Log in

View Full Version : ffdshow tryouts project: Discussion & Development


Pages : 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 [151] 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308

jmartinr
29th May 2009, 23:21
I know its off topic but do you know of a tool to edit the mpeg2 screen AR flags that must be set incorrectly in alot of my mkv's now. Will be nice to fix them all.

Dvdpatcher (http://www.wincesoft.de/html/dvdpatcher.html) maybe? You'll have to demux the mkv's first of course.

tetsuo55
29th May 2009, 23:40
Dvdpatcher (http://www.wincesoft.de/html/dvdpatcher.html) maybe? You'll have to demux the mkv's first of course.That tool looks good, i hope it works.

It's not 100% off topic we basically concluded that to get the file working withing FFdshow it has to be physically changed.

haruhiko_yamagata
29th May 2009, 23:49
haruhiko, rev2976 makes seeking slower. Are dropped frames on seeking really a problem?
Yeah, it's slower. But it was really a problem.
Try this (http://x264.nl/h.264.samples/premiere-paff.ts) with Haali's splitter, with 8threads. Haali's splitter drops non ref frames until the first image is delivered to the video renderer. If you use 8 threads, frames are heavily dropped.

mark0077
30th May 2009, 00:26
Dvdpatcher (http://www.wincesoft.de/html/dvdpatcher.html) maybe? You'll have to demux the mkv's first of course.

Excellent, well just a final note on this, dvdpatcher set to patch the "entire file" did the trick. Strangely though the only combination of "Aspect Ratio:" and "Size:" that gave fullscreen 16\9 with a DVD I am using to test, was 4:3 and 720*576.

I'm a bit skeptical about these settings... it works and looks fine when passing through ffdshow... but this need to set to 4:3 and an optional bitrate setting in dvdpatcher has me worried I will ruin all of these mpeg2 files of mine :P I would assume the bitrate is just for information purposes rather than effecting quality? Thanks for the input guys, learning alot, bloody DVD's and their messed up AR's.

EDIT: Again when setting the entire mpeg2 file (no mkv containers involved), to 16/9 and 720x576, ffdshows decoders gets the AR wrong, it adds black bars top and bottom. Using mpc-hc's decoders, this plays perfectly fullscreen. I still think ffdshow has a bug somewhere. Setting the entire mpeg2 file to 16/9 and 720x576 with no mkv container shouldn't show any black bars in ffdshow decoders but it does.

haruhiko, with the entire mpeg2 stream set to 16/9 and 720x576 with no mkv containers to mess things up, and with ffdshow mpeg2 decoders, shouldn't I get 16/9 fullscreen? It should know from this that the 720x576 needs to be stretched from 1.25:1 to 16/9 right? Instead it gets displayed at what looks like 1.25:1 WITHIN 16/9 container which adds the black bars, basically ignoring the stretch to fill 16/9. Basically if its 16/9 throughout the mpeg2 file, and I use 16/9 screen I shouldn't see black bars if they arn't already encoded into the stream of course. Seriously looks like a bug to me. Let me know if you need a sample if you want to look into it. Thanks for your time on this issue.

tetsuo55
30th May 2009, 00:55
I have been reading a lot about mkv today and it seems mkv passes on whatever the video stream tells the muxer.

So the stream info and mkv container info will always match.
If DVD needs special treatment the mpeg2 files will have to be edited before muxing to mkv.

However what mark just mentioned does sound like a bug to me

Snowknight26
30th May 2009, 01:21
So the stream info and mkv container info will always match.

Not always the case.

the_corona
30th May 2009, 06:11
Is there a way to do Motion Interpolation in FFDShow, kinda like MotionFlow on TV's? Obviously not to 120hz, but maybe up to 60hz or 50hz.

Only way I found was using AVISynth scripts, but I would guess a "native" method would be alot faster?

Ever since I saw that Tech at a friends TV I've been trying to recreate it, but without much sucess.

