ffdshow tryouts project: Discussion & Development

[clsid]

No. There are currently no developers working on ffdshow.

[SledgeHammer_999]

No. There are currently no developers working on ffdshow.

oh I didn't know that. Bummer.

[kirakami]

Are there any plans to improve ASS/SSA rendering or use libass for that instead?

agreed. FFDShow really needs to improve SSA/ASS render.
there are still problems rendering new versions of SSA/ASS subs
also vobsub hangs. ffdshow r3978 20110825
would be good if there comes option to use libass instead.

[kirakami]

libass Render Download for windows
http://srsfckn.biz/assrender/

Start =>Run => regsvr32 "C:\assrender.dll"

[Stephen R. Savage]

AssRender.dll is an Avisynth plugin. You can't register it or use it in DirectShow, or anything like that.

As for ffdshow needing a new subtitle renderer, there's nothing stopping you from using the MPC-HC one or DirectVobSub. ffdshow development has halted and it is already planned to be replaced by LAVFilters anyway, so there's no sense in petitioning for feature changes.

[egur]

clsid:
Hi,
I'm releasing a test build from the latest ffdshow build that include Intel SandyBridge accelerators.
Eventually, I would like to create a separate installation build that defaults to the Intel decoder for supported codecs.
Could you please point me to where I can find the installation script? I've looked at the SVN, but I can't seem to find it.


Everyone:
Here is doom9 release thread for ffdshow DLLs supporting hardware acceleration with the SandyBridge CPU:
http://forum.doom9.org/showthread.php?p=1523738

[Eliminateur]

ffdshow development has halted and it is already planned to be replaced by LAVFilters anyway, so there's no sense in petitioning for feature changes.

what..?, so ffdshow-tryouts is essentially... dead?, what happened to all the developers?, so we can expect no new versions with nothing new in them? :confused:

[pirlouy]

I need help with one little issue: I've selected "ffdshow audio processor" in MPC external filters. But I can't access options window. With another computer, it works.
I've tried uninstall and reinstall (with ICL or MSVC2010) without success. Any idea ? I have no problem if I choose ffdshow audio decoder. :/

[ryrynz]

what..?, so ffdshow-tryouts is essentially... dead?, what happened to all the developers?, so we can expect no new versions with nothing new in them? :confused:

Yup Stephen is correct, it's basically dead. There haven't been "developers" for ffdshow in a long while, a handful of people fixing small issues, but mostly just updates to Libav.

[fastplayer]

Could you please point me to where I can find the installation script?
Look for ffdshow_installer.iss in /trunk/bin/distrib.

[kirakami]

so ffdshow really no more
& here i thought it will reach r4000

[ikarad]

AssRender.dll is an Avisynth plugin. You can't register it or use it in DirectShow, or anything like that.

As for ffdshow needing a new subtitle renderer, there's nothing stopping you from using the MPC-HC one or DirectVobSub. ffdshow development has halted and it is already planned to be replaced by LAVFilters anyway, so there's no sense in petitioning for feature changes.

Directvobsub is dead since 2008 and doens't support bluray subs.
MPc-HC sub rendere is dead since many years and no developpers works on sub renderer in mpc-hc.


If ffdshow is dead (how do you know that it's dead?), there is not alternative.

I doubt that lavfilters replace ffdshow because it's not the same thing. lav filters is just for decode audi oand video but there is no filters and no support of avisynth.

[ikarad]

No. There are currently no developers working on ffdshow.

Why?

is it too difficult?

one year ago only, there were lots of progress with ffdshow (support of HD subs, support of dxva, support of HD audio, support of ffmepg-mt etc...)

Even some developpers like albain speak about the fusion between mpc-hc and ffdshow

[kirakami]

can't ffdshow keep going at least until r4000?
there r still issues with subtitles rendering
Vobsub hangs
still no support for SSA/ASS subs version 3 & 4 yet.

[nevcairiel]

Its not a matter of choice, there is just no-one capable and willing to work on it.

[Thunderbolt8]

what about these directvobsub builds, arent they working rather well? http://xhmikosr.1f0.de/index.php?folder=bXBjLWhjL21wYy1oY19hcHBzL3ZzZmlsdGVy

anyone able to explain why ffdshow is going to be discontinued? are its features superfluous now or kind of included in other tools as well and people dont want to waste time for the same thing twice?

if so which projects are going to replace (most of) the functions ffdshow ? LAVsplitter? mpc-hc?

[pirlouy]

Bah. Don't be that sad. One day, someone will feel a need for the same thing as you, and he will develop something. Think about Madshi, Nevcairiel; some time ago, nobody heard of those. Now you have good splitters, audio & video decoders, and a good renderer. In some time, someone will work on a good subtitle renderer (but maybe the best solution if it's developed by the one who works on renderer), and who knows, someone will rewrite post-processing stuff...

[hoborg]

Yes, subtitles need some good solution. But it should work undepended on aplication.

