CINEMA CRAFT Encoder SP3

[Lyris]

Good grief, this needs work.

I'm encoding a movie right now which is about 3 hours long and has to fit on one DVD9. So I'm spending a lot of time on getting the best settings for the problematic parts. Has anyone else had these issues?

1. Black screen in source window. I don't know what causes it, but it happens sometimes. I have to use VirtualDub to find the in/out frames and refer to the frame number.

2. This is the worst one: sometimes the encoder DOESN'T APPLY the settings you specify. I made a bunch of different outputs of one scene using differnet Residual/Luma settings, making sure to change the "Picture" type in the list, then encoding the file. Ran them through SSIM comparison, and it's encoded the EXACT SAME THING on all files, ignoring my "Picture" setting. FACEPALM.

Edit: I've found out that this happens when you "trim" the source video, that is, when encoding does not start from frame 0 of the source. Using AviSynth to trim right now as a workaround. Will report to CTC when I get a spare second.

3. With 23.976fps source, "Bit Allocation" window doesn't show the correct part of the video at the correct time. Again I have to use VirtualDub to find the right point.

[Lyris]

4. Sometimes in the Bit Allocation window, the slider can't be moved past the first "screen", meaning that unless you want to change settings at the very start of the video, you need to close and try again. Typing the timecode into the "Move to" box does not fix it, the position stays stuck.

5. Sometimes when you reload a project file which is NTSC 23.976fps, the UI resets the frame rate to 29.97. This doesn't seem to have much practical limitation if the VAF file has already been created though.

Just keeping a log before I send a report!

[Lyris]

6. Segment re-encode sometimes seems to insert a strangely interlaced version of the scene. I've not verified the result on a television, but the entire movie is 23.976fps with pulldown flags.

7. It's just this minute crashed when clicking the viewer window (film strip icon).

8. It refuses to import this AVISynth script (which loads fine in VirtualDub, etc):
part1=avisource("june_part1.avi")
part2=avisource("june_part2_no-jaundice.avi")
part3=avisource("june_part3.avi").trim(0,72536)
part4=avisource("snowscene_insert.avi")
part5=avisource("june_part3.avi").trim(72790,163826)
part6=avisource("june_part4_CREDITS.avi").trim(0,2218)
return part1+part2+part3+part4+part5+part6


It works fine with other AVISynth scripts with TRIM functions though? :confused:

I really hope they fix this GUI. I'm not getting any of the increased productivity they promised.

[Lyris]

Major new release posted on CTC's web site for those who missed it.

#3 appears to have been fixed and "Shutdown when finished" added. Looks like they are paying attention.

Thanks CTC!

[kolak]

Major new release posted on CTC's web site for those who missed it.

#3 appears to have been fixed and "Shutdown when finished" added. Looks like they are paying attention.

Thanks CTC!

More to come :)
The only one encoder, which has not being tested against SP3 is Toshiba- none of others, industry standard encoders can hold on with SP3 for film sources.

[nimrodim]

I am utterly frustrated with this.
There seems to be memory issues with this version.
I am using windows 7 (64 bit) with 4 gigs and a q6600 at 3.6 Ghz.
I have encountered numerous problems:
1. Works fine on small AVI and Avisynth frame served stuff. - but when I try to stream any large videos the program crashes at the end of the pass complaining about "not enough storage is available to process this command".
2. Tried to split the process with larger files into 2: creation of a vaf file and then another pass. - But the vaf files comes out incomplete...and video encoding on the second pass stops at random points.
3. This version really hogs a lot of memory and drives CPU usage on all cores much higher compared to sp2. (30-50% compared to 60-90%)
4. Program crashes if you try to enlarge the progress window during encoding.
5. Tried the 1 Pass VBR encoding on Avisynth frame served video - again process stops randomly at different points in the video.