PS: don't want to hear the old "but thats not how its supposed to look preachings" :-)

tetsuo55
30th May 2009, 09:00
Not always the case.it will only not match if you used a muxer that allows manual override(and you entered a value yourself) or the muxer had a bug.

Default behaviour is copy stream info

STaRGaZeR
30th May 2009, 14:07
Yeah, it's slower. But it was really a problem.
Try this (http://x264.nl/h.264.samples/premiere-paff.ts) with Haali's splitter, with 8threads. Haali's splitter drops non ref frames until the first image is delivered to the video renderer. If you use 8 threads, frames are heavily dropped.

I have no dropped frames with that sample with Haali's splitter and ffmpeg-mt with 4 or 8 threads, playback is just fine using rev2975. In fact with rev2976 I get garbled frames at the beginning, and after second 10 or so there is a lot of stuttering and dropped frames with VMR9 and EVR Custom.

haruhiko_yamagata
30th May 2009, 14:57
I have no dropped frames with that sample with Haali's splitter and ffmpeg-mt with 4 or 8 threads, playback is just fine using rev2975.After seeking, within first one or two seconds, several frames are dropped.

In fact with rev2976 I get garbled frames at the beginning, and after second 10 or so there is a lot of stuttering and dropped frames with VMR9 and EVR Custom.
Please rebuild ffdshow.ax.

STaRGaZeR
30th May 2009, 17:05
After seeking, within first one or two seconds, several frames are dropped.

I can't reproduce. Only 1 or 2 frames are dropped here with 2975 when seeking, and not always. No big deal IMHO. Same with other files. With 2976 0 are dropped.

Please rebuild ffdshow.ax.

That worked, thanks.

Snowknight26
30th May 2009, 19:52
ffdshow seems to crash 3-4 seconds into playing this VC-1 stream from the Cloverfield Blu-ray using libavcodec. wmv9 is not affected; r2975.

http://stfcc.org/misc/cloverfield.sample.m2ts - first 20,000,000 bytes.

ikarad
31st May 2009, 20:27
1st question: 3 months ago, only mlp decoder existed in ffdshow and allowed to decode trueHD streams. But I notice that now trueHD decoder also exists and I d'ont know the difference between the two decoders.

what is the difference between "mlp" decoder and "trueHD" decoder? Who is the best?

2nd question
to decode ac3 there are two decoders in ffdshow audio: liba52 and libavcodec.
What is the difference between the two decoders?

clsid
31st May 2009, 22:52
MLP and TrueHD are two different (but very similar) audio formats.

ikarad
1st June 2009, 08:24
MLP and TrueHD are two different (but very similar) audio formats.

What is the best (the best sound quality)?

FredThompson
1st June 2009, 09:56
Nope, I mean CPU in ffdshow.
It is placed in "ffdshow (DirectShow) video decoder configuration", Info & CPU page.

I have tested simultaneous reading/writing of HuffYUV with ffdshow and various combinations of the CPU flags as you suggested. It appears MMX must be enabled for decoding HuffYUV. When that is the ONLY flag enabled, the erroneous right-hand "empty" blocks are created with ffdshow_rev2975_20090528_clsid. Is that helpful information?

clsid
1st June 2009, 11:45
What is the best (the best sound quality)?http://en.wikipedia.org/wiki/Dolby_TrueHD

haruhiko_yamagata
1st June 2009, 12:06
I have tested simultaneous reading/writing of HuffYUV with ffdshow and various combinations of the CPU flags as you suggested. It appears MMX must be enabled for decoding HuffYUV. When that is the ONLY flag enabled, the erroneous right-hand "empty" blocks are created with ffdshow_rev2975_20090528_clsid. Is that helpful information?I still can't reproduce.
Could you try with this sample (http://ffdshow-tryout.sourceforge.net/doom9/huffyuv800x600.avi)?
The file plays fine for me.

FredThompson
2nd June 2009, 00:37
I still can't reproduce.
Could you try with this sample (http://ffdshow-tryout.sourceforge.net/doom9/huffyuv800x600.avi)?
The file plays fine for me.I still have the problem. This happens even if I use GSpot to raise the ffdshow VFW preference to preferred.