[Keiyakusha]

Bah. Don't be that sad. One day, someone will feel a need for the same thing as you, and he will develop something. Think about Madshi, Nevcairiel; some time ago, nobody heard of those. Now you have good splitters, audio & video decoders, and a good renderer. In some time, someone will work on a good subtitle renderer (but maybe the best solution if it's developed by the one who works on renderer), and who knows, someone will rewrite post-processing stuff...

Noone really interested in subtitles that much cause there is no reliable standard to begin with. There was attempts by aegisub devs to start the project on new fast crossplatform renderer but it was put in hold until standard fully developed and finalized. The problem is, they don't want to release any specifications till its done so noone will implement partial support in vsfilter or something (which is right thing). So someone need to motivate them i guess...

[Thunderbolt8]

so which are the lastest available versions of ffdshow and which problems do they still have? 3972 can be found on xvidvideo.ru

[clsid]

ffdshow has not been discontinued. There simply won't be any significant changes for the foreseeable future. That will remain the case until new developers join the project.

Latest builds can be found on the SourceForge project page.

[Midzuki]

so which are the lastest available versions of ffdshow and which problems do they still have? 3972 can be found on xvidvideo.ru

r3979 can be gotten from Dirk Paehl's site, or from
software d0t vladbazar d0t com :-/

[Thunderbolt8]

theres no ICL12 version though, arent these supposed to be faster for multithreading intel CPUs?

[Gser]

theres no ICL12 version though, arent these supposed to be faster for multithreading intel CPUs?

Doesn't make much of a difference since its quite light weight to begin with. The only thing that doesn't work fast enough with my i7 is SPP deblocking. But I use deblock QED anyways.

[clsid]

ICL12 does not have any real-life advantages. It only helps in very specific cases that do not apply to the majority of people.
Get the latest build from SourceForge and you will have a proper build with working deband. Builds you get elsewhere are possibly broken.

SPP deblocking strength can be adjusted with the slider on the PP page. Move the slider to the left to lower CPU usage.

[Thunderbolt8]

well I use sharpening and sometimes deinterlacing. if it helps with those, then I'd need it

[clsid]

Probably little difference for those.

[Thunderbolt8]

little is better than nothing. no multithreading for VC-1, no DXVA with interlacing, no nothing. I need every bit I can get

[nevcairiel]

little is better than nothing. no multithreading for VC-1, no DXVA with interlacing, no nothing. I need every bit I can get

Then go buy the compiler and built it yourself.

[Thunderbolt8]

nöööö

[kirakami]

Real Audio Cook Decoder Codec is Buggy
audio while playing creates dub..dub..dub sound

[j19861986]

I don't know if I can post this....
I have modified an IDCT matrix that can be loaded to FFDSHOW that you might try out... it is modified and used against a calibrated screen (with hardware calibrator)...
It is developed by me over one year, that makes it good, I think it is quite precise...
It works very well generally... for all video formats
sure to be check it out?
I have uploaded it in depositfiles.com
please check this out for me
her is the link to IDCT Matrix
JT_IDCT_0.98.xcm (http://depositfiles.com/files/50c1b4e60)
Want the developers to FFDSHOW to check it out, and requesting the file to be available in the "custom matrices" folder (when installed FFDSHOW) and Call it like "Precise IDCT Matrix" or something... :D

[j19861986]

Little Guide How to use the IDCT Matrix:
Open up FFDSHOW Video Decoder
Click the "DCT" in the left menu and check mark it..
Check mark from the upper window that says "Process Whole Image"
then select: "MPEG quantization" and Quantizer to "1" in the main screen
then click "Quantization Matrix" in the main screen
new window will come up, and click "Load matrix..."...
select the IDCT Matrix that you have downloaded and load it...
The load screen will close, and click close to the Quantization Matrix Window...
And Apply in the Main Screen
Then you have applied the IDCT Matrix

[kirakami]

What are proper steps to Enable DXVA Decoder Configuration for H.264 with subs & how do i confirm it?
using ffdshow r3978.

[Gser]

I don't see why we would need to apply a IDCT again.

[LigH]

IDCT = "Inverse Discrete Cosine Transformation" is applied while decoding. If you want to use a quantization matrix while encoding, omit the "I" = "Inverse".

Waiting 60 seconds to download 128 bytes - how lame ... you could have exported the matrix via CQME (http://forum.doom9.org/showthread.php?t=73666) and pasted its text representatipon in any pastebin.

Hmm, I wonder why such a matrix should be "better" than others, it is almost flat, therefore won't reduce high frequencies which are usually mostly part of noise. Especially I wonder if you knew what you did, because the Intra coefficient (0,0) is not 8 in this matrix, and the Inter matrix is not coarser than the Intra matrix (which is usually useful because interframe quantization errors are less persistent).

If you are looking for finer matrices to preserve more details, yet with compliance to general rules of DCT quantization matrix dynamics, you may try instead:

Didée's SixOfNine (and the HVS variant)
Fox Home Entertainment (although that one is a bit exaggerated for the Inter part)
Professional
Selur - high datarate

from my matrix collection.

[Sven75]

Just a bug report...

Crashing with ffdShow, but working fine with current ffplay.exe:

MS RLE (http://www.mediafire.com/?c4y5wyow7n233y7)

[j19861986]

IDCT = "Inverse Discrete Cosine Transformation" is applied while decoding. If you want to use a quantization matrix while encoding, omit the "I" = "Inverse".