Not happy with this....
:(:(:(

[kolak]

I am utterly frustrated with this.
There seems to be memory issues with this version.
I am using windows 7 (64 bit) with 4 gigs and a q6600 at 3.6 Ghz.
I have encountered numerous problems:
1. Works fine on small AVI and Avisynth frame served stuff. - but when I try to stream any large videos the program crashes at the end of the pass complaining about "not enough storage is available to process this command".
2. Tried to split the process with larger files into 2: creation of a vaf file and then another pass. - But the vaf files comes out incomplete...and video encoding on the second pass stops at random points.
3. This version really hogs a lot of memory and drives CPU usage on all cores much higher compared to sp2. (30-50% compared to 60-90%)
4. Program crashes if you try to enlarge the progress window during encoding.
5. Tried the 1 Pass VBR encoding on Avisynth frame served video - again process stops randomly at different points in the video.

Not happy with this....
:(:(:(

For the beginning stop using Avisynth- render your script to YUY2 and feed CC-SP3 with AVI.
If you have problems report to CTC- they will fix it.


Andrew

[nimrodim]

Just wanted to add that here is no option for setting a drop frame flag - as in the previous version of cinemacraft.

[Koutsoubos]

For the beginning stop using Avisynth- render your script to YUY2 and feed CC-SP3 with AVI.
If you have problems report to CTC- they will fix it.
,


I almost shelled out money on impulse for this beta release program. Luckily I was playing around with it on an associate's computer. Now I won't even consider it until they get all bugs sorted out.

Do NOT feed your CCE-SP3 with YUY2 AVI files unless the levels have already been corrected to 16-235. If the levels are 0-255, the program does not correct this (since it only does so on the internal RGB-->YUY2 conversion)

Really, I have not seen anything quite this broken in a while.

[kolak]

,


I almost shelled out money on impulse for this beta release program. Luckily I was playing around with it on an associate's computer. Now I won't even consider it until they get all bugs sorted out.

Do NOT feed your CCE-SP3 with YUY2 AVI files unless the levels have already been corrected to 16-235. If the levels are 0-255, the program does not correct this (since it only does so on the internal RGB-->YUY2 conversion)

Really, I have not seen anything quite this broken in a while.

Who feeds YUY2 with RGB levels?- bit against whole logic of formats.
Also did you change setting to 0-255 when you done it?
YUY2 is the best format to feed SP encoders.
SP3 has unmatched quality for film sources, after you pass few GUI bugs.


Andrew

[Koutsoubos]

Who feeds YUY2 with RGB levels?- bit against whole logic of formats.


This happens if all one does is to do a straight RGB-->YUY2 conversion without squashing the luma, assuming (wrongly) that CCE would take care of it. CCE won't do that (wasn't designed to) and there will be considerable problems down the line.

The bugs are quite substantial. While I have no doubt that the encoding will be good when it works (with chapter customizations and frameserving and all that), right now I have to agree with Lyris' current assessment that there is no "increased productivity" to be gained.

[kolak]

This happens if all one does is to do a straight RGB-->YUY2 conversion without squashing the luma, assuming (wrongly) that CCE would take care of it. CCE won't do that (wasn't designed to) and there will be considerable problems down the line.

The bugs are quite substantial. While I have no doubt that the encoding will be good when it works (with chapter customizations and frameserving and all that), right now I have to agree with Lyris' current assessment that there is no "increased productivity" to be gained.

SP3 has much betetr quality than SP/SP2 for film sources and if you don't do 10 encodes a day bugs are not that problematic. GUI is not the best, I also prefer old one, but quality is unmatched.

Andrew

[kolak]

This happens if all one does is to do a straight RGB-->YUY2 conversion without squashing the luma, assuming (wrongly) that CCE would take care of it. CCE won't do that (wasn't designed to) and there will be considerable problems down the line.

The bugs are quite substantial. While I have no doubt that the encoding will be good when it works (with chapter customizations and frameserving and all that), right now I have to agree with Lyris' current assessment that there is no "increased productivity" to be gained.

This is more user problem/error than CCE issue.

[Lyris]

The new version is a huge step forward, that's for sure. I'm much happier now.

[mp3dom]

How can I set a progressive encoding? Even if I set Picture Type = Progressive (In Window->Picture...) the resulting encoding flag is always "Interlaced".