Maybe there is a conflict somewhere. I'm a little puzzled why ffdshow_rev2645_20090131_sse_icl10 works and later versions do not.

Have you looked at the code base to see if anything might have changed unintentionally; typo, compiler change or something like that?

Here is my list of installed codecs.

clsid
2nd June 2009, 11:18
I don't know what exactly you did in GSpot, but VFW does not have a merit system.

haruhiko_yamagata
2nd June 2009, 12:48
I still have the problem. This happens even if I use GSpot to raise the ffdshow VFW preference to preferred.

Maybe there is a conflict somewhere. I'm a little puzzled why ffdshow_rev2645_20090131_sse_icl10 works and later versions do not.

Have you looked at the code base to see if anything might have changed unintentionally; typo, compiler change or something like that?

Here is my list of installed codecs.
Thank you for testing. But excuse me, CPU flags control from GUI was broken. I have fixed it at rev 2982. Please wait for a next build and try again.
Just after rev 2645, SIMD code has been changed.
To confirm that ffdshow vfw decoder is being used, you could enable level filter and make it very bright or something.

FredThompson
2nd June 2009, 23:50
Thank you for testing. But excuse me, CPU flags control from GUI was broken. I have fixed it at rev 2982. Please wait for a next build and try again.
Just after rev 2645, SIMD code has been changed.
To confirm that ffdshow vfw decoder is being used, you could enable level filter and make it very bright or something.OK, I will test again. VirtualDub 1.8.8 only supports VFW so there should be no question that VFW is the method. We're looking for verification of a problem in SIMD code?

FredThompson
2nd June 2009, 23:58
I don't know what exactly you did in GSpot, but VFW does not have a merit system.If you look at the list of installed codec with GSpot there is a VFw category. Those were all set to 20. I set ffdshow to default, 80, I think, something like that. I've had to do this with quartz and past builds of Sound Forge which trumped it, thus preventing PowerPoint and other old VFW methods of embedded video from working. Haven't had to do that in a long time so I guess Sony fixed that problem. It was just something I thought of from years ago that helped with clashes. Basically, I went looking for conflicts which might be there. Found that stupid Pixie pack and removed it, too. TeamViewer, SUPER or MediaCoder installs that piece of junk.

haruhiko_yamagata
3rd June 2009, 00:06
If you want to confirm ffdshow is creating the black blocks, the easiest way is to disable huffyuv in VFW decoder configuration.

FredThompson
3rd June 2009, 03:33
If you want to confirm ffdshow is creating the black blocks, the easiest way is to disable huffyuv in VFW decoder configuration.I don't have the HuffYUV codec installed because ffdshow supports the format. Good idea, though.

clsid
3rd June 2009, 11:56
If you look at the list of installed codec with GSpot there is a VFw category. Those were all set to 20. I set ffdshow to default, 80, I think, something like that. I've had to do this with quartz and past builds of Sound Forge which trumped it, thus preventing PowerPoint and other old VFW methods of embedded video from working. Haven't had to do that in a long time so I guess Sony fixed that problem. It was just something I thought of from years ago that helped with clashes. Basically, I went looking for conflicts which might be there. Found that stupid Pixie pack and removed it, too. TeamViewer, SUPER or MediaCoder installs that piece of junk.
That must be a hidden option in the Registry then. I know ACM codecs can be set in a preference order in the Device manager, but there is no such option in there for VFW codecs.

Blight
4th June 2009, 19:17
Hi Devs,
I'm trying to implement ffdshow as a subtitle filter in Zoom Player.
I'm using the COM API (IffDecoder) in order to interact with ffdshow settings.

To get the basics down first, what I'm trying to do is play a video from one directory and then instruct ffdshow to load a subtitle file from a different directory.

