crop and resize in Gordian 0.28

[folio]

Ive got a query regarding the crop and resize of Gordian 0.28...

when i open my d2v file in Gordian after cropping and resizing, compressing test, i get a resolution of 576x320 for a 1hr 30 min film (1 CD encode) 931kbits.
and 0.202 in the bits/frame - 0.48 next to load button and 0.420 compressibilty test, the aspect error is -0.1%. Now ive always used around a bits/pixel frame value of 0.17 for my 1 cd encodes as stipulated in the Doom guide for Gordian...so when adjusting the slider to around 0.17 the closest it gives me is 0.182, with 43.4 next to load button and 0.42.0 compressibility
resolution of 608x336 with aspect ratio of 1.810....i run another compressibility test and heres what i get :

bits/pixel frame value of 0.182
load button 41.9
compressibility test 0.435

now it says in the guide that a value between 65 and 45 % should be used...im not too sure what to do...? if i move the slider to get a higher value than 41.9 i increase the bits pixel frame thus making it harder to create a 1 cd encode!?!

im using the Lanczos resizer as it was recommmended...if encode with the value of 41.9 i might not get great quality? Could anyone advise me on this? i used to use Gordian 0.27 and never had these problems at all ....always had a value of of aorund 60% and had a resolution of 688x288 depending on the source and aspect ratio of film etc...but never had to go below 544xXXX resolution to get a great quality....thanks in advance..
folio

:confused: :(

[jggimi]

If you're anything like me ... (and I know I am) ... you'll use the b/p*f value as a guideline prior to your first compression test, only. And then, once the compression test results are in, use the test results, and ignore b/p*f values for all future tests and final encoding.

Lanczos resizing is perhaps the sharpest, but it is also likely to be the least compressible. You might like the look of sharp bicubic or neutral bicubic, which, while they have somewhat softer resizing, may provide more compressibility for you.

There are many filters that may improve compressibility, some at the cost of the video image, of course. You can spend a lot of time experimenting with filters; some may produce results you like, others may not.

And you may find that sometimes, a 1-CD encode is not the best solution, and that only more bits will solve a particular "quality .vs compression" problem.

In addition, you may find you like the results at 40% compression; as "quality" is highly subjective and content dependant.

You don't have to encode the entire contents to test things. You can use the Preview button to see the results of filters (including cropping and resizing), you can encode a small D2V project by using DVD2AVI's [] buttons, or you can encode small sections of the existing project with the Trim filter in your .avs script.

[folio]

cheers for the advice will do some testing...reckon that some 1 cd encodes look the biz but then it al depends upon the content and how it reacts to being encoded...thanks again
folio;)

[folio]

got a another hitch to report! after creating the d2v i get a red filled box next to the load button!? even after the compress test, it wont go even when i reduce the output resolution with the slider , it wont go. In fact it only goes when i have the following valuse in the bit/frame and compress boxes
bit/frame : 0.132
load button : 99.5 !
compress: 0.133
these values just dont mae sense?! 0.132 is too low for sure no?!

The film is only 1hr 33 min long? so its not a case of it being a bad compress for 1 cd. I made a Divx3 copy with Vidomi and its great...
i'm not an amateur at all, i know that normally when you get the red filled boxes it because the film is way too long to put on a 1 cd encode...as the resolution does not permit it...but this is 1hr 33min (16:9) film of which ive been able to encode with simple encoders like Vidomi!!? id aprreciate any help..got me confused here?! ive created the d2v file 3 times now with the same outcome!
thanks again...
folio

[hakko504]

Now you're on the other end of the scale, compression test has given you a too high value (99.5%) and you need to increase the resolution and or use sharper resize/less filters.

[folio]

will increasse resolution and see where it takes me...was using the neutral bicubic so will try lanczos, though i did the test without filters as well...
thanks for the tip....

[folio]

Quote "Now you're on the other end of the scale, compression test has given you a too high value (99.5%)
and you need to increase the resolution and or use sharper resize/less filters"

Im not sure what is happening but it seems Gordian 0.28 wont let me achieve a value decent enough to start encoding? as stated in the
issue i had before, ive increased the resolution right till 720 and im still getting a compress value of 78 and boxes filling with red!
ive used the Lanczos resizer
and not used filters at all. Its strange ive done loads of encodes with Gordian and its happened just with the last two films
one was interlaced and the other progressive?! Ive read posts on the forums and came across a post which came to the conclusion
it had to be a 2 cd encode..but this is not the case...its a 1hr 33min film!

Ive just done another compress test and its given me a resolution of 416x224 with a value of 64.2% of 0.583 compress test, though
its now got 0.378 as the value for the bits/pixel frame...aspect ratio : 1.857 and aspect error of -2.8%

surely after the compress test a value of 65% is appropriate for a good encode?
its not making sense at all...ive always encoded with a value of between
17 and 20 in the bits/pixel frame box after the compress test. and ive always had decent resolutions...544x304,576x320 (16:9) and
512x384,4:3 I know that
not one film wil encode in the same way but
ive never had to go into resolution that went higher than 640xXXX...im perplexed is that ive not got gordian 0.28
configured properly?...am i dummy or what?? or maybe its one of those weeks for me..?! any help appreciated
ive read thru the forum posts by the way...!


:confused: :confused: :confused:

[jggimi]

Try following my suggestions, which I've already made (mostly): First, take a deep breath, try to calm down.

Prior to your very first compression test, use the b/p*f value to set an initial resolution and size. Note that the bitrate doesn't change as you change resolutions. When you have a usable resolution and usable b/p*f, execute your first compression test.

Ignore b/p*f values after your first compression test. Use the % factors only.

Using only the % results from the compression test, set a new resolution, and possibly, a new size.

Any time you make a significant filter change, such as type of resizing or noise reduction, or a significant change in resolution or bitrate ... do another compression test.

To test your video filtering prior to any encoding, use the Preview button.

To run a test encoding of a small section of video, edit the .avs, selecting a small set of frames with the AviSynth Trim filter. For example:Trim(10001,11000)will produce a 1000 frame video, starting at frame 10001. You can either edit the .avs manually, then use the Add Job button on the Encoder tab, or you can use the Edit button on the Save & Encode window. Note: obviously, don't bother muxing any audio in this test encoding job.

[folio]

hey jggimi...i am calm now by the way...its just i never had these issues with Gordian 0.27 and divx pro 5.0.2...i also know that no two films encode the same...I did as you said and came up with a resolution of 448x256, it came out fine though there were discernible traces of heavy edge enhancement which might have something to do with using the Lanczos resizer,In oyur opinion do you think that the Lanczos is a good resizer to use for short 1hr 30 min films? or would it overcompress too much for a short film?
maybe it'd be better to use the sharp or soft bicubic in this case as the edge enhancement was pretty heavy to say the least, maybe playing with a few filters would help as well..anyways thanks for the helping and calming hand!
cheers:D

[jggimi]

I will always start with the Lanczos resizing filter, then move to sharp or neutral bicubic as needed. Edge enhancement is a difficult issue to deal with; I've sometimes had good luck using the Convolution3D filter.

Experimentation is key -- I really like previewing my filters before encoding, or only encoding a small portion of the content -- because I can make many iterative tests very quickly.

We're just lucky that this is an art more than a science. It makes each encoding a new experience. :rolleyes: