What sample was that Phil :!:
Because every sample I've done (posted), looks far worse when done with TMPEG, at the same file size :!: :!:

-kwag

Hi everybody

Just wanted to know if the ma script is usefull now or not.

Depends on your source material, or if you want to reduce the file size even further :)


-kwag

It seems you removed it !

I commented it there :
http://www.kvcd.net/forum/viewtopic....r=asc&start=22

And you suggested some test to do but I don't do them for the moment.

The problem is not to tell that tmpgenc is better or not. The problem is to stop to say "with ffvfw that's the end of prediction" "3 two hours movie on a DVD" "this codec is awesome" blah blah blah...

I thought we were looking for perfect encoding here. Not "perfect according to the circumstances" nor "perfect compared to everything that exists nearby". Okay other encoders would have done it worst, but what I see in this sample is not perfect and will never be on any DVD I burn. It looks like a bad Divx :!:

People are so crazy about this new "toy" that they forget what is the goal in this. And the goal is certainly not to do divx. Else let's use the Divx codec !

Note: I'd rather the sample you did a little before (http://www.kvcd.net/ffvfw-high-action-sample.m2v)

That's the matrix. Not a sample :twisted: Sorry Phil, but I think you also have a PLAYBACK CODEC issue :!:

-kwag

Ok, but it looks marvelous even with quantizer = 4!!
And thats the reason for a more blocky surfaces picture. ;-)

Stay cool!
These Pics are just Impressions to assume also how the video as a whole does behave. If there would exist dancing blocks, ants etc. I would say it.

The problem is that my free provider doesn't permit that big needed traffic for this which you already can assume.

But the videos themself are intact! I wouldn't post some " 8O 8O 8O " if the quality also in motion wouldn't be that good, .... as I'm like you a little perfectionism ;-)

Don't take the words that seriuos Phil, ist a yellout of being "amazed" AND we also did say that we're just at the beginning of testing and developing settings on CBR mode of FFvfw mpeg2 .... you see.
And testing also means that on some movie sources the "being amazed" could turn to failures. .... So up to now there have been failures and successes ... and therefore spoken seriuosly: I still did not come to a conclusion as Im waiting for the next upcoming tricky behaviour of that encoder. And thats why I also replied in some posts that up to know I do not recommend anything according to settings or even encoders ;-)

And as you know treating Captures is also my passion and I will see this weekend which potential that Encoder really gots on material more worse than good mastered consumer DVD as thats what we right now do use when testing sources.

Lol. My mistake, it wasn't were I think it was.
I can't find it for the moment (I don't have mpeg2 software on my office's PC) but I think the title was "ffvfw-notch-kdvd.m2v"
(read my comment about it, it's a sample with Carrie Ann Moss and Val Kilmer, you should probably find what it is !)

No Karl you CAN'T say that ! It's too easy :!: I'm in video (especially MPEG2) since years and I can see when there is a problem and find if it's due to playback or not. BTW, I didn't change the softwares recently and others samples you did don't suffer such problems. As all my previous KVCD and KDVD !

Again read my comment, watch the sample and you will see all the defaults I gave.

The only comment you can have and would be correct is that some of these defaults probably wouldn't be on a TV set. We all know that. But that does not make this sample good.

I know that, I'm happy to see that you know that too, Karl probably know that also but the problem is that everyone looking the thread and watching your (or others) stills can't know that.

Look at in the spanish forum for instance : people that never post try to use this codec ! Why ? Because your stills are too perfect (for sure you post only the perfects one, that's logical :-)).

And that is only a part of the problem. The other part is what I said before : you lost your focus. From "doing awesome KDVD" we go to "doing better KDVD than CCE" and now "doing better KDVD than tmpgenc". Doing things better than something known to be very bad can lead to do things that are "rather good" or "not so bad" if you don't pay attention. I still looking for "good" things.

I can't tell for your own samples, I'm just talking about that sample of kwag I can't even find for the moment but I think you found out.

Yes and for sure Kwag had also a lot of failrues :-). But once posted here, and described with words like the one you quoted at top of your post, a sample is supposed to reflect a success. And this one wasn't.

I just wanted to "refocus" the competition.

Well, as for me, I'm yet to find the "perfect" encoder, and I'm pretty sure we never will.