[bigotti5]

Create a preset with progressive setting in Window->Picture and save it.
Double click in main window - marker window appears - mark Picture - select your preset (as shown in my picture)
encode

http://members.aon.at/video.digital/sp3.png

[mp3dom]

Thank you very much! So all settings available in the Window menu (Segment/Picture/Effect) are mainly saved presets that need to be set in the segment window right? Thanks again!

Edit: Tried your suggestion but doesn't work as expected! Seems that Frame 0 is always ignored because it keeps flagged as interlaced! You can see this with ReStream. Could you confirm that?
A workaround could be to set frame 1 as progressive so from frame 1 to the end it would stay correctly flagged as progressive.

[hank315]

Edit: Tried your suggestion but doesn't work as expected! Seems that Frame 0 is always ignored because it keeps flagged as interlaced! You can see this with ReStream. Could you confirm that?
A workaround could be to set frame 1 as progressive so from frame 1 to the end it would stay correctly flagged as progressive.
This seems to be an old trick just to fool a (HW) decoder. A lot of PAL DVD's have the first frame or first GOP encoded as interlaced, the rest is encoded as progressive.

I'm running tests with SP3 but for now it's impossible to change intra/inter matrices, I've read the manual, tried a lot of settings but it refuses to use the matrices as set in the Window/Segment settings.

[mp3dom]

I haven't tried it, but I think that the encoder change your preferred matrix every GOP if necessary. For example I started with standard MPEG matrix but the resulting file had a FOX-similar matrix at the end. Have you tried to set the matrix as 'fixed' and 'use the same matrix on the entire stream'? (maybe it's a bug related to frame 0 being ignored by the settings, you could try to change matrix from frame 1 :))

[hank315]

Yep, I set it to "fixed" and use "throughout the entire stream".
I had natural 2 matrix set but what ever I try it always uses the natural 1 matrix (the standard CCE matrix).

[mp3dom]

Weird, but the GUI seems a bit bugged at the moment.

[kolak]

Yep, I set it to "fixed" and use "throughout the entire stream".
I had natural 2 matrix set but what ever I try it always uses the natural 1 matrix (the standard CCE matrix).

Have you set it in main window/ segment window as bigotti5 showes?

You have to create in Segment Settings custom presset: choose your matrix and in Quant Matrixes options choose Fixed. Save with some name.

In the main window apply this presset for entire source.
I've tried and does work fine- it keeps matrix which I have chosen for whole file.

Andrew

[mp3dom]

Kolak, can you confirm my 'problem' regarding progressive encoding? Maybe I'm doing something wrong...

[kolak]

Kolak, can you confirm my 'problem' regarding progressive encoding? Maybe I'm doing something wrong...

What is your problem?
Progressive sequence is not set to 1?
Looks like CTC encoders don't do it and I never had problems with it.
Frame Structure= Frame, Frame Type= Progressive- should be enough. Many DVD players ignore flags anyway and do its own picture analyse and make decision based on this- not flags.

Andrew

[mp3dom]

No, it doesn't flag 'Frametype progressive' so the stream appears always 'Interlaced' even if I force progressive encoding. At this point I don't know if internally SP3 is calculating motion vectors as interlaced or progressive.
The only way is to set a new segment from frame 1 and apply the progressive preset. At that point starting from frame 1, the progressive frame is set (DGIndex and other tools detect the stream as progressive).
If I set the progressive preset from frame 0 (without others segments) the stream will be detected as interlaced on the whole stream (you can see it clearly with DGIndex and other tools that will show always interlaced TFF).
So it seems that preset at frame 0 is ignored (at least regarding progressive/interlaced encoding)

[kolak]

