RealVideo 9 Information Thread Discussion [Archive] - Page 9

karl_lillevold

27th June 2003, 19:40

i tried to, at first it did not work, because I was missing the SMIL2 tag. The code below works (change the audio offset), but probably doubles decode complexity, since it may decode the video twice. If someone can think of another way to achieve this, speak out :)

<smil xmlns="http://www.w3.org/2001/SMIL20/Language"
xmlns:rn="http://features.real.com/2001/SMIL20/Extensions">
<head>
<layout>
<root-layout backgroundColor="black" width="704" height="288"/>
<region id="Video" >
<region id="VideoChannel1" soundLevel="100%" z-index="0"/>
<region id="VideoChannel2" soundLevel="0%" z-index="1"/>
</region>

</layout>
</head>
<body>
<par title="test">
<audio src = "clip.rmvb" region="VideoChannel1"/>
<video src = "clip.rmvb" region="VideoChannel2" begin="0.5s"/>
</par>
</body>
</smil>

RadicalEd

29th June 2003, 01:42

oh yes, your method works, but sirber's is still buggy for me :o :|

Sirber

29th June 2003, 01:46

What Karl did is a smile (:D)... :confused:

karl_lillevold

30th June 2003, 07:20

Re-wrote the beginning of the first post, to include a section on what I am doing in this forum, for three reasons: (1) to say thanks to Doom9 and the moderators for all the time spent running this site, which (2) has resulted in a forum with an incredible amount of useful discussions and ideas, and (3) to make sure everyone knows I do not want to sell anything to anyone. Yes, I do let you know when there is a new improvement, and about all the optimizations in RV9. Maybe that sounds like I am selling RV9.. So, maybe I am - it is free after all.. Otherwise, the old RealPlayer stigma would prevent anyone from trying what I know is worth trying, no..? Anyway, thanks to everyone who has given RV9-EHQ a fair evaluation. Based on current feedback, I don't think anyone has been disappointed.

kilg0r3

9th July 2003, 08:50

could somebody please tell me what the "file roll" options do? I read the docs but can't seem to understand what 'roll' means in this context.

Thanks

Dark-Cracker

9th July 2003, 12:01

normaly it's the prebuffering in milliseconds.
if i am wrong karl will surely correct me :)

++

karl_lillevold

9th July 2003, 15:57

i was confused too when I first ran across the 'roll' option in TFC's HPG, because it's an option I have never tried. I think this is the meant to roll over to new output RM files every so often (in terms of filesize or time), and intended for live encoding to file, or when archiving a live encode. This way too large RM files are avoided. The RM file format does not support larger than 4 GB compressed RM files (there are no problems with large source files to Producer on Win32).

Sirber

9th July 2003, 16:43

Wrong!!! :p I can't compress a 2.7 GB wave with Helix GUI :D It crashes and return to desktop.

karl_lillevold

9th July 2003, 17:12

right you are indeed! I have filed a problem report (https://helix-dna.helixcommunity.org/issues/show_bug.cgi?id=451).

EDIT: Hmm, I thought it would work in Avisynth though, with WavSource, but a quick experiment shows that Avisynth as well fails to open a WAV files larger than 2GB. That might be quite problematic for 6ch encoding.

EDIT2: See this thread (http://forum.doom9.org/showthread.php?s=&threadid=57243) in the Avisynth Usage forum.
Until fixed, for long multi-channel encodes, the only option in Producer is to use the method of separate audio and video inputs, audio being an AC3 track decoded on the fly to producer by ac3filter, video an Avisynth script.

EDIT3: or use HeadAC3 to convert AC3 to 6ch WAV. This seems to produce "good" +2GB WAV files that Avisynth can read.

justin

13th July 2003, 03:39

I have a question... how do you find out how to script your own rvmb profiles? I want to test out how some cartoons look, and I don't want to use Sirber's tool for making it easier, because I like things nice and complicated. I know that I am a nincombpoop :eek:

And also, is there any way to know for sure that you're encoding with EHQ, even if it's setup right, I keep wondering if it's not being used. This is my script, just one second... ah there it is:

