RealVideo 9 Information Thread Discussion

[karl_lillevold]

yes, those big jumps are not good for smooth video playback since we rely on the hardware timer for accurate time measurements. On my Pentium systems the 'tick' test has always provided relatively smooth numbers. Were you doing anything CPU intensive while you ran the tick.exe test program?

example from my system:

140515109 16 140513183 7 140513183 7 140513183 7
140515109 0 140513191 8 140513191 8 140513191 8
140515125 16 140513199 8 140513199 8 140513199 8
140515125 0 140513207 8 140513207 8 140513207 8
140515140 15 140513215 8 140513215 8 140513215 8
140515140 0 140513222 7 140513222 7 140513222 7
140515156 16 140513230 8 140513230 8 140513230 8
140515156 0 140513238 8 140513238 8 140513238 8
140515171 15 140513246 8 140513246 8 140513246 8
140515171 0 140513254 8 140513254 8 140513254 8
140515187 16 140513262 8 140513262 8 140513262 8
140515187 0 140513269 7 140513269 7 140513269 7
140515203 16 140513277 8 140513277 8 140513277 8
140515203 0 140513285 8 140513285 8 140513285 8
140515218 15 140513293 8 140513293 8 140513293 8
140515218 0 140513301 8 140513301 8 140513301 8
140515234 16 140513308 7 140513308 7 140513308 7
140515234 0 140513316 8 140513316 8 140513316 8

[bilu]

AcroTray.exe 1752 C:\Program Files\Adobe\Acrobat 5.0\Distillr\AcroTray.exe
flashget.exe 1928 C:\Program Files\FlashGet\flashget.exe
iexplore.exe 3508 C:\Program Files\Internet Explorer\iexplore.exe
avpcc.exe 1456 C:\Program Files\Kaspersky Lab\Kaspersky Anti-Virus Personal Pro\avpcc.exe
avpcc.exe 1704 C:\Program Files\Kaspersky Lab\Kaspersky Anti-Virus Personal Pro\avpcc.exe
avpm.exe 2812 C:\Program Files\Kaspersky Lab\Kaspersky Anti-Virus Personal Pro\avpm.exe
persfw.exe 1736 C:\Program Files\Kerio\Personal Firewall\persfw.exe
PrcView.exe 3652 C:\Program Files\Process Viewer\PrcView.exe
sgbhp.exe 2032 C:\Program Files\SpywareGuard\sgbhp.exe
sgmain.exe 1980 C:\Program Files\SpywareGuard\sgmain.exe
Explorer.EXE 1292 C:\WINDOWS\Explorer.EXE
alg.exe 1436 C:\WINDOWS\System32\alg.exe
csrss.exe 464 C:\WINDOWS\system32\csrss.exe
CTHELPER.EXE 1532 C:\WINDOWS\System32\CTHELPER.EXE
lsass.exe 544 C:\WINDOWS\system32\lsass.exe
mmc.exe 928 C:\WINDOWS\system32\mmc.exe
nvsvc32.exe 1596 C:\WINDOWS\System32\nvsvc32.exe
services.exe 532 C:\WINDOWS\system32\services.exe
smss.exe 408 C:\WINDOWS\System32\smss.exe
spoolsv.exe 1000 C:\WINDOWS\system32\spoolsv.exe
svchost.exe 720 C:\WINDOWS\system32\svchost.exe
svchost.exe 764 C:\WINDOWS\System32\svchost.exe
svchost.exe 880 C:\WINDOWS\System32\svchost.exe
svchost.exe 912 C:\WINDOWS\System32\svchost.exe
svchost.exe 1836 C:\WINDOWS\System32\svchost.exe
winlogon.exe 488 C:\WINDOWS\system32\winlogon.exe
MozillaFirebird.exe 900 D:\mozilla\MozillaFirebird\MozillaFirebird.exe
thunderbird.exe 2496 D:\mozilla\thunderbird\thunderbird.exe


CT Helper is from Creative
nvsvc32.exe is "NVIDIA Driver Helper Service"
Kerio Personal Firewall
Kaspersky Antivirus Monitor
PrcView.exe is Norton Utilities Process Viewer, wasn't running then, but it's useful to generate this kind of reports :)

I'll try disabling those items for moments to see what happens in tick.


