dvdshrink v3.2 is released

[ddlooping]

Glad to hear it, Robert, thanks. ;)

[motster]

@dvdshrink & team...

Awesome release! I love the output options! If you get a spare moment (ha!), can you have a pref. setting for setting Verify when using DVD Decrypter to burn?

The other feature I've been thinking about, but I know isn't easy to implement, would be to flag titles that are referenced in the First Play sequence. It stinks when you Still Image some titles, only to find that you'll be sitting through 4 minutes of "This video removed" because you didn't realize they're the stupid intro trailers ;-)

I know, I could figure out the titles myself using a player, but if he gets bored and wanted more features to add... ;-)

[ddlooping]

mb1, I must confess I did not test v3.2 with interlaced material, but with non-interlaced sources, DVD2One 1.5 was well behind in terms of output quality.
Comparative tests were done for each DVD2One quality options, including the new ones.

[quantum]

I just compared the mixed extras from Shrek 1 using 3.2 and 3.1.7 reducing both to 60% of their original size. The newer version is obviously and significantly improved in both interlaced and progressive scenes. Macroblocking is reduced everywhere.

It also appears the annoying "pulsing" effect from 3.1.7 and earlier versions is gone. Version 3.2 seems to have a more even distribution so you don't have the 3 good frames and 1 bad frame syndrome which caused the pulsing in previous versions.

My previous comparisons with IC7 and Shrink 3.1.7 showed they were both fairly close, with IC7 maybe being slightly in the lead. I don't intend to retest IC7, but my guess would be that Shrink has at least caught up with IC7 and possibly jumped ahead.

Really fine work dvdshrink.

[jsquare]

All I can say is that this new version of DVDShrink "kills" IC7 and probably all other transcoders in every aspect.

I was trying to convert the last 2 discs of the LEXX series(S4V5 and S4V6) and was having a lot of problems with them, macroblocks all over the place with IC7, IC8 and Shrink 3.1.7, I even tried DVDRebuilder which did a good job but I didn't wanted to go back to my DVD2SVCD days of re-encoding.

So this new version came right on time, took around 1.5hrs for each disc and the results were better than expected, wish I had this version of Shrink when I started doing the whole series.

One more thing, I don't think you can call DVDShrink a "transcoder" anymore, the new engine have very similar results to those made by CCE with DVD-Rebuilder, so it may have become a fast "re-encoder" instead.

[Lagoon]

Well you still have to call it a transcoder since it IS technically one :p

[Lazza]

Originally posted by valnar
OK, I'm impressed........... Way to go DVDShrink team!

-Robert
Glad you like it, I'm sure none of us would object if beers were handed around. :D http://smilies.sofrayt.com/%5E/g0/beer.gif

[Solo]

Very nice. I have not done a full movie yet, but I loaded an image and took a look. A lot of new options. Looks awesome.

Thanks :D

[DVD Maniac]

I just tried a "Power" test with Terminator 2 which has always been a problem for any encoding / transcoding tool I have used previously. The interlaced source requires hefty compression if you want to keep DTS and the directors track (55%) and all my previous attempts were very dissapointing - even CCE struggled.

I watched source and Shrink outputs together and I must say the two really are very difficult to distinguish. There is some very slight blocking on the action scenes and the colour is very slightly washed out (but nowhere near what you got with previous versions). Also, the "mosquito" noise that I got very badly on this title with Shrink before is nowhere to be seen.

All I can say is "awesome". Congrats to the team

Can I ask those close to the development where we should be posting development suggestions? Mine are as follows -

1. Single frame title AND menu vob replacement (ala Menuedit etc)
2. A batch mode (now the run times are approaching 2-3 hours it makes sense to have this feature for running overnight batch jobs)
3. Automated / Semi-Automated Menu button removal / disabling (thats a tough one I know - but this would make the app the ULTIMATE backup solution)

My Pinnacle crapware and other redundant apps are being removed right now!