What I've done so far:
* Enable subtitle support in ffdshow (working):
IffDecoder.putParam(IDFF_isSubtitles,Integer(True));

* Set ffdshow to file mode (working):
IffDecoder.putParam(IDFF_subAutoFlnm,Integer(False));

* And now the difficult part, passing along the subtitle file name to FFDShow. Since this is a string rather than a boolean, I've tried using "IffDecoder.putParamStr(IDFF_subFilename,SubFileName)" function, but the specs don't mention the character format. I tried ANSI, UNICODE and UTF8, all fail (the file field isn't filled with the file name and no subtitle is displayed).

I also tried calling this after setting the file name:
IffDecoder.putParam(IDFF_subFlnmChanged,Integer(True));

But it really didn't make a difference. What am I missing?

Blight
4th June 2009, 19:27
clsid:
Is it by design that the daily builds have a lower version number than beta 6?

ffdshow_rev2981_20090601_clsid.exe = 1.0.5.2945
ffdshow_beta6_rev2527_20081219.exe = 1.0.6.2524

clsid
5th June 2009, 11:44
Version is indeed wrong. I'll fix that.

Blight
5th June 2009, 11:57
clsid:
What about loading the subtitles, how can it be done?

clsid
5th June 2009, 12:40
Sorry, I can't help you with that.

haruhiko_yamagata
5th June 2009, 12:42
Hi Devs,
I'm trying to implement ffdshow as a subtitle filter in Zoom Player.
I'm using the COM API (IffDecoder) in order to interact with ffdshow settings.

To get the basics down first, what I'm trying to do is play a video from one directory and then instruct ffdshow to load a subtitle file from a different directory.

What I've done so far:
* Enable subtitle support in ffdshow (working):
IffDecoder.putParam(IDFF_isSubtitles,Integer(True));

* Set ffdshow to file mode (working):
IffDecoder.putParam(IDFF_subAutoFlnm,Integer(False));

* And now the difficult part, passing along the subtitle file name to FFDShow. Since this is a string rather than a boolean, I've tried using "IffDecoder.putParamStr(IDFF_subFilename,SubFileName)" function, but the specs don't mention the character format. I tried ANSI, UNICODE and UTF8, all fail (the file field isn't filled with the file name and no subtitle is displayed).

I also tried calling this after setting the file name:
IffDecoder.putParam(IDFF_subFlnmChanged,Integer(True));

But it really didn't make a difference. What am I missing?
I think you are using IffDecoder_com.pas in delphi directory.
Binary compatibility of that interface has been broken since 1126 (and again at rev 2852 which has been fixed at rev 2942). I'll fix this.

@developers: please do not add new lines in the midst of an interface. New lines must be added at the end of the interface (or more ideally create a new interface with a new IID).

Blight
5th June 2009, 15:22
haruhiko:
How about a very limited interface just for getting/putting config parameter values, basic control over the settings, something that wont ever break on newer builds.

In essence it would only need 2 functions (set/get) which will pass an identifier and a pointer to a pre-allocated memory block. If the pointer is null, the returned function value can be the buffer size required.

Then just write a doc listing what value each identifier expects (UTF8 String, unicode string, word, dword, float, etc...).

This will make things easier and would ensure forward compatibility. And the concept is nearly identical to the existing interface.

In any case, please let me know which build has this fixed so that I can implement the subtitle control. Also, does the code accept UTF8/Unicode strings, it's important for file names, but if it can't, I need to know so that I can pass the 8.3 filename for compatibility.

albain
5th June 2009, 18:08
Hi Devs,
I'm trying to implement ffdshow as a subtitle filter in Zoom Player.
I'm using the COM API (IffDecoder) in order to interact with ffdshow settings.

To get the basics down first, what I'm trying to do is play a video from one directory and then instruct ffdshow to load a subtitle file from a different directory.

What I've done so far:
* Enable subtitle support in ffdshow (working):
IffDecoder.putParam(IDFF_isSubtitles,Integer(True));

* Set ffdshow to file mode (working):
IffDecoder.putParam(IDFF_subAutoFlnm,Integer(False));

* And now the difficult part, passing along the subtitle file name to FFDShow. Since this is a string rather than a boolean, I've tried using "IffDecoder.putParamStr(IDFF_subFilename,SubFileName)" function, but the specs don't mention the character format. I tried ANSI, UNICODE and UTF8, all fail (the file field isn't filled with the file name and no subtitle is displayed).