Bilu

[bilu]

I've sent you a lot of tick files, but I wasn't able to find the problem :( Maybe it's hardware related...


Tick files:

a_noKAV.txt - no Antivirus Monitor
b_noNV.txt - No NVidia service
c_noHQNV.txt - Changed from Best Quality to Best Performance in NVidia Control Panel
d_noCT.txt - No Creative Helper
e_noMOZ.txt - Closed Mozilla and other running programs listed in the post
f_noSG.txt - Disable Spyware Guard
g_noKAV2.txt - no Antivirus Control Panel
h_noPFSG2.txt - no Firewall, killed Spyware Guards process
i_stdwin.txt - only Windows process running
j_nonet.txt - disabled networking, since I use Connection Sharing
k_safe1.txt - Safe Mode
l_safe2.txt - another tick in Safe Mode
m_bios.txt - after changing BIOS setting, ran in Safe Mode


Bios Settings changed:
Processor In Order Queue Depth from 4 to 1
L2 Cache ECC disabled
SDRAM Cycle Time Tras/Trc 7/9 to 5/7
RAS-to-CAS Delay 3 to 2
RAS Precharge Time 3 to 2
CPU Latency Timer Disabled
PCI Delayed Transaction was already disabled


Best regards,
Bruno

[bilu]

I'm using HPG for the first time.

Unique features I love in HPG:
- job and audience generators;
- integrated Dropfilter and the possibility to use RV9 prefilters ( although I'm not using them);
- Vorbis support;

Unique features I love in AutoRV9:
- compressibility test;
- subtitle integration (could be better);
- better audio integration than HPG;

My dream: have a nice variable framerate filtering so I can auto-IVTC without worrying about hybrid movies and/or speed animes, and then integrate a tool with Robot4Rip that grabs the AC3 and D2V file,creates an RV9 job and starts encoding. This tool could burn the raw Vobsub subs or place them in the same directory, so that RV9 could be played nicely with external subtitles I didn't even OCRed. I was close once with my R4R_ENC batch tool, except for auto-IVTC :)

That is my dream of laziness: an encode by clicking Start :)


Bilu

[Sirber]

If you find something about hibrid movies, I'll be glad to add it to Surreal UI :D

[bilu]

@Sirber

Take a look at this interesting thread:

@VFW/DShow experts: VFR in Avisynth could be good news for Matroska
http://forum.doom9.org/showthread.php?s=&threadid=63541


@Karl
What could be done both in RV9 and Avisynth to increase compressibility to fit a certain size? I almost get the impression that whatever I do I'll lose quality so I'd better stick to EHQ=80, fixed vbrquality value, and split for CDs :)

If bitrate is not good enough for RV9, it washes the picture as needed.
Filtering to increase compressibility in RV9 would do the same I think, so there's no need to do it twice :)

I have to try C3D, Deen and Undot with RV9 to see if it's worth the compressibility gain and how does it look. Also I've heard that resizing doesn't help that much.


Bilu

[karl_lillevold]

@bilu: time permitting, we are planning to come out with 2-pass VBR improvements to the encoder. We also have another small enhancement to both encoder and decoder that will be out shortly.

C3D works very well for RV9. Without having used the compressibility measurements too much, I am not sure how much I would trust them :( Do not believe everything you hear, resizing works very well to increase compressibility. Like I have mentioned a few times, with EHQ=80 I resize such that I get 0.13 bits/pixel or higher. For grainy, noisy, lots of action material, or very low rez (w=400 or below), I go higher 0.18-0.20 bits/pixel.

Hope this helps.

[bilu]

Originally posted by karl_lillevold
@bilu: time permitting, we are planning to come with 2-pass VBR improvements to the encoder. We also have another small enhancement to both encoder and decoder that will be out shortly.

I wasn't complaining, it's just that resizing would be my best option since movies already look so clean :)

I was thinking that filters that help compressibility under MPEG-4 wouldn't necessarily help in RV9, but maybe it's time for me to try.

I know that compressibility tests on AutoRV9 take vbrquality 75 as reference like a first-pass. Is quality decreasing linear, i.e. if you decrease from quality 75 to quality 50 will it decrease mostly the same percentage in size in every movie?