[Lazza]

Originally posted by DVD Maniac
Can I ask those close to the development where we should be posting development suggestions? Mine are as follows -

1. Single frame title AND menu vob replacement (ala Menuedit etc)
2. A batch mode (now the run times are approaching 2-3 hours it makes sense to have this feature for running overnight batch jobs)
3. Automated / Semi-Automated Menu button removal / disabling (thats a tough one I know - but this would make the app the ULTIMATE backup solution)

Don't think that's going to happen sadly m8, but only dvdshrink himself can answer that Q about any possible future development of DVD Shrink. ;)

[dragongodz]

just curious what ideas that were suggested on the dvdshrink forum are actually used.

i can see the higher requant to higher original quants has been for example and error propigation reduction for P frames. the settings dont exactly say though what was found useful and what wasnt. :)

[TheSeeker]

Am I just seeing things or does shrink tend to output a slightly grainy picture? It may just be that I was compressing by 30% though. Most likely thats it. Would the smooth or max smooth option be better for longer movies that require more compression?

[nwg]

I have done ROTK (PAL) with Shrink 3.2 and got much better results than with DVD Rebuilder and CCE or other transcoders (DVD2One, CloneDVD and Rejig). I used the default sharp setting and compressed the film down to 58%.

I have also done Once Upon a Time in Mexico (PAL) with the film at 88% and the extras at 45% and the same quality setting. The picture looks excellent for both the film and extras.

[TheSeeker]

So sharp default setting seems to be good in pretty much any situation? I wonder if smooth would be good for anime? Cause i know that the mosquito noise can really become apparent right in the middle of those flat color areas if you compress too much.

[geffroman]

It's really hard to compare SHRINK and ReBuilder unless you are going to include the number of passes and settings you use with CCE... and what version of CCE...

[motster]

Does Shrink have an easy way to tell which title is WS vs FS on a disc? Sometimes doing Re-Author it's not always easy to tell (especially on my machine where i use Remote Desktop and can never preview titles :()

[KYUSS]

Originally posted by Mephiston
So far all i can say is WOW!

If DVDShrink gets any better i will not even both with CCE for movies under 2 Hours anymore, and if Shrink goes retail, i would be willing to buy a copy now.

Great Work!

what more could be added for retail release?

maybe its own burning module?

Fantatsic proggy

cheers
KYUSS

[Gang$ta]

This new Ver.3.2 is amazing..
Just did a movie,165 mins using D/A & AOC (default}

Fcnkin Awesome...

A big THX to Mr DVDShrink

[bugsy_pal]

I am relatively new to DVD burning, but DVD Shrink has been my most used tool. With 3.1.7 I had made a copy of "The Third Man" - the compression level was something around 50-55%, from memory, and there were lots of artefacts and smearing between frames. The new version with AEC does appear to have reduced these effects.

Just a query - do black and white movies present more problems than colour? I backed up Big Fish using the new setting (sharp), and the result was outstanding, but The Third Man just seems to be a really difficult one to get looking clean, unless the compression factor is 80% or higher. Admittedly, there is a lot of detail in the B&W picture, but I would have thought that that would also be the case with a colour flick like Big Fish.

[Joergen]

Again more testing, and to give an easy-reading comparison:

DVD The Gladiator R2

-DVDShrink 3.1.7: without the commentary track (just the 5.1 track) the ratio is 72%, and the results very blocky in scenes like the opening battle in germania = hopeless.

-DVDShrink 3.2 = WITH commentary and 5.1 track the ratio is 69% and at default sharpness the quality is excellent in all scenes.

NOTE: Interestingly switching on the "maximum smoothness" setting that forces the newfangled encoder on every frame makes the quality far worse. The maximum smoothness reminds me of IC7 quality (or as I like to call it, lumpy block porridge -quality).

[dvdshrink]

Just a query - do black and white movies present more problems than colour?
Color usually accounts for 1/3 of the compressable data in a movie, but it's quite possible that DVD Shrink would gain more compression than that from the color components, perhaps up to 1/2. If there is no color data, the rest will undoubtably suffer more.

[vaylon]

Thanks shrink!!!!

Just tried 3.2 and loved it.
I have dvd2one and ic8and7. Neither compare for both quality or speed.

have shrunk several disc for comparison against 3.1 and quality is much better(although 3.1 was great). On full disc backups like lotr-rotk the quality is very noted. But what surprized me was the reauthoring movie only quality.

On 3.1 overall quality was excellent with no compression artifacts at all.Except in dark areas it looked a little washed and grey.

on 3.2 overall quality was excellent with no compression artifacts at all. dark areas were much more true to the originals ,didn't get the wash look. Colors also stayed more vibrant than the earlier version.

another note is about cpu usage- 3.1 hardly used any cpu cycles at all( the most I ever saw it use my xp2800 was maybe 10%)and that was rare.
But 3.2 used 33% right on start up and spiked up to 67%. This is not a complaint, it nice to know a program is actually using my chip. But if someone else has a problem with3.2 locking or crashing their system they may want to quit multitasking and cutoff a few programs.