I also tried calling this after setting the file name:
IffDecoder.putParam(IDFF_subFlnmChanged,Integer(True));

But it really didn't make a difference. What am I missing?

Hi,

the string format depends on how ffdshow has been compiled, so it should be unicode for windows 2000/XP/Vista/7.
I guess that it should work again when Haruhiko will have fixed the problem

Otherwise you can use the FFDShowAPI (based on windows messages) or FFDShowAPI.Net which contain all the primitives

Soulhunter
6th June 2009, 03:50
I know its already possible to switch profiles via hotkeys...

But, could you maybe add a option to assign/bind hotkeys to profiles?

haruhiko_yamagata
6th June 2009, 09:55
haruhiko:
How about a very limited interface just for getting/putting config parameter values, basic control over the settings, something that wont ever break on newer builds.
It's a good idea. Even if I fix the broken compatibility, it breaks compatibility for applications that are built for beta6 (non delphi). If a new interface exists, the application can know what to do.

In essence it would only need 2 functions (set/get) which will pass an identifier and a pointer to a pre-allocated memory block. If the pointer is null, the returned function value can be the buffer size required.

Then just write a doc listing what value each identifier expects (UTF8 String, unicode string, word, dword, float, etc...).

This will make things easier and would ensure forward compatibility. And the concept is nearly identical to the existing interface.

In any case, please let me know which build has this fixed so that I can implement the subtitle control. Also, does the code accept UTF8/Unicode strings, it's important for file names, but if it can't, I need to know so that I can pass the 8.3 filename for compatibility.
wchar_t* should work.

haruhiko_yamagata
6th June 2009, 13:29
Also, does the code accept UTF8/Unicode strings, it's important for file names, but if it can't, I need to know so that I can pass the 8.3 filename for compatibility.
Lol, it accepts only char*. I'll create a new one that accepts unicode.

_xxl
7th June 2009, 16:11
I have a question:
Why does ffdshow subtitles filter use 60-80 Mb of ram?
Tested with XP sp3, Amd Xp 2000+, fx5200, 512Mb ram, avi Xvid + mp3 sample + .srt subtitle, latest ffdshow 2985 ,MPC-HC, default video renderer.
Subtitles on:
http://i44.tinypic.com/2sacygh.jpg
Subtitles off:
http://i44.tinypic.com/33eiikz.jpg

avivahl
7th June 2009, 17:31
I have a question:
Why does ffdshow subtitles filter use 60-80 Mb of ram?
Tested with XP sp3, Amd Xp 2000+, fx5200, 512Mb ram, avi Xvid + mp3 sample + .srt subtitle, latest ffdshow 2985 ,MPC-HC, default video renderer.
Subtitles on:
http://i44.tinypic.com/2sacygh.jpg
Subtitles off:
http://i44.tinypic.com/33eiikz.jpg
Wow, I just checked it and same here. Vista+SP1. WMP11. SRT subtitle.

Mirabel
7th June 2009, 19:11
I'm sorry, no offense, but recent ffdshow builds really maul :eek::eek: Windows Explorer, Windows Media Center and WMP. They crash when it comes to thumbnail rendering (I have thumbs for FLV, MKV etc. enabled).

I tried it two times, the last build is ffdshow_rev2984_20090605_sse_icl10, I don't know which build was the first one that didn't worked. I assume it's from the 2nd part of May. I'm now using "ffdshow tryouts revision 2941 May 9 2009 02:47:13 (icl 10, x86, unicode, r)". Both times I had to use system restore, because installing an older build broke the entire thumbnail rendering, even of WMV files.

Explorer just stops rendering thumbs or crashes, the other mentioned apps crash.