However, this codec to me seems a lot better than TMPGEnc and CCE, if I view the VIDEO samples side by side. I don't think something is wrong when I say that. I'm sure noone has lost his focus here. We all encoded with CCE and TMPGEnc for years, and the fact that we all test out ffvfw right now to me seems a great thing. I actually love the enthusiasm in this forum, and I love to participate in it. It's ok if one disagrees with some of the achievements here. But I still believe that there's actually no difference in finding the "perfect encoder" and finding an encoder which is better than TMPGenc or CCE right now. Because until now TMPGEnc and CCE were proven to be the best encoders out there, although far away from perfect. Now every encoder which does a better job, let it be only in High Bitrate MPEG-2, is something great, and also something to be enthusiastic about. :roll:

@ Jell

ok, but I tested yesterday only on one movie where the results where in my eyes awesome, .... but the next movie "The Core" was very hard to handle ... ok I wanted to end up with a 480x576 K"V"CD ;-)


@ Phil

Well my frames do look that good continously in that stream, I just selected frames where the encoder shows different difficultinesses (ähm english?? ;-) ) to handle.
Also look at my last 480x576 on "The Core" no blocks at all and still a q curve very low for that kind of movie treatment and length. BUT to receive that picture I had to blur it a little... and that schould not be our final prediction method, do rise negative unfilter() settings for prediction. That sample did look even when using Notch much more worse in TmpgEnc compared to FFvfw mpeg2.

I now what do you mean by sayin that others will get confused by hearing those impressive yellouts :wink: But ... well lets party a bit ... its like new girl-friends which by the time will still be even more interesting or maybe the other way und you leave her :lol:

Now to Kwags sample ......

@ Kwag
sorry Kwag, I didn't took a look at it as I first had to catch up the last postings or if yours was posted already yesterday I overflew it cause of my being exited.

I have to agree with Phil according to the "gibbs" at the edges at the beginning or the surface over Val Kilmers left shouder at sec 6.
Im at work now and Im not shure if you did mention your quantizer and what would be the approx. final full filesize for that movie?

Did you add noise to that sample?

I've just tested the 1-pass-quantizer-mode (@2)
At first view, it seems to be great. Filesize is about 40% less. I have to verify that 8O

EDIT: Was a bug in VDub (bombed and encoded only 2/3 of the video :roll: )

I'm sorry but ffvfw does not do better work than CCE on a 1 movie-DVD, and CCE is still a very good challenger for 2-movie DVD (taht probably depend on the movies but ffvfw should loose on somes and win on others).

ffvfw if far better than CCE when we are talking about quite small target filesize (let say 3 movie-DVD). That's a reality : at this, CCE is awfull, at this ffvfw is better. But... at this ffvfw results are still unusable.

As I said, doing better than "not good" for the moment just lead to reach "rather good". And I just rang the bell because I see too much big words and nothing that bring us back to the real world.

This morning I tried to run virtualdubmod and it didn't even start the process ! But I hadn't find the time to look further the reason.

No, I didn't add noise. I think :roll:

But apparently, it's a source related issue :!:
Look at my screenshot from the original VOB ( different aspect, because VOB is anamorphic ) :

http://www.digitalfaq.com/archives/i.../2004/02/5.png

Now look at ffvfw's screenshot:

http://www.digitalfaq.com/archives/i.../2004/02/6.png

And TMPEG's screenshot:

http://www.digitalfaq.com/archives/i.../2004/02/7.png

Anyway, you can clearly see that around the letters and some edges on the original, there are slight artefacts. So ffvfw is doing a great job reproducing them too, but not creating new ones :)

-kwag

:lol: okay if the source in the cause, that's different. A script can handle that easily :-)

And for all others problem I related ?

Kwag

Have you tried 528x576 on dvd? I've tested it on my pana dmre 50 and it works. I've used bitrate 6426 and i frame interval 25.

No, haven't tried that yet :lol:
What I'm going to try in a little while, is encoding one of my Panasonic DMR-E80 captures, to see how it looks :cool:

-kwag

You can find the here:

TMPGEnc-0.11.20.97.zip (Main Program) http://www.pcphotovideo.com/Download...0.11.20.97.zip

tmpg_en0725.zip (Update to .12 and Japanese to English Translation) http://www.pcphotovideo.com/Downloads/tmpg_en0725.zip

Hope this works for you.

Racer99

I don't know if it will make any difference but I have been using VirtualDub-Mpeg-2. I know some of you have had problems with VirtualDubMod, it might be worth trying.

http://fcchandler.home.comcast.net/

This is also what I use.

@ Racer 99

Thank you very much for the links!!! ;-)

But in CBR mode we don't suffer from that VBV issue any more.


But anyhow, thanks a lot.



Inc.