No, it doesn't flag 'Frametype progressive' so the stream appears always 'Interlaced' even if I force progressive encoding. At this point I don't know if internally SP3 is calculating motion vectors as interlaced or progressive.
The only way is to set a new segment from frame 1 and apply the progressive preset. At that point starting from frame 1, the progressive frame is set (DGIndex and other tools detect the stream as progressive).
If I set the progressive preset from frame 0 (without others segments) the stream will be detected as interlaced on the whole stream (you can see it clearly with DGIndex and other tools that will show always interlaced TFF).
So it seems that preset at frame 0 is ignored (at least regarding progressive/interlaced encoding)

Hmm- I've just done test. I have only one segment- whole movie (so 2 markers: 0 and end of the movie). It does work fine in this case. I also set it for the last one- probably not needed. I remember also doing some encode, where I had many chapters and it was also fine. Make sure you set Progressive not Auto detect (of course if your source is 100% progressive).

What is your stream timecode set to? 00:00:00:00 or other?
Do you have latest version ?

Andrew

[mp3dom]

Yes, same scenario here, with Progressive set (not Autodetect) but shows this strange behaviour (at least with latest version, TC set to 00:00:00:00 on both source/mpeg settings)

[kolak]

Yes, same scenario here, with Progressive set (not Autodetect) but shows this strange behaviour (at least with latest version)

No way- I've just tried few times- all fine :)
Send me your project file.
Works fine all the time for me. Do you use AVI or avisynth as a source?

Andrew

[Lyris]

Question - has anyone ever had a huge amount of success in improving difficult scenes by changing Quantize Matrix parameters? Every time I've done comparative tests, the default settings have proven to be the best.

[kolak]

Question - has anyone ever had a huge amount of success in improving difficult scenes by changing Quantize Matrix parameters? Every time I've done comparative tests, the default settings have proven to be the best.

SP/SP2/SP3 will find the best matrix for you.


Andrew

[kolak]

It looks like Progressive sequence=1, should not be set. I've checked stream from Harry Poter, which was done with Toshiba reference encoder and it has the same flagging as CTC encoders.


Andrew

[Lyris]

Glad it's not just me then :)

[hank315]

Choosing another matrix seems impossible, if someone has figured that out, please tell.
A lot of GUI settings are ignored, also random "out of memory" errors (8GB installed using WIN7-64bit).
For those who actually use this for creating DVD content, better check the generated stream to see if your settings really have been applied.

IMO CCE-SP3 isn't ready for release, too many bugs.

[kolak]

Choosing another matrix seems impossible, if someone has figured that out, please tell.
A lot of GUI settings are ignored, also random "out of memory" errors (8GB installed using WIN7-64bit).
For those who actually use this for creating DVD content, better check the generated stream to see if your settings really have been applied.

IMO CCE-SP3 isn't ready for release, too many bugs.

Yep, I set it to "fixed" and use "throughout the entire stream".
I had natural 2 matrix set but what ever I try it always uses the natural 1 matrix (the standard CCE matrix).

Have you set it in main window/ segment window as bigotti5 showes?

You have to create in Segment Settings custom presset: choose your matrix and in Quant Matrixes options choose Fixed. Save with some name.

In the main window apply this presset for entire source.
I've tried and does work fine- it keeps matrix which I have chosen for whole file.



Have you seen my response?
All works fine. What version do you have? Do you have all required 3rd party software?
There are some GUI bugs, but never had any errors during encode- also Win 7 64bit with 8GB of RAM.

IF you have problem contact CTC. IF you're using cracked version- don't complain :)

Andrew

[mp3dom]

What settings are ignored? Just to know and check. I've never had any out of memory error, but I'm working on episodes so no more than 30min of continuous video.
Besides my problem of 'progressive flag' being ignored (seems that I've found a fix for that), other settings seems fine. Only a doubt about the change from Normal to Active in complex areas... both seems to produce almost the same quality but I cannot check for sure (the final output is visually the same as Xtream)

[kolak]

What settings are ignored? Just to know and check. I've never had any out of memory error, but I'm working on episodes so no more than 30min of continuous video.
Besides my problem of 'progressive flag' being ignored (seems that I've found a fix for that), other settings seems fine. Only a doubt about the change from Normal to Active in complex areas... both seems to produce almost the same quality but I cannot check for sure.