Please fix it or I'll have to stick with what I current use. I use ffdshow for ages now, and never experienced that problems. I guess there are other apps which will encounter those problems.

I'm running Vista SP2, and I have this computer not longer than a year and I did not blow it up with codec packs or other stuff which may have broke the thumbnail stuff - reverting to the old ffdshow build fixes it as described. In my eyes it is safe to assume that the problem lies within ffdshow, sadly.

Atak_Snajpera
7th June 2009, 19:48
Could you check Generic r2975?

haruhiko_yamagata
8th June 2009, 11:01
haruhiko:
How about a very limited interface just for getting/putting config parameter values, basic control over the settings, something that wont ever break on newer builds.

In essence it would only need 2 functions (set/get) which will pass an identifier and a pointer to a pre-allocated memory block. If the pointer is null, the returned function value can be the buffer size required.

Then just write a doc listing what value each identifier expects (UTF8 String, unicode string, word, dword, float, etc...).

This will make things easier and would ensure forward compatibility. And the concept is nearly identical to the existing interface.

In any case, please let me know which build has this fixed so that I can implement the subtitle control. Also, does the code accept UTF8/Unicode strings, it's important for file names, but if it can't, I need to know so that I can pass the 8.3 filename for compatibility.
I have fixed IffDecoder to match delphi. It breaks binary compatibility for recent applications though.
I have added IID_IffDecoder2 which can be queried on an instance of DirectShow video decoder and returns IffDecoder. It can be used to check the version of IffDecoder.
If IID_IffDecoder2 is implemented, getIffDecoder2Version, getParamStrW and putParamStrW should work.
Test build and patch (http://ffdshow-tryout.sourceforge.net/doom9/ffdshow_rev2987_20090608_Q.zip).
Index: src/IffDecoder.h
===================================================================
--- src/IffDecoder.h (revision 2987)
+++ src/IffDecoder.h (working copy)
@@ -7,6 +7,8 @@

// {00F99063-70D5-4bcc-9D88-3801F3E3881B}
DEFINE_GUID(IID_IffDecoder , 0x00f99063, 0x70d5, 0x4bcc, 0x9d, 0x88, 0x38, 0x01, 0xf3, 0xe3, 0x88, 0x1b);
+// {5DD3A966-7365-46d0-B79F-D4973BD88E8D}
+DEFINE_GUID(IID_IffDecoder2, 0x5dd3a966, 0x7365, 0x46d0, 0xb7, 0x9f, 0xd4, 0x97, 0x3b, 0xd8, 0x8e, 0x8d);

struct Tpreset;
struct TglobalSettingsDecVideo;
@@ -66,8 +68,6 @@
STDMETHOD (compat_cpuSupportsMMXEXT)(void) PURE;
STDMETHOD (compat_cpuSupportsSSE)(void) PURE;
STDMETHOD (compat_cpuSupportsSSE2)(void) PURE;
- STDMETHOD (compat_cpuSupportsSSE3)(void) PURE;
- STDMETHOD (compat_cpuSupportsSSSE3)(void) PURE;
STDMETHOD (compat_cpuSupports3DNOW)(void) PURE;
STDMETHOD (compat_cpuSupports3DNOWEXT)(void) PURE;
STDMETHOD (compat_getAVIfps1000_2)(void) PURE;
@@ -132,6 +132,11 @@
STDMETHOD (compat_cpuSupportsSSE42)(void) PURE;
STDMETHOD (compat_cpuSupportsSSE4A)(void) PURE;
STDMETHOD (compat_cpuSupportsSSE5)(void) PURE;
+ STDMETHOD (compat_cpuSupportsSSE3)(void) PURE;
+ STDMETHOD (compat_cpuSupportsSSSE3)(void) PURE;
+ STDMETHOD (compat_getIffDecoder2Version)(void) PURE;
+ STDMETHOD (compat_getParamStrW)(unsigned int paramID,wchar_t *buf,unsigned int buflen) PURE;
+ STDMETHOD (compat_putParamStrW)(unsigned int paramID,const wchar_t *buf) PURE;
};