Yes it is. Picture quality is also pretty acceptable (litle blured vs CBR). But again we have VBV at 7 as with Quality mode :(

For the other hand:
I've tested CBR mode with VBV 20 and 24. I always get bitrate peek at ~4000 even above 5000. I think that for KVCD this is unacceptable. Can I look forward for litle explanation (as short as possible) of VBV part in encode/decode process? Even with bitrate at this high level, this files are playable on standalone. Maybe I'm lucky :)

So,
No posts since 05-02...that's it :?:
Can't believe that the thread is dead 8O
Even because we're still missing 20 pages if we want to match the record of CQ vs CQ_VBR :!:
Come on guys. Cheer up and post your ideas.
C ya

Well as for me, I'm done with this Codec... Sorry to say.

After heavy testing, I don't see a way to limit those bitrate peaks (15000 and more). When we find a way to fix that it'll be great. Until then, I already spent too much time on it! :wink:

8O I never had a peak of over ~8,500, when using KVCD's Q matrix :!:
Is that at 704x576 :?:
Maybe that's why, because at 704x480, I never reach the MAX (for DVD) which is ~9,000Kbps (9,800?, don't recall now).

-kwag

That's 704x576 Quantisizer 2, with the 6000 something Bitrate set (for 224 kb VBV) and the Notch matrix and a simple resizing script. Even with Quantisizer set to 3 i get peaks of ~9000 and the texture quality for plain surfaces decreses a lot. :roll:

Well, I don't seem to ahve any problems w/ this codec :P

It's ok, but I do have one or two quams about it..

1) - Debuging "blindly" is a prblem w/ me.. (ie, no PREVIEW on some items)

2) - it takes approx 1 minute for it to respond back, when I have to go back
and make a change in the compression. It's like it goes to sleep and then
finally wakes up, "oh, I'm ready now.. you mak enter.." or something like
that.

3) - I can't complain about quality. I found it very close to my TMPG encoding.
That's for MPEG-2 source. Can't say for MPEG-1, cause TMPG is suppose
to be much better w/ those :P but I'm not a fan of MPEG-1 anyways :P

4) - lacking in better interface. There's need for a GUI. I'd do it, but I'm
not familiar w/ this .dll communiction of opening up a codecs Config box
and directly modifying it contents (as though you went through the vdub
and Compression route. I know this is possible, because I've done it once
before (beta-testing an idea some time backwards ago) - - Like I said, I'd
cook one up if I knew how to call it from my Delphi app. As my understanding
goes, most of the settings get saved in the .reg file anyways. But, you still
have to call the Config (or is it Settings) to make the actual settings/changes.
Mind you, I was researching this, but then I thought it was more important
to find a MUX/deMUX/Multiplexer for these .avi wrapper files. (see below)

5) - better (cleaner) method to MUX/deMUX/Multiplex the source for a proper
DVD author project. Something needs to be created that will allow a user
to just drag 'n drop (my favorite app feature) and let it do it's thing (based
on a profile, that is user-configurable)

Basically, the above is a short recepee (or wish list) for tweaks/features.
Feel free to add to it, if need be.
.
.
My favorite are two.. Better GUI (need to create one) and MUX/Multiplexer

But, I still prefer my TMPG encodes to MPEG-2 in the mean time.
Well, we'll see what new things happen, but there's room for others here to
contribute (ideas/methods/process etc etc)

-vhelp

@ jell and kwag..

In an effort to come to a resolution (not screen'wise) perhaps, if you both
agree w/ a DVD source, and both do the same scene, using the same params
and see where you both compare (or go off)

Next, you can then fine-tune the issues.

How does that sound to you guys ??
Heck, I'd jump in too, just to see what I get. Then, we three can post our
final BitrateViewer results here - - Yeah ??
I've lot loads of DVD's just aiking to be push into battle hehe..

-vhelp

@vhelp: Well I'd surely join your testings, but I don't see how we can "fine-tune" that codec, so that it won't produce such high bitrate peaks. :roll:
Even if it will work with the material that we all use for testing, there's no way I can get my DVDs here to work with it. I limited the bitrate peaks to about 11.000 but that's not what I want at all... :?

Hi Jellygoose,

Correct me if I'm wrong, but...

Shouldn't our goal be to "mimic" as closely the bitrate values that are in
the DVD, but at a lower projected size (ie, SVCD or re-encoded dvd proj) ??