Bilu

[karl_lillevold]

I was running some StackVertical comparisons with two DirectShowSource (rmvb) videos, and suddenly Avisynth started crashing with an unknown exception. So I simplified my script to display just one video, still no luck, unknown exception 4 out of 5 times, both in mplayerc and VirtualDub. Finally I realized when I re-installed Avisynth late last week, and I said 'yes' to install ffvfw. I did not notice any immediate problems at that time. So then I uninstalled ffvfw, and now everything is working again. Maybe I could have figured out how to make it work, perhaps by disabling ffvfw 'raw' video, or installed a newer version, but since I don't really need it, this was quicker. So, just something to keep in mind, if Avisynth starts crashing while using DirectShowSource, Gabest's RealMediaSplitter, and rmvb files.

[Rrrough]

After experimenting with RV9 on a Windows box I'm quite impressed by the quality. Especially with EHQ + new 2-pass. Unfortunately, transcoding to RV9 seems to be a "Windows-only" solution, because on a Linux box, AFAIK it's only possible to transcode an MPEG2-stream into raw avi and feed that into Helix producer for Linux. A nicer solution would be an export module for transcode (http://www.theorie.physik.uni-goettingen.de/~ostreich/transcode/).
Also 5.1 RealAudio sounds great, but transcoding from AC3 under Linux could be easier with an liba52 (http://liba52.sourceforge.net/) -interface for Helix Producer, resp. a transcode-export-module :D
Are there any plans for enhancing Producer on the linux platform ?
Anyway, the "support" for this codec within this forum is great !

cheers

[karl_lillevold]

@Rrrough: thanks! and yes, you are right about "Windows-only" :(

since the RV9 encoder runs as fast on Linux as on Windows, it is really unfortunate Producer on Linux (and OS X) supports only uncompressed AVI (and MOV) files and the filesize even has to be less than 2GB. We would really like to see this improve, for instance in the direction of your suggestions, but the Producer team is fully booked with other tasks in the foreseeable future.

I have already mentioned this before: since most of the Helix system is open source, it would be relatively easy for a skilled programmer with Linux experience, to work on these issues, and contribute the improvements back to the Helix Community. I would be more than happy to help someone get started, with pointers on where to start and the right contacts. Just let me know.

[Rrrough]

hey karl, thanx for your answer. too bad that my favorite OS isn't really supported (yet), so I've to stick with a seperate Windows-partition a bit longer, but the codec's quality is definetly worth it.

tusen takk og ha det bra (IIRC :) )

[karl_lillevold]

In Producer 9.2 Milestone 8 changes to encodingComplexity were made (again).. So now only 'low', 'medium' and 'high' work. 'very-high' does not work. The old codec specific registry keys and codecProperties "encoderComplexity" in the audience file, with 85 as the complexity, still work though.

More changes are expected, so for now, there is no need to upgrade. Just use Milestone 7 with my VBR fix (http://forum.doom9.org/showthread.php?s=&threadid=64163).

EDIT: and even 'high' and 'medium' and 'low' have been mapped differently, so the only working methods to set encoding complexity are the registry and codecProperties.

[phrentec]

Is there anyway to convert bink videos(.bik) into realvideo. I have tried to use the wrpr codec to frameserve from the radvideo to an avi but it seems to crash before it can be frameserved. I know mediaplayerclassic can play bink videos(even in the .exe form) through using the binkw32.dll so is it then possible to open bink videos with helixproducer through avisynth (this may belong in the avisynth forums i guess) via the binkw32.dll.

I am just looking for a way to convert bink videos to realvideo without having to use a lossless codec in the process.

[iwod]

It seems that M8 has alot of changes and somewhere in that releasenote suggested that EHQ is now not part of RV9 and is put into a new Codec.

Is that Rv10?? :D

[Sirber]

can you paste the text here?

[RadicalEd]

RealVideo EHQ/Encoding Complexity Changes
The number of levels available for encoding complexity was reduced from 5 (very-high, high, normal, fast, vary-fast) to 3 (high, medium and low) for several reasons:

Only some of the 5 options were triggering unique behavior in most of the codecs that supported the encodingComplexity property. Thus, users might change the setting from fast to very-fast with no change in behavior. This is not a very good user experience.
The EHQ technology added to RealVideo 9 was separated out into a new codec - itself with 3 ranges. The new codec is yet complete but RealVideo 9 is still available in this build. RealVideo 9 no longer includes the EHQ technology so there is no method in Milestone 8 to exercise EHQ. Please use Milestone 7 if you wish to encode with RealVideo EHQ.