Overall I would have to give this version 5stars *****.
Great job guys.

[TheSeeker]

I recently backed up Harry Potter and the Sorcerers stone at compression levels of about %64 percent i think, using max sharpness. I have to say the result was darn near indistinguishable from the original. Maybe I was just tired and seeing things but the result was great. Alot of detail preserved. Still not too sure about what settings work in what situations better but that will come with experience. Anyone who has used the final 3.2 alot have any opinion on this matter?

[valnar]

I'm really interested in the technological advances they made in the transcoder. I know that the old DVD Shrink, DVD95Copy and DVD2One all used something very similar - it's like they read the same white paper on how to do it. Quality was comparable. IC7/8 was ahead of the pack because they did a "deeper" transcoding technique, which also took longer, but made it look nicer.

(Forgive the laymans description, but this is what I've read over the year.)

So my question is what changed at the core in the new DVD Shrink 3.2 to make it better than IC7/8? Is this a brand new development in the transcoding process? Or just an evolutional enhancement?

Inquiring minds want to know.
Robert

[smok3]

i was just testing the disk1 of 'blue planet' - lots of water and stuff like that - that did not look very good on original either, compression required was under 60% and with deep ticked and the use of aec the quality remains pretty much the same - most noticeably there is no color bleeding.
edit: powerdvd's interpretation (cropped):
http://somestuff.org/tmp/PDVD_033croped.png
(~215k)
(ok, this is all very subjective, i would have to do a lot more testing to conclude anything...)

[mb1]

Little test update (http://people.freenet.de/mb1svcd/compare1.htm)

I stopped comparison tests for now - because too much time consuming (as long as Shrink does not recognize that no double encoding is necessary for this full disc backup).

But I will further investigate into Shrinks behaviour with MProbe analyzing software.

[mrbass]

Originally posted by valnar

So my question is what changed at the core in the new DVD Shrink 3.2 to make it better than IC7/8? Is this a brand new development in the transcoding process? Or just an evolutional enhancement?

Inquiring minds want to know.
Robert

@dvdshrink
Can you post partially in your 3.2 beta10 annoucenement describing technically what exactly each option does. You said "not to post it" so I'm not gonna to post it. Can you give just a basic summary of which method is compressing B,I, and P pictures more and/or less like back in the old days with dvdshrink 1.03?

[TheSeeker]

Are these findings valid for interlaced material only? Because just did a test with Star Trek First Contact. I went and did a full backup with deep analysis and aec max sharpness, and the output look much much better than I thought it would. The only actual noticeable flaw when playing it back on my comp was that some really extreme details (facial hair, wrinkles, etc.) would go a little smooth or fuzzy but then immediately correct itself. Which is what aec does it compares the frame it just encoded to the original to see if any artifacts are introduced and then corrrects it so artifacts dont propogate. I dont know, but Im highly skeptical that DVD2One could get results as good as this. Those times you quote seem a little high to me although I am running a Athlon64 so that might have something to do with my faster encode times.

EDIT: Sorry my compression rate was 50%. That might be important to know.

[dvdshrink]

Sorry mrbass, I've been passing such questions by. The reason is, I don't wish to discuss in public the technical details of DVD Shrink transcoder implementation.

I'm sure that experts like mb1 (above) will uncover most of the internal workings anyway ;-)

I'd appreciate your understanding.

[dragongodz]

I don't wish to discuss in public the technical details of DVD Shrink transcoder implementation.
thats fine to some degree but i will quote myself

just curious what ideas that were suggested on the dvdshrink forum are actually used.
since you asked for and got suggestions isnt it also then fair to aknowledge what you found useful ? i do not mean the fine detail but the general concept or idea that was posted.

[dvdshrink]

The main thing that came out of that discussion was AEC, which is that it is necessary to compensate for errors introduced by the compression process. I believe IC and DVD Shrink are the only ones doing that now (I may be wrong). I'm sure everyone will be doing this soon.

Other ideas were tried and rejected - compress the edges of the frame more, for instance. It didn't work out. I need to read the thread again, because it did go on for a while after I left it.