<?xml version="1.0" encoding="UTF-8"?>
<audience xmlns="http://ns.real.com/tools/audience.2.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://ns.real.com/tools/audience.2.0 http://ns.real.com/tools/audience.2.0.xsd">
<avgBitrate type="uint">350000</avgBitrate>
<maxBitrate type="uint">760000</maxBitrate>
<streams>
<stream xsi:type="videoStream">
<pluginName type="string">rn-videocodec-realvideo</pluginName>
<codecName type="string">rv9</codecName>
<encodingType type="string">vbrQuality</encodingType>
<encodingType type="string">vbrBitrate</encodingType>
<quality type="uint">95</quality>
<codecProperties type="bag">
<encoderComplexity type="uint">80</encoderComplexity>
<customPacketSize type="uint">16000</customPacketSize>
</codecProperties>
<maxStartupLatency type="double">60</maxStartupLatency>
<maxFrameRate type="double">24</maxFrameRate>
<maxKeyFrameInterval type="double">20</maxKeyFrameInterval>
<enableLossProtection type="bool">false</enableLossProtection>
</stream>
<stream xsi:type="audioStream">
<pluginName type="string">rn-audiocodec-realaudio</pluginName>
<codecName type="string">cook</codecName>
<codecFlavor type="uint">25</codecFlavor>
<streamContext type="bag">
<presentationType type="string">audio-video</presentationType>
<audioMode type="string">music</audioMode>
</streamContext>
</stream>
</streams>
</audience>

ok, and also, when I did an encode test of an anime, it turned out to be 350kbps, but I want it to be on quality mode of 95% instead. So do any of you guys know what's happening out there?!

kilg0r3

13th July 2003, 09:39

personally, I use the rpad files generated by autorv9 and modify them using the two relevant documentation pages included with the producer.

karl_lillevold

13th July 2003, 17:49

Originally posted by justin
I have a question... how do you find out how to script your own rvmb profiles? I want to test out how some cartoons look, and I don't want to use Sirber's tool for making it easier, because I like things nice and complicated. I know that I am a nincombpoop :eek:

I like your sense for challenges. kilg0r3 made a good point. Start with an example, and read the documentation. Note the difference between an audience file and a job file. In job files everything is specified, incl inputs and outputs and pre-filters, and are ultimately the most flexible solution. Audience files are more limited and specify only the target audience, but simpler to get started with. For an advanced job file example with two sources, check this thread on RA Multi-channel (http://forum.doom9.org/showthread.php?s=&threadid=56684), and here's another one with the DropDupe prefilter (http://forum.doom9.org/showthread.php?s=&threadid=56564).

And also, is there any way to know for sure that you're encoding with EHQ, even if it's setup right, I keep wondering if it's not being used. This is my script, just one second... ah there it is:

Well, based on your audience it should be enabled. If you really want to make sure :) check that the erv4.dll that's in the codecs folder of the producer you are using says "RealVideo 9 EHQ" in the description when right clicking on it. Also, if you encode a short test, run with both encoderComplexity set to 65 and 80. With 80 it should take 3-4X as long to encode.

ok, and also, when I did an encode test of an anime, it turned out to be 350kbps, but I want it to be on quality mode of 95% instead. So do any of you guys know what's happening out there?!
That's because you have both vbrQuality and vbrBitrate in your audience. Remove the line with vbrBitrate and it should get you quality 95 (this should result in a pretty high bitrate, and you probably will not see much difference for a little lower quality, but experiments are a great way to learn). In AutoRV9's compressibility test, what is used for the reference quality is 87.

justin

14th July 2003, 20:27

thanks for the feedback, yeah. Just about everything was in the help file, but at least I understand the syntax completely now :) (XML certainly is straitforward)

Wow, yeah 95% is a hog! I scripted up a job file instead and tried it again at 65% and it still came out pretty good (minus any action scenes).

Ok, this might sound awkward, I hope I can explain this ok... at 65% you get really sharp quality stills, but any scenes with motion are all garbled up. At 80%, those high motion scenes become a little clearer, but the still scenes become sharper too. Is there any way to get 65% quality stills and have crisp motion scenes as well?

Without using 2 pass, I'm not sure if two pass helps in this situation, but I wouldn't like to encode in two pass because it takes twice as long.

karl_lillevold

14th July 2003, 20:34

for vbrQuality make sure you set the maxBitrate very high

<maxBitrate type="uint">99999999</maxBitrate>

Otherwise, the maxBitrate will limit the quality during high action scenes. This sounds like what you are seeing.

SeeMoreDigital

17th July 2003, 12:42

Karl

I apologise in advance if you have already been asked about this. But have you had chance to test the new version of this filter with HelProd 9.2?

Cheers

karl_lillevold

17th July 2003, 15:17

yes, I tested ac3filter 0.69b a few days ago, and it works fine as input source filter for AC3 to Producer, I don't know which change that fixed it. I edited the RA5.1 thread (http://forum.doom9.org/showthread.php?s=&threadid=56684), but did not post a reply, which is why you did not know there was new information.

trbarry

21st July 2003, 17:54

I'm afraid I'm guilty of pretty much ignoring RV9 so far.