If something is ignored report to CTC and try to use AVI, not Avisynth. Avisynth support seams to be most affected.

I have not found any setting, which is ignored, but I use AVI as a source.

Andrew

[mp3dom]

Yes, using Canopus CLLC seems fine.

[kolak]

Yes, using Canopus CLLC seems fine.

It should, because Canopus Lossless decoder outputs YUY2 (and it's very fast), which is the best input format for SP3.

Again- please report any bugs, requests to CTC. They will fix it or add new features if they get official report/complain.


Andrew

[hank315]

Functions like "Close all GOPs" and "fix GOP structure" don't seem to work if you do multipass without deleting the existing vaf file, previous versions could do this automatically...
Progressive/interlaced settings work OK here.
I also found the solution for the matrix settings, there's a difference between
"use the same ones throughout the entire stream" and
"use the same ones throughout the same segment".
The first one doesn't work if you have only one segment, it will always choose natural 1 matrix, "use the same ones throughout the same segment" works OK when choosing a custom preset.

For most setting changes you have to delete the vaf file.
About input, I always use Avisynth scripts, according to the manual frameserving (Avisynth, Virtualdub) is supported so there should be no need for direct AVI input.

IF you have problem contact CTC. IF you're using cracked version- don't complain :)
LOL, I've been asked by a small video company to check their encodes because they experience inconsistent results using this new SP3 release so for me it's just testing, they can do the complaining :). Their version is 1.0.3.1.

First impression is that CCE-SP3 has a lot of nice features but apart from the GUI bugs some weak points are still not solved, still too noisy and the motion estimation needs some work.

[kolak]

Functions like "Close all GOPs" and "fix GOP structure" don't seem to work if you do multipass without deleting the existing vaf file, previous versions could do this automatically...

These are quite big changes and I think SP3 behaves properly in this case, but worth to note.
Closed and Fixed GOPs should be used only for multiangle encoding or when there is specific need. Adaptive and open GOP help a lot for quality.


Progressive/interlaced settings work OK here.
I also found the solution for the matrix settings, there's a difference between
"use the same ones throughout the entire stream" and
"use the same ones throughout the same segment".
The first one doesn't work if you have only one segment, it will always choose natural 1 matrix, "use the same ones throughout the same segment" works OK when choosing a custom preset.

I will pass it to CTC.

I use 3rd option as there were no reported playback problems and it does improve quality.


For most setting changes you have to delete the vaf file.

This is correct behaviour. Manual should be updated to describe, when you don't and when do need to re-create vaf file (as with SP version)


About input, I always use Avisynth scripts, according to the manual frameserving (Avisynth, Virtualdub) is supported so there should be no need for direct AVI input.

Yes- but I'm not sure if CTC is aware how complicated avisynth scripts can be and this can cause stability problems.
Does it have problems with simple scripts also?


LOL, I've been asked by a small video company to check their encodes because they experience inconsistent results using this new SP3 release so for me it's just testing, they can do the complaining :). Their version is 1.0.3.1.

Please tell them to report all findings- CTC does listen.


First impression is that CCE-SP3 has a lot of nice features but apart from the GUI bugs some weak points are still not solved, still too noisy and the motion estimation needs some work.

Engine is brilliant for films sources, but not for interlaced. It looks like motion estimation algorithm could be improved. I dont have real problem with noise, but I don't use default settings. Some settings are very sensitive and you have to find correct for specific source. Once you do it quality is amazing. It's not ProCoder- just press encode. SP3 is not for everyone- you need some knowledge and experience to use it. You would (or not) be surprised how many companies do encoding :)

There is need for big GUI fixes and I hope it will come soon.

Andrew

[mp3dom]

I dont have real problem with noise, but I don't use default settings.

For example? What settings are you using or, in general, are you fine-tunings? Thanks.

[kolak]

For example? What settings are you using or, in general, are you fine-tunings? Thanks.

Depends on the source. For typical film source I have very good quality even at 4Mbit average. Never seen any sharper encoder or better preserving grain. It does need tweaking for each source.