Waiting 60 seconds to download 128 bytes - how lame ... you could have exported the matrix via CQME (http://forum.doom9.org/showthread.php?t=73666) and pasted its text representatipon in any pastebin.

Hmm, I wonder why such a matrix should be "better" than others, it is almost flat, therefore won't reduce high frequencies which are usually mostly part of noise. Especially I wonder if you knew what you did, because the Intra coefficient (0,0) is not 8 in this matrix, and the Inter matrix is not coarser than the Intra matrix (which is usually useful because interframe quantization errors are less persistent).

If you are looking for finer matrices to preserve more details, yet with compliance to general rules of DCT quantization matrix dynamics, you may try instead:

Didée's SixOfNine (and the HVS variant)
Fox Home Entertainment (although that one is a bit exaggerated for the Inter part)
Professional
Selur - high datarate

from my matrix collection.

Here is the export (intra first, inter last) I made in text:
8,18,18,19,20,22,21,23
17,17,19,19,22,22,23,24
18,19,21,20,22,23,23,25
19,21,20,22,23,24,26,26
20,22,22,24,26,27,27,28
20,21,23,23,27,28,28,30
22,23,23,25,27,28,30,31
23,25,24,28,29,31,32,34

15,18,18,19,20,22,21,23
17,17,19,19,22,22,23,24
18,19,21,20,22,23,23,25
19,21,20,22,23,24,26,26
20,22,22,24,26,27,27,28
20,21,23,23,27,28,28,30
22,23,23,25,27,28,30,31
23,25,24,28,29,31,32,34

Save it in text file (txt file)
Import via CQME (http://forum.doom9.org/showthread.php?t=73666), just check-mark "Text (Comma......", then the import screen comes up, select files of type to "Text files", load the text file you saved, import options: just check-mark "Comma" and click "OK", and "OK" again in "Text (Comma...." check-marked
Save it in xcm file
load it then in ffdshow
(one thing is that the intra's first number isn't the same as the inter's first one), the modification is from ffdshow idct modifier... and saved the xcm from FFDSHOW, is it an issue or what? suppose to be same?

[Midzuki]

...

(one thing is that the intra's first number isn't the same as the inter's first one),

Many encoders/decoders will refuse an intra-matrix whose very-first coefficient is not 8. mpeg2enc is one of the rare applications that ignore this "rule".

[LigH]

CQME assumes the Intra coefficient (0,0) always being 8. Many MPEG2 encoders assumed so as well. It is a kind of common level, basis of comparison, for the other matrix coefficients.

If anyone will prove to me that this is not necessary, and there are encoders which respect different values in this coefficient (especially Xvid does), I may publish a new version which releases this restriction.

Nevertheless, my doubts persist that this matrix is so remarkably useful as it is. Quoting it here was not outstanding by itself; adding a link to its origins and to a description of its purpose would be very appreciated.
__

Wait ... is that a matrix for MPEG4-AVC (H.264)? -- That one is probably not optimized for use with DCT in MPEG2 or MPEG4-ASP.

[kirakami]

request support for ASS subs v3/4.
why ASS ver 3/4 subs renderer still not added to ffdshow?

[kirakami]

ffdshow r3981 fails to decode 10bits depth all green screen :(
r3978 works fine

[clsid]

request support for ASS subs v3/4.
why ASS ver 3/4 subs renderer still not added to ffdshow?There are no developers available that are interested in working on the subtitle functionality. You should use DirectVobSub instead.

ffdshow r3981 fails to decode 10bits depth all green screen :(
r3978 works fineFixed.

[kirakami]

There are no developers available that are interested in working on the subtitle functionality. You should use DirectVobSub instead.

Fixed.

already using DirectVobSub for years
if ffdshow doesn't render ass v3/4
same happens DirectVobSub (using latest ver 1.5.3.3644)
only potplayer properly

wont merging libass Render into ffdshow work?
http://srsfckn.biz/assrender/

[clsid]

I will repeat it again: there are no developers

[Midzuki]

wont merging libass Render into ffdshow work?
http://srsfckn.biz/assrender/

That's the site of lachs0r alright — so,

¿ why don't you use his mplayer2 instead ? :confused: :confused: :confused:

[sneaker_ger]

There also seems to be some opposition to new ass extensions:
http://blog.aegisub.org/2010/02/old-and-present-vsfilter.html

[Keiyakusha]

There also seems to be some opposition to new ass extensions:
http://blog.aegisub.org/2010/02/old-and-present-vsfilter.html

Maybe i'm wrong but i guess ass3/4 means ScriptType: v4.00+ which is nothing new and this is the default format of the aegisub. Just ffdshow's subtitles filter is broken compared to vsfilter

[sneaker_ger]

Ah, ok, I thought he talked about some extensions. Well, it's dead. libass+madVR with MPC-HC would of course be nice.

[nevcairiel]

The main issue is that most ASS scripts are designed for vsfilter, and designed around all the bugs in vsfilter, therefor probably not looking 100% correct in libass