I always thougth that as long as you maintained the same level of bitrate
spike (spike for spike) that, THAT is the ultimate goal. So, say our DVD
disk is spiking between 1kb through 7.4kb, that would be considered normal
for a DVD title.. BUT, now we want to re-encode to say, dvd (notice how I
use small letter to signify our re-encode) and our new goal is now to produce
bitrate spike (per our new encoding method/process plus N.Matrix) to our
new 700 to 4.5kb, we should now have the same spikes, but w/in a lower
threashold.
Now, I do realize that this may not work in our real-world encoding because
we have different MPEG encoding engines/algorithems etc., but I still think
we should always have the goal to come close. But still, it's prob. pointless
because we are encorporating Filters and different methods of resizing and
setting our Aspect Ratios, of which may alter the bitrate pathways etc.

Sheesh.. I hope I made some sense here. If not, blaim it on my NEW mobo..
I woke up to a dead pc this morning, and I was really pissed, cause I had
lots of things open that I needed to finish up on.. RaTs

-vhelp

Well I agree with you in most parts vhelp. But that is exactly the problem of this Codec for me. It does NOT produce DVD-Compliant Bitrate peaks. On the other side it goes down to 200kb/sec and that is also not DVD compliant I believe. And I don't see a way how we can fix this yet.
So I'm asking you what you intended with your post. :roll:

Just to clarify: the DVD bitrate peak is 9800, considering audio :!: So if you have a 192 kbps audio track for instance, your max peak for video should be 9608 kbps. :wink:

Actually, from the link below :arrow: "Maximum video bit rate is 9.8 Mbps."
http://www.dvddemystified.com/dvdfaq.html#3.4 ;)

-kwag

But it doesn't make sense... So if you decide to put eight 448 kbps audio tracks (in theory), total bitrate would go up to 13384... I'll mail the guy and confirm.

Hay guys, TGIF..

I'm curious. I've ben still testing ffvfw out w/ my setup, and so far, I can't
really complain. But, ...

I'm wondering which movie's are giving your guys the high spike, and at
what bitrate were you using at the time ??

I'll give you an example of a scene for you guys to test this spike out with.
.
.
Good movie to test spike w/ is Fifth Element. Very end of Chapter 6, I believe.
If there is a spike lover in a given snenario, this is it. Just rip Chp 6/7
together, cause the spike runs through the transition to Chp 7. Only a minutes + worth of your encoding time :)

I'll post the spike, based off my ffvfw encode if you like !!
-vhelp

For me i get 4000 to 6000 bitrate spike at 352x288 res on cartoon (Futurama). And with movies i get the same values at DVD res. But i use quantizers no highrer than 3 or 4 (for beter quality). I know that rising quantizer to 6/7 gives reduced spikes to 50% of this.

This is just not what we aim for. If we lower the Quantisizer to 5/6 of course the spikes will be under 6000kb. But that is due to the lower AVERAGE bitrate. The spikes are still there, most likely as high as with Quantisizer 2 if you compare the spike to the average Bitrate.

hay Jell..

question..

But, where are you getting the spikes from ??

I'm assuming that your spike issues are from scenes that are not "spike-worthy"

IOW.. are your scenes slow-moving, but spiking in them ??
Or, fast-moving, and spiking w/ good reason.. but maybe just TOO spiking ??

Here is screen pic of btviewer's results of my previous discussion of the
movie, "The Fifth Element" (chp 6/7) ...

http://www.digitalfaq.com/archives/error.gif

-vhelp

Jell,

If you've sceen the scene, you'd understand why the bitrate went up that
high. Fire does that !! But, the nature of THAT scene called for such a high
bitrate. I agree w/ it's decision to 6.5k it.
.
.
But, w/ respect to that pic I posted above, and that scene, IMO, lowering
the spike (I mean, bitrate) would only make that scene blocky ??

I have the 4.1mb clip too.. if you wanna see that, but if you have the dvd,
then you don't need it, right ??

Anyways..
-vhelp

@vhelp:

You're totally right, a spike of 6.5k in that scene (fire) is totally ok, and might be needed for such a scene.
The spikes I'm talking about DO appear in such high-action/movement scenes, and also might be totally rectified for such a scene. That doesn't change the fact that a spike of ~14.000 is just not DVD-compliant. By the way, I also got an average bitrate higher than yours, which totally fits my theory. for an average bitrate of 1663 a spike of 6500k is extremely high. almost 4x the average bitrate. Now if you need an average bitrate higher than that (noisier source for example), let's say 2100 kb/sec that would also be a spike of 8400kb/sec max. bitrate. and my spikes are even higher, as said.
:arrow: no way to specify max. bitrate :arrow: this sucks 8)