Then the discussion went into private beta testing forum for several weeks, where many useful ideas were proposed and some of them implemented. There are very many ways to do AEC, some of them better than others, which is why you'll find considerable differences between DVD Shrink and IC.

I'm afraid I did rather lose contact with the original thread (truth is I forgot all about it, I should get back there and make amends).

[mrbass]

Totally understood dvdshrink. About the thread...not only was the last post about the technical stuffed changed so you never did post the final details anyways so it's a moot point. Only you dvdshrink could post what each one does and doesn't so I totally respect that.

Only thing I can say is it's come a long long long way since dvdshrink 1.03 (October 2003). Here's the table for dvdshrink 1.03 'customize streams' section.

Technical Note on Encode Levels

DVD Shrink uses a simple re-encode algorithm, which operates at one of two levels of compression, on each mpeg picture type. The following chart indicates what each encoding level means.

Level I-pictures P-pictures B-pictures
Level 1: No Compresion No Compresion Level 1
Level 2: No Compresion No Compresion Level 2
Level 3: No Compresion Level 1 Level 1
Level 4: No Compresion Level 1 Level 2
Level 5: Level 1 Level 1 Level 1
Level 6: Level 1 Level 1 Level 2
Level 7: Level 1 Level 2 Level 2
Level 8: Level 2 Level 2 Level 2

So my point is that AEC is nothing like the above because there is so much more going on. Thanks dvdshrink for the intense 3 weeks of development you spent on it which was way more than anyone anticipated when you initially made the announcement.

[dvdshrink]

Ah yes, I remember the levels!

You could never get the output size you wanted because one level was too much and the next was not enough. Things have improved :-)

Anyway, it is my personal belief that AEC with "max smoothness" option most closely resembles the output from IC8, thus you could hazard a guess that the internal algorithms are similar, although I wouldn't know or assert that to be true, since I'm not an IC8 developer.

It came out during beta testing that "max smoothness" was not always the best choice for encoding hollywood movies, in fact many improvements could be made. Ultimately "sharp" was chosen as the default setting, while "maximum sharpness" seems to do a pretty good job too, and may even be preferable to some.

There is always a trade-off, so ultimately it depends on both your viewing equipment, which may show some types of defect more than others, and your own personal taste - and of course the particular movie you are encoding.

I've been asked to explain which AEC options are best in which circumstances, and the truth is, I don't know. I understand what the options do from a mathematical perspective, but this does not translate into real-life recommendation. I'll be watching reports on this forum and elsewhere with interest: like everyone else, I'm keen to learn which options I should use for my backups.

So it's over to you - mrbass, or anyone else - if you have ideas which AEC modes work best for different material (interlaced content, serials, anime, old movies, hollywood movies, etc) please post them!

[mb1]

Changes are easy to analyse but a little complex to write down:

DVDShrink 3.2.0.15

Compression ratio changes I-frame changes P-frame changes B-frame

90% no aec very slightly no around -15%

90% sharp very slightly very slightly around -15%

90% max smooth around -10% around -10% below -10%

90% max sharp very slightly no around -15%

80% sharp no no around -30%

70% sharp around -20% around -17% around -35%

60% sharp around -38% around -38% around -39%

57% sharp around -42% around -42% around -39%
smallest possible
for example movie

57% max smooth around -40% around -41% around -41%

Of course these values can vary ±2% depending on source material.
The values are based on shrinking and analysing progressive hollywood movie 'Starship Troopers' (PAL with GOP15).

There are also some interesting little findings.
B-frame before and after an I-frame doesn't change its size any more below 80% ratio.


I think it would be nice, too if some users develop a database for movie deep analysis data where you can download the analysis data you just need (if it doesn't change for future releases). Would save the community a lot of time and cpu power :D
Maybe any thoughts on that ... :rolleyes:

[TheSeeker]

It takes 10-15 minutes to do a deep analysis. Its not that tedious that we need to spend all the time and effort to make a database is it?

Not to mention there is like 50 different version dvd's for every movie it seems.

Another thing. Wouldnt those compression ratios vary quite a bit depending on the movie and the amount of action/inaction and color, and brightness?

[nwg]

mb1 wrote,
I think it would be nice, too if some users develop a database for movie deep analysis data where you can download the analysis data you just need (if it doesn't change for future releases). Would save the community a lot of time and cpu power

It has been suggested before on the Shrinks forums.

There is to many variables to take into account such as what TheSeeker mentions above.

