Avisynth + ffavisynth.dll not working?

[NoX1911]

Mpeg2Source("movie.d2v")
LoadCPlugin("C:\AviSynth\Plugins\ffavisynth.dll")
ffdshow("default")
LanczosResize(512,384)
converttoyv12

FFDshow doesn't connect to the