From the M8 release notes (https://distribution.helixcommunity.org/files/documents/50/761/ReleaseNotes9.2M8.htm)


No more rv9 EHQ, that is a bit disturbing :\

[karl_lillevold]

No more rv9 EHQ, that is a bit disturbing
it is still there and will never go away :D, it is available via codecProperties or registry settings, but unless you want to use those, stick with Milestone 7 + my VBR improvement.

[Sirber]

Why does the compagny want to remove EHQ? It's like going back to the dinosaur age ;)

[iwod]

Originally posted by Sirber
Why does the compagny want to remove EHQ? It's like going back to the dinosaur age ;)

Simply Becoz RV9 is total crap.....

And So one day they break the bitstream, uncover somthing with Fast motion. A New PP build out of Codec. Better EHQ with even higher EHQ setting. Better VBR, other general improvement..... etc.... little by little.......

We begin to see a new age..... No we are not in Diniousor age....... we went to Mars......

Welcome to RV10 :D

[Sirber]

RVX will be RV9 + EHQ?

[RadicalEd]

Not too sure. I had the privelage of coming across Karl's roommate in a random IRC chat one day, so I got a bit of a personal Q&A. I asked about RV10 and Karl said they weren't really focused on it yet because RV9 with EHQ was so much better and almost like a completely different codec.

It's anybody's guess I suppose :]

[RadicalEd]

There seems to have been an update allowing for EHQ in milestone 8.

https://distribution.helixcommunity.org/servlets/ProjectDocumentList?folderID=169&maxDepth=&paginated_page=4

scratch that, the update was just karl's twopass updates. Stevemc screwed up :P

[karl_lillevold]

yes, updates made too late at night without enough communication can sometimes result in not very useful results. Producer M7 + VBR fix is still the recommended encoding combination.

EDIT:

i created another erv4 (RV9) encoder DLL to replace the patch, that
remaps and corrects the complexity levels such that

M8 complexity notation -> M7 complexity notation

high -> very-high
medium -> normal
low -> fast

in this way, Milestone 8 with this patch, now allows for both the
latest 2-pass VBR improvements + very high encoding complexity.

Notes:
without this patch, the complexity levels in M8 are like this

high (default) -> normal
medium -> fast
low -> fast

[iwod]

Originally posted by Sirber
RVX will be RV9 + EHQ?

That was beocz i was boring and spread a bit of rumors. :D I do believe Rv10 should be in development ( at least on paper ) but i really don't think they would come out so soon.

Considering that there are other little improvement hopefully karl said b4 that is coming up. I belive Rv9 still has some time b4 it goes.

To me, i am very happy with my anime rips with VBR fix and EHQ at my current bitrate. And with some of the upcoming enhancement i believe Rv9 's best has yet to be seen.

( And with Gabest ' MPC and Mkv format Rv9 is simply the best. )

[phrentec]

When encoding using producer.exe i get the error
"Warning: Channel 0 clips moderately."
What does that mean and how can I stop it from happening?
Thanks.

@Sirber: Is there a link to the most current binary of SurrealUI.exe because the link at http://sirber.no-ip.com/~sirber/SurrealUI.exe seems to be down.

[karl_lillevold]

Warning: Channel 0 clips moderately
information from producer mailing list, where this question was just asked:

Those messages are warning from one of the audio checks producer does.
We call these audio watchdogs. You can find more information about the
audio watchdogs in the Producer help page by typing 'producer -m' and
looking at the "-daw" option and reading about the 4 audio watchdog
checks that producer does.

In particular, it looks like your audio signal is too loud and is
clipping.

[phrentec]

thanks for the reply
sorry but what is meant by "clipping"?

[Sirber]

IIRC, sound goes out if it'S range, like if you normalize too much.

[karl_lillevold]

clipping is a well known digital audio problem when the levels are too loud. Imagine a nice smooth sine wave getting clipped off, ending up with square form instead. This sounds very bad. As long as you are encoding from a file, don't worry about this warning, especially, when it's only "moderate". Producer has a gain control pre-filter with a high quality limiter, if you need to reduce the input levels before compression.