But I'm curious. Is anyone aware of any tests with RV9/EHQ on highly detailed HDTV material, say at least 1280x720p, at maybe 5-6 mbps?

Sorry if this has already been answered here somewhere.

- Tom

Sirber

21st July 2003, 19:46

I tryed :)

Can't be played on my 2000+. I have framedrop at 5000kbps. You'll need a P4 3GHz.

SeeMoreDigital

21st July 2003, 19:57

Hello Tom,

I recently encoded the hi-def 1440x816 WMV9 Terminator trailer, you know the one that can be found on the M$ web site.

When I encoded it to DivX5 using Dr DivX at 6500kbps 1440x816 2pass VBR it looked pretty good but did not playback very smoothly.

However, when I encoded the same trailer again using RMV9 (not EHQ sorry) at 6500kbps 1440x816 2pass VBR it looked and played great.

God knows who much better it would look with EHQ. Shame I could not manage to get the HP+GUI application to see .wmv files.

Maybe I should try again - Hope this helps!

karl_lillevold

21st July 2003, 20:09

The clip Sirber tried to play back was an example I transcoded from one of the many wmv9 HD samples as well, the Coral Reef demo at 1280x720p. I encoded it at 5 Mbps with RV9-EHQ, while the wmv9 source is 7 Mbps.

This clip plays back fine on a dual 2.2 GHz system, but probably requires something like a 2.8 GHz single P4 to avoid codec CPU scalability. This is a really complex clip.

kilg0r3

21st July 2003, 20:25

when can we expect multi codec hardware decoders? :D

wasn't there something like a joint venture of ati and real building part of the decoding into radeon chips?

trbarry

21st July 2003, 22:19

Yes! Hardware accel for something other than MPEG-2 is really holding up our ability to make and use our own HDTV clips. It means we have to focus more on playability than compressibility. Though I suppose once we all have 3.5+ Ghz machines it won't matter as much.

- Tom

Sirber

21st July 2003, 22:25

Originally posted by kilg0r3
when can we expect multi codec hardware decoders? :D

wasn't there something like a joint venture of ati and real building part of the decoding into radeon chips? You can build a system with a tv-out / sound-out with ffdshow and MPC. You just need a mouse connected :)

SeeMoreDigital

21st July 2003, 23:35

Agreed,

Yes! Hardware accel for something other than MPEG-2 is really holding up our ability to make and use our own HDTV clips. It means we have to focus more on playability than compressibility. Though I suppose once we all have 3.5+ Ghz machines it won't matter as much.

I remember the days (1997/98) when my old 333MHz P2 PC, with 16MB ATI video card was not good enough to playback DVD's. So I went out and purchased a Creative DVD Encore 'hardware' card. It cost me a small fortune at the time!

It's quite unbelievable to think that even todays 3.06GHz P4's struggle with high def images.

RealMedia Video9 seems to be the best of the bunch so far with HD. However, I would like to give WMV9 VCM (.avi) a go with HD, as some of my standard encode tests with this version of codec proved to run more smoothly than the equivalent standard .wmv version.

Apparently MingCL is investigating!

Cheers

Sirber

22nd July 2003, 02:12

Think about H.264 for HDTV :D 12 GHz? :D :D :D

kilg0r3

22nd July 2003, 08:46

Originally posted by Sirber
You can build a system with a tv-out / sound-out with ffdshow and MPC. You just need a mouse connected :) currently doing just that :)
cpu: amd athlon 700; ram: 368mb; vga: ati rage pro; sound: hercules 4 channel (need only two though); dvd: liteon something; Mouse: Logitech wireless trackball!!:D

Valky

23rd July 2003, 09:52

Just a simple question: Is it possible to have dual language audio in rmvb or not?

Using forums 'search'-function I red all the 54 threads that came from 'dual language' but none of them doesn't seem to consider RealMedia.

I am asking this, because Dark-Cracker says in his Autorv9 advertise the following "support 2 audio streams".

Well, I tried his latest beta, but I couldn't find second source-input box anywhere? (And I was little bit disappointed anyhow, cause actually I tried to encode a simple movie with it, but it came without sound at all when using wav as input..which was ok and encoded with other great dark-cracker-tool 'Besweet_Gui_light', but thats unfortunately another story. Produced avs-file seemed to lack this audio-file completely, but I noticed this too late this morning after 8 hours of encoding...)