#ifdef __cplusplus

haruhiko_yamagata
8th June 2009, 11:19
I have a question:
Why does ffdshow subtitles filter use 60-80 Mb of ram?
Tested with XP sp3, Amd Xp 2000+, fx5200, 512Mb ram, avi Xvid + mp3 sample + .srt subtitle, latest ffdshow 2985 ,MPC-HC, default video renderer.
Subtitles on:
http://i44.tinypic.com/2sacygh.jpg
Subtitles off:
http://i44.tinypic.com/33eiikz.jpg

40MB is used for buffering image. I'm not sure about the rest, but ffdshow uses 10x or more of the text size. It's not very efficient, but we don't have trouble with 100MB on current PC.

haruhiko_yamagata
8th June 2009, 11:23
I'm sorry, no offense, but recent ffdshow builds really maul :eek::eek: Windows Explorer, Windows Media Center and WMP. They crash when it comes to thumbnail rendering (I have thumbs for FLV, MKV etc. enabled).

I tried it two times, the last build is ffdshow_rev2984_20090605_sse_icl10, I don't know which build was the first one that didn't worked. I assume it's from the 2nd part of May. I'm now using "ffdshow tryouts revision 2941 May 9 2009 02:47:13 (icl 10, x86, unicode, r)". Both times I had to use system restore, because installing an older build broke the entire thumbnail rendering, even of WMV files.

Explorer just stops rendering thumbs or crashes, the other mentioned apps crash.

Please fix it or I'll have to stick with what I current use. I use ffdshow for ages now, and never experienced that problems. I guess there are other apps which will encounter those problems.

I'm running Vista SP2, and I have this computer not longer than a year and I did not blow it up with codec packs or other stuff which may have broke the thumbnail stuff - reverting to the old ffdshow build fixes it as described. In my eyes it is safe to assume that the problem lies within ffdshow, sadly.
If libavcodec is selected for VC1, try disabling it.

haruhiko_yamagata
9th June 2009, 14:31
I had problems a few builds back with DTS over SPDIF creating a jerk (like a dropped frame) in the video every 6 seconds. I ended up using AC3Filter for DTS and it doesn't do that. My post about the issue I don't think were ever responded to. I haven't tried a recent build to see if it was fixed, but maybe someone was working on it and didn't complete the fix?

Cheers
The REAL Joe@tetsuo55 and therealjoeblow
I have fixed at rev 2990. Thank you for report and sorry for late reply.

tetsuo55
9th June 2009, 14:40
@tetsuo55 and therealjoeblow
I have fixed at rev 2990. Thank you for report and sorry for late reply.thanks, as soon as a 2990 build is out i will test. Also want to test the AR fixes asap.

therealjoeblow
9th June 2009, 15:50
@tetsuo55 and therealjoeblow
I have fixed at rev 2990. Thank you for report and sorry for late reply.

Thank you haruhiko, I'll test it a bit later.

Cheers,
The REAL Joe

Thunderbolt8
11th June 2009, 04:20
Then it alters timing. I think it is not a bug of ffdshow.
The graph state change from pause to playing when first video frame is delivered. It may be triggering a bug in ReClock. The timing first frame is delivered differs a great deal between libavcodec and ffmpeg-mt.
As the crashing module is ReClockHelper.dll, it is most likely a bug of ReClock. If ffmpeg-mt is causing delayed crash (ie breaking heap), it should crash even without ReClock.
I posted this on slysoft forums and the answer there was

"Just disable the DShow estimator."

so how can I do that and does this have any disadvantages?

haruhiko_yamagata
11th June 2009, 09:36
I posted this on slysoft forums and the answer there was

"Just disable the DShow estimator."

so how can I do that and does this have any disadvantages?
What is "DShow estimator"?

cca
11th June 2009, 11:03
What is "DShow estimator"?

That would be "Directshow estimator", it is a way to extract the fps of the video stream by building a directshow filter chain and query the appropriate filter for the fps. At least this is my understanding of it.