[karl_lillevold]

in other news, upcoming RV9 encoder DLLs (erv4.dll) will be compiled with the Intel Compiler (ICL) version 7.1, which we have now finally fully integrated in our build system. Previously, only certain files were compiled with this compiler, since MSVC 6.0 did not support the required inline asm code.

Since all CPU intensive code is already fully assembler optimized, the improvement is around 5%. Still, worth including, just by changing the compiler (anything to switch from [...])

There is also some good news about the RV9 encoder speed on new hyper-threaded CPUs from Intel, but more about that later.

[Sirber]

Any news about AMD 3DNOW implementation? ;)



I can't wait to try it :D

[karl_lillevold]

No specific AMD improvements, but since AMD without SSE2 is slower in EHQ mode than a P4, this improvement should be very useful:

RV9-EHQ with fast 1st pass : http://forum.doom9.org/showthread.php?s=&threadid=66132

[Sirber]

Hi

the splitter from the first pass isn't closed, so there is 2 splitter loaded in the second pass. :scared:

[edit]

With RealAnime, I can't encode 2-3 OGM in a row. Producer IDLE coz old splitter are still active.

[Sirber]

producer fail to encode a file with "&" in it's name :rolleyes:

[edit]

about the last OGM bug, I'll use AVISynth so everything will be ok.

[slavickas]

Originally posted by Sirber
producer fail to encode a file with "&" in it's name :rolleyes:



i think it's present all the time, i noticed and forgot it about year ago

[Sirber]

@Karl

Could it be possible to have settings for:

1) Force "keep high freq"
2) Force minimal bitrate

Thanks!!! :D

[karl_lillevold]

@Sirber: that's a good suggestion... I will maybe not add exactly those settings, but perhaps a new "high bitrate" mode, where the codec's adaptivity is less aggressive, specifically for B frames, and choice of quantizers. This adaptivity was previously tuned for medium to low bitrates.

[Sirber]

128 Kbps Stereo Music - RealAudio 10 raac 0 44.1 kHz RealPlayer 8
64 Kbps Stereo Music - RealAudio 10 raac 0 44.1 kHz RealOne Player

Might be a bug, cause Codec Flavor is the same for 2 different setting...

[damrod]

Originally posted by Sirber
producer fail to encode a file with "&" in it's name

i think it's normal since producer.exe is a cmdline soft no?

when i launche producer i use " in my cmdline

"c:\producer\producer.exe" -i "c:\test;avs" -o "c:\test.rmvb" ....

[Sirber]

nope, RealAnime use jobfile since 1.2x IIRC. The problem still exist...

[damrod]

oh yes so i know the problem

it's the format of jobfile which is the same of audience files....

XML !!!!!

some keywords to not be used in it : & for exemple

bad luck...lol

oh avec the accents...ok there's no accents in english...it's not my fault if it's a limited language ;-)

[Sirber]

the easiest fix would be to rename files prior encoding. hum... something I could do :D

[Lagoon]

Even better, rename file to a generic name during encoding, and give back its original name after the encode :D

Just a thought, I don't know at all if it's hard to code :confused:

[damrod]

good idea

btw i add the possibility to demux ra10 (aac inside) to aac file

just need mp4creator to create ipod m4a file....

[Wilbert]

I read somewhere the following line of text:

real has given us (*) some funding and is working on integrating the vorbis and theora in their Helix player.
(*) refers to the xiph people.

I'm a bit puzzled about the meaning of this. Does that mean that it will be possible to mux rv and ogg, and play it?

[damrod]

you can encode ogg with producer....

[kilg0r3]

Is it just me or ist it still impossible to get to the binaries and the producer documentation on helixcommunity.org? I'd like to read up on the 1 pass quality mode but I don't have a producer distor on the box.

[damrod]

see rv10 elysium thread...:-(((

sniff sniff

[kilg0r3]

Hi!

I am backing up some of my dvds on dvd-r. So i don't require filesize prediction for all of them. since it suffices to restrict the fielsize of the last file on a disc.

What I want to do with the producer should be roughly equivalent to a single pass constant quant 2 (standard mpeg matrix) encode with some b-frames in xvid etc. The most similar encoding mode in producer seems to be "constant quality", which appears e.g. in AutoRV9 and Realbatch etc. However, is not present in the producer docs. So how would I do it? And, which setting would be _roughly_ equivalent to the settings mentioned above in xvid?

TIA, kilg0r3.

PS. Do you never close your computer case?