So if it's possible could it be done with avs-file if I wanna use RMFactory? (which is excellent software too, but usually crashes when adding second avs-file to joblist..with one input it works great)

My current avs file is this:

video=mpeg2source("E:\disney\mouses\mouses.d2v")
audio=wavsource("E:\disney\mouses\mouses.wav")
video=audiodub(video,audio)
video=AutoCrop(video,0,wMultOf=4, hMultOf=4)
video=TemporalCleaner(video,3,6)
video=undot(video)
video=unfilter(video,30,30)
video=bicubicresize(video,480,352,0,0.5)
video=mipsmooth(video)
video=limiter(video)
return(video)

kilg0r3

23rd July 2003, 10:45

muxing into rmvb does not work. you can do two audio streams with a smil file or mux rv9 and two audio streams in a matroska container.

Nic

23rd July 2003, 11:58

OT: Got round to finally installing and using RV9 last night. Its easy to install the RealOne player so its non-obstructive, so you can forget about it and use Media Player Classic. RV9's quality is excellent, at around 256kbps ive never seen anything better. I strongly recommend trying it for those, like me, who kept putting it off.

ps
Basically making it non obstructive is using the customized install option (rather than express), turn everything off, tell it you have no network connection. Quiting the install after its installed everything, (quit it as soon as it mentions anything about registering/passes), Going to the task manager, kill realsched.exe. Then goto C:\Program Files\Common Files\Real\Update_OB and remove the realsched.exe there. Also goto the registry and search for tkbell/realsched.exe in HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Run and remove it.

Dark-Cracker

23rd July 2003, 12:11

it's not possible for the moment to mux 2 audio stream in a .rmvb file the only solution is to encode the 2 audio track separately (.rm) and the video track (.rmvb) and use some .smil file to launch the video + desired audio track .

> but I couldn't find second source-input box anywhere.

in the audio panel select your file and click on "ADD" button , select your second audio file and click again on the add button. u will have a list with the name of the audio file.

Bye.

PS: feel free to post a msg in the respected threads of these tools for your bug report.

Sirber

23rd July 2003, 12:55

@D.C.

Could you do a RMVB-->MKV converter (like you did) but for commandline uses? I'd like to use it for future projects :)

Valky

23rd July 2003, 14:04

Originally posted by Nic
OT: Got round to finally installing and using RV9 last night. Its easy to install the RealOne player so its non-obstructive, so you can forget about it and use Media Player Classic. RV9's quality is excellent, at around 256kbps ive never seen anything better. I strongly recommend trying it for those, like me, who kept putting it off.

I am happy to hear this kind of feedback from such xvid-devoted guy like yourself. I am sure that this kind of feedback gives much more to whole encoding-community than those codec-against-codec competitions and their-codec-sucks threads cause there isn't just one best-of-the-best codec for every situations.
Very refreshing to hear this kind of stuff!

Ps. To DC: I think that concerning my whole previous post it served better community in this thread than just making a bug-report to your beta-tool, cause I am not even sure yet if it was my own mistake or not. I am investigating more... :)

karl_lillevold

23rd July 2003, 21:41

@Nic: thanks for taking the time! And for the excellent summary on installing RealOne and to stop Tinkerbell :p.
My summary (http://forum.doom9.org/showthread.php?s=&threadid=48179) is a little more awkward, and requires starting up RealOne to set some options, which is a good idea to make R1P behave nicely if you use it (for instance for streaming or DRM content).

Nic

23rd July 2003, 21:49

Never saw that post! :) Your post is alot more complete. I did alot of encoding in WMV8 for a client at 128&512kbps, compared them with a RV9 clip I made at the same bitrate of one of the same sources. RV9 was lots better (especially the 128kbps clip, res was only 352x288, so if it had been higher there may have been more of a difference at 512). Guess ill have to read up on your licensing, for my company's next transcoding venture. :)

Ill encode a 1 CD of Road To Perdition over the weekend in a few formats. Ill report back :)

Keep up the good work,
-Nic

ps
Talking to one of the guys at work about it today, his reply was "Aren't Real evil?" ;) You may really have to change the marketing of the player (easy to download, non-intrusive, etc) to change that perception :(

karl_lillevold

23rd July 2003, 22:05

@Nic: Re encoding: remember EHQ ;) Re marketing of player: I absolutely agree. Me and others have tried to change it, and the right people are taking notice, slowly...

SeeMoreDigital

23rd July 2003, 22:18

Cor blimey me!

Me thinks if Karl and Nic carry on much more like this we could see the beginings of a dedicated RMV9 section!

I must admit I was an ardent DivX man myself. But RealMedia is certainly comming up with the goods more quickly than the competition.

Great fun and great work!

Cheers

iwod

24th July 2003, 04:03

Now that real can be put into Matrosaka then the real player doesn't really matter. Once we can get Matroska into good shape.

However I hope RealONE 3 will finally elminate all ( not some ) of those annoying things like sign up etc....