I have my procedure to find good settings, but confidential :) All based on CTC manual and observing encoded files.


Andrew

[Lyris]

I always make sure to turn the Lowpass Filter (on by default) OFF. Hate, hate, hate lowpass filtered video. Then again, I usually have a full DVD9 to play with.

CCE SP3 is such good quality that the LPF is not needed at a high bit rate - never in my experience anyway.

Also, the presets are set up for Interlaced, so I have to change that for Film material.

[mp3dom]

LPF settings on SP3 are weaker than SP/SP2 standard (40 vs. 32). It becomes visible only when you're watching your video at a higher resolution (SD material on HD monitor or similar). At plain SD is quite invisible. Anyway it helps to keep mosquito quet especially on noisy animation when noise is already visible near sharpen edges. At high bitrates and really complex material SP3 produce visible artefacts like all other MPEG2 encoders but I'm speaking of complex CG/animation, not film with actors where surely the motion is more predictive.
Regarding preset setup for interlaced, do you mean resize filters and so on? Or even the quantizer matrix/LPF are setup for interlaced? In that case setting up the "Picture..." tab to progressive resolve the problem?

[Biggiesized]

Question - has anyone ever had a huge amount of success in improving difficult scenes by changing Quantize Matrix parameters? Every time I've done comparative tests, the default settings have proven to be the best.

I thought SP3 and SP2 share the same quantization matrices. I've opened them both up side by side and it appears true.

[name123]

i really like the new sp3, quality is outstanding, it is perfect the way it is, the only problem is that they do not support avs directly with video sources being over 2 gbs, it would cause CCE to freeze on 1st pass or show not enough storage, but there is a work around it (feed your avs to virtual dub then frameserve it) the Virtual multipass is one of the most superb update, after the intial pass, the following pass becomes faster and consumes less time.

[mp3dom]

Some day ago I've used an AVS file that point to a 10GB AVI without any problem.
Anyway, regarding my "issue" of Progressive flag being ignored, I've found the method to replicate (at least on my system).
If you use the progressive preset at the beginning (starting from frame 0) without using any other preset in the middle then all is ok (progressive flag is set). But if you set also other points where you change something, then the first progressive preset is ignored (thus not enabling the progressive flag).
For example made 2 preset: Progressive Default and Progressive Active (the same as default but with Active Motion Estimation). Then assign at frame 0 the Progressive Default and at frame X the Progressive Active. You'll see that prior of the 'Progressive Active' preset, the frames will be flagged/encoded as interlaced, and will switch to progressive just at the beginning of Progressive Active.

[kolak]

Some day ago I've used an AVS file that point to a 10GB AVI without any problem.
Anyway, regarding my "issue" of Progressive flag being ignored, I've found the method to replicate (at least on my system).
If you use the progressive preset at the beginning (starting from frame 0) without using any other preset in the middle then all is ok (progressive flag is set). But if you set also other points where you change something, then the first progressive preset is ignored (thus not enabling the progressive flag).
For example made 2 preset: Progressive Default and Progressive Active (the same as default but with Active Motion Estimation). Then assign at frame 0 the Progressive Default and at frame X the Progressive Active. You'll see that prior of the 'Progressive Active' preset, the frames will be flagged/encoded as interlaced, and will switch to progressive just at the beginning of Progressive Active.

Is this from avisynth script a sa source? Try with AVI.
It looks like ti's only issue with avisynth as a source.

[Video Dude]

CINEMA CRAFT Encoder SP3 1.0.4.1

[Fix] Sometimes two-pass inverse 3:2 pulldown failed because i32 file could not be opened. This problem has been fixed.
[Fix] In some QuickTime files, frame number was counted one frame longer. This problem has been fixed.
[Fix] Fixed other minor problems.
[Change] When both resizing and deinterlacing are specified, deinterlacing will be performed first from this version.

[kolak]

CINEMA CRAFT Encoder SP3 1.0.4.1

Problem with "use the same ones throughout the entire stream" has been also fixed.


Andrew