TheSeeker wrote,
It takes 10-15 minutes to do a deep analysis

Wow, it take 30 minutes on my setup.

[TheSeeker]

Athlon64 3000+ system. My new toy. I love it so much. I can do encodes around 3 times faster now. A 4 pass encode in CCE only takes like 3 hours now. Its great.

[nwg]

TheSeeker wrote,
Athlon64 3000+ system. My new toy. I love it so much. I can do encodes around 3 times faster now. A 4 pass encode in CCE only takes like 3 hours now. Its great.

Impressive.

I had to make a PC on a budget, I managed to put a AthlonXP 2000 in it.

A 2 pass CCE will take 3 hours.

[TheSeeker]

Yea I used to have an athlon t bird 1.2 ghz.. so for me this is a HUGE step up in the right direction.

[ookzDVD]

Just did Mummy Return Region 1 DVD, ~65% on the main movie,
and ~40% on the extra. AEC - Sharp mode enabled.
The result is awesome!

Thank you dvdshrink :)

[fozzieb]

Not to mention there is like 50 different version dvd's for every movie it seems.

Every version of a DVD has a code just like cd's so the dvd could look up a database find if it has the detail available and download it.

[DVD Maniac]

Having thought about this and contrary to an earlier post, I am not convinced that a database compilation for AEC settings is really viable (or useful).

Just think about all the varibles we have -


DVD Type -
Hollywood Movie (High Quality)
Old Movie (B&W?)
Anime
TV Series (High quaility)
TV Series (Low Quality)

Format -
Interlaced (Most PAL!)
Progressive

Watched On -
PC CRT
PC TFT
TV Tube (old)
TV Modern Tube widescreen
Plasma
Projector

That's 5x2x6 = 60 possible generic types of scenarios. OK, you can probably get this down to 40 or poss 30 by eliminating the less popular combinations but thats still a lot of work. Add to this the "personal preference" factor and its just not a goer.

What do others think?

[mb1]

I did not speak of database for 'best aec settings'. That would be in fact nonsense.
I meant 'deep analysis data' which DVDShrink stores for every analysed dvd.

The average analysis on an average pc will last around 30 minutes (sometimes more if structure is complex).

Take 'Matrix RC1' for example. Thousands of people made a backup of this blockbuster. Thousands of times this 30 minute deep analysis could be used for better things. Ok, this dvd is a past one. But think of all the future blockbuster releases, Star Wars IV-VI, Kill Bill 2, Troja, The day after tomorrow, Harry Potter and many more.

I would find that extremely useful. My opinion.

[m99]

Originally posted by fozzieb
Every version of a DVD has a code just like cd's so the dvd could look up a database find if it has the detail available and download it.

This is something I have missed since I bought my first DVD. Why isn't it something for DVD like CDDB is for CD?

[smok3]

i agree, that would be a usefull option, similar stuff was recently incorporated into 'some' audio ripping tool, also 'central database' like - just for a litlle different purpose
check:
http://www.hydrogenaudio.org/forums/index.php?showtopic=4789
(but in the case of video dvd - the solution would be easier - as the deep analisis could just compare few seconds generated localy to ensure that input is correct - i guess).

[fozzieb]

I used to work for a company "www.moviemine.co.uk" and we were sent a program that read the id of the dvd looked it up on the studios web site (or database) and filled in all the info we needed, movie title, director, year of release......

it was very cool and i'm sure it was a shareware app rather than some custom software.

i think a database with the code number and the deep analysis data is all it would need, so if i had LOTR 1 it would see my dvd code and get th deep analysis data, but if you had LOTR 1 extended it would have a different code and get different deep analysis data from databse.

or like whatever :D

[infoscapeone]

IC7 had such an analysis database but they had to switch it off due to leagal reasons

[smok3]

Originally posted by fozzieb
[B]i think a database with the code number and the deep analysis data is all it would need
it would be better to use some other logic, think about movies that get reauthored by just adding some subtitles, when video stays (hopefully) the same..., for example let user choose (or even search) for the approapriate title, then the compare can be made with partial local analysis.

edit: (ok, i dont know how related is the analysis to the actual dvd structure at this point)

[smok3]

Originally posted by infoscapeone
IC7 had such an analysis database but they had to switch it off due to leagal reasons how about decentralized p2p way? :p

[TheSeeker]

Would it really be worth it to save 30 minutes?