But i know it is not the programming people that makes the decision. ( Karl )

Nic

24th July 2003, 10:23

Rob Lanphier <robla at real com> wrote:
What's not legal is to put RealAudio or RealVideo data in a Matroska container.

On the helix licensing mailing list that was written after a question by Christian. Christian's reply stated:
"does really not help us here to make rv2mkv available in any
form, as its clearly a license violation of the Real specs license"

So sadly maybe not. Maybe something will be resolved about it, I hope so. But otherwise we'll have to play Real in the Real container (and play it in MPC ;)), which kinda makes sense for now.

-Nic

ps
more promising results from rv9 even at reasonably high rates:
http://forum.doom9.org/showthread.php?s=&postid=348719#post348719

Sirber

24th July 2003, 12:45

It's illegal to store Real content in Matroska? Datas is only taken on the fly and redirected :)

karl_lillevold

24th July 2003, 15:35

thanks for pointing me to the other thread. I posted a reply there, regarding EHQ and the noisyEdgeFilter, which really needs to be disabled for PSNR measurements.

News: in Producer Milestone 5, there are several new <codecProperties> settings for tweakers. I will publish them shortly, but the first one is

<codecProperties type="bag">
...
<calcPSNR type="bool">true</calcPSNR>
...
</codecProperties>

This can be used to verify independently calculated PSNR numbers, with for instance DirectShowSource. DirectShowSource decoding could potentially be the victim of CPU scalability in the decoder, which may reduce PSNR, and this is a nice verification, since it is the encoder's accurate PSNR measurement, relative to its source. This setting also automatically disables for instance the noisyEdgeFilter. I will verify how well DirectShowSource PSNR measurements shortly.

p.s. there is no reason whatsoever for RV9 to excel only at low bitrates, even though it works really well in that range. For instance, take a look at Sagitairre's PSNR comparison (http://forum.doom9.org/showthread.php?s=&threadid=57687).

This is a long time topic of discussion, and I always get back to this. The ringing in MPEG-4, and the lack thereof if RV9, can provide a difference in sense of sharpness, depending on the viewing distance or screen resolution. On my 1600x1200 LCD screen, or our demo 42" plasma at work, I really can not stand ringing, but when I connect a laptop to my oldish 36" Sony TV, ringing is much less annoying, hardly noticable.

Re Matroska: I think the binary EULA that users who download binaries from Helix, and possibly also the binary EULA that users who install RealOne click through, both have language in them stating that these binaries shall not be used to transcode RealMedia to another format. This is what Rob Lanphier referred to. These EULAs were constructed with great caution long ago. I don't know the current status, but work is going on to try and loosen them up a little.

EDIT: to enable PSNR calc, the flag has to be 'true', not 'false' ...

kilg0r3

24th July 2003, 16:29

Putting RM content into matroska is not exactly transcoding the content, is it. :)

karl_lillevold

24th July 2003, 17:24

I just found out that the previous method to enable EHQ was accidentally removed in Producer Milestone 5. So, please DO NOT USE the erv4.dll from Milestone 5. Use erv4.dll from Milestone 4, or the one posted in this thread (http://forum.doom9.org/showthread.php?s=&threadid=58196), which also has more information about the new method to enable EHQ.

the new and "official" way to set encoderComplexity is to have an <encodingComplexity setting in the <videoStream> section of the audience.

Values:
very-fast, fast, normal, high, or very-high

Type:
string

Example:
<videoStream>
...
<encodingComplexity type="string">very-high</encodingComplexity>
...
</videoStream>

Note that this goes in the videoStream section as opposed to the previous encoderComplexity number, which was inside codecProperties (inside videoStream).

Sigmatador

4th August 2003, 18:01

is there any hope for a linux version of the real producer ?

karl_lillevold

4th August 2003, 18:13

producer for Linux is already available. Helix DNA Producer is released for Linux, OSX, and Win32. The Linux and OSX versions are a little limited due to filesize restrictions, and they can not read compressed AVI and MOV files. We are looking for contributors in those areas, since the Producer team is very resource limited.

Sigmatador

5th August 2003, 11:01

oops, i didn't check before posting ^^

but only uncompressed avi for input is a bit restrictive, i will looking for a linux frame server.

what's about this file size restriction ??

karl_lillevold

5th August 2003, 15:28

Originally posted by Sigmatador
what's about this file size restriction ??
Good question. It's probably relatively simple to resolve, and may in fact already been fixed (at least I can not seem to find any open issues), but I have been told even the Linux kernel did not handle > 4GB files too well until recently. And most native tools still don't handle such files. Input, research, and testing to find out what works or not would be very nice.