSynth reader, I believe it's the first and maybe the only encoder (besides QSVEncC) with VapourSynth reader, this should allow me to remove the piping solution making the copy command line and execute command line feature in the options dialog work again.
JohnLai
30th August 2015, 04:11
Hey everybody,
I uploaded a new release, it does not have big new features but many small improvements and updated tools of course, sorry if I missed some posts.
https://github.com/stax76/staxrip/releases
No "QSV --avqsv --output-res -o - -c raw | NVEncC --y4m"?
T_T
NikosD
30th August 2015, 07:48
Hey everybody,
I uploaded a new release, it does not have big new features but many small improvements and updated tools of course, sorry if I missed some posts.
https://github.com/stax76/staxrip/releases
Nice work and thanks for the fixes.
A few things regarding the new version:
BUGS
1) This HEVC .ts clip (and a few others like this) crashes badly the default (automatic) video source (LWLibav).
https://www.sendspace.com/file/49zvqy
It gives this error and closes StaxRip x64.
Exception
Arithmetic operation resulted in an overflow.
System.OverflowException: Arithmetic operation resulted in an overflow.
at StaxRip.MainForm.OpenVideoSourceFiles(IEnumerable`1 files, Boolean isNotEncoding, VideoFilter preferredSourceFilter) in D:\Projekte\GitHub\staxrip\Forms\MainForm.vb:line 2097
Also it crashes the LSMASH video source giving this error but without closing the app (I can select another source)
------------------------------------------------------------
Error opening source
------------------------------------------------------------
System exception - Access Violation
(D:\Video\Out of BD spec clips\L5.1\Beyond 1080p\4K (3840x)\H.265\140512_hm130_4s_sao_dbf temp files\140512_hm130_4s_sao_dbf_Source.avs, line 2)
LSMASHVideoSource("D:\Video\Out of BD spec clips\L5.1\Beyond 1080p\4K (3840x)\H.265\140512_hm130_4s_sao_dbf.ts")
The other video source options (non-commercial) FFVideo and DSS/DSS2 work fine.
2) Unfortunately, I wasn't so lucky regarding some 10bit HEVC files.
For example this file (and a lot of others) can't be opened by any non-commercial video sources:
https://www.sendspace.com/file/8vv7qp
3) This clip can be handled without problems, using the "single-file" mode of StaxRip x64 by eac3to automatically (no user selection)
https://www.sendspace.com/file/vb152k
BUT when I choose the "batch-file" mode selecting this clip along with others, at the final conversion the audio part is missing completely (without errors) (!)
The video part is OK, only the audio is missing.
Requests for QSVEncC GUI:
1) I think it would be easy and probably useful to add, in your GUI, the following parameters:
--level <string> set codec level, default auto
--profile <string> set codec profile, default auto
H.264: Baseline, Main, High
HEVC : Main
MPEG2: Simple, Main, High
2) The three different codecs that QSVEncC supports right now (H264, H265, MPEG2) and probably VP8 and MJPEG that is going to support in the next versions, don't share the same encoding features.
For example there is no ICQ encoding mode for MPEG2.
You can check the encoding features available, by executing the command --check-features or directly from the GUI by selecting "Check Features" from the arrow (it's not working using latest v2.11 for x64, only for x86 - I have already mentioned that to the developer and he is going to fix it in the next QSVEncC version)
I wonder how difficult would be to adjust the QSVEncC GUI according to the codec selection and change the name of the template to single "Intel H.265/H.264/MPEG2" instead of different templates "Intel H.264" and "Intel H.265"
Missing updates:
MediaInfo is at 0.7.76 version.
Yups
30th August 2015, 09:49
Seems the artifacts I get with the Intel HEVC encoder happen only in the first 10 seconds so I can make a codec comparison.
Yes there is a problem somewhere within the first 10 seconds which I believe is caused by fps*10= GOP Length from QSVEnc. Try Intels default 256 GOP, it seems to work.
2) The three different codecs that QSVEncC supports right now (H264, H265, MPEG2) and probably VP8 and MJPEG that is going to support in the next versions, don't share the same encoding features.
For example there is no ICQ encoding mode for MPEG2.
You can check the encoding features available, by executing the command --check-features or directly from the GUI by selecting "Check Features" from the arrow (it's not working using latest v2.11 for x64, only for x86 - I have already mentioned that to the developer and he is going to fix it in the next QSVEncC version)
I wonder how difficult would be to adjust the QSVEncC GUI according to the codec selection and change the name of the template to single "Intel H.265/H.264/MPEG2" instead of different templates "Intel H.264" and "Intel H.265"
Lookahead doesn't work either with HEVC, I don't know if this is an Intel or QSVEnc limitation though.
stax76
30th August 2015, 13:23
Yes there is a problem somewhere within the first 10 seconds which I believe is caused by fps*10= GOP Length from QSVEnc. Try Intels default 256 GOP, it seems to work.
I tried 256 but it didn't help.
Generally my first impression of the Intel H.265 encoder isn't that good, both speed and quality is somehow disappointing. :(
No "QSV --avqsv --output-res -o - -c raw | NVEncC --y4m"?
I'll add it if it's not too much work, regarding the current piping solution, maybe it's better to keep it as option instead to remove entirely.
@NikosD
Thanks for investigating, I'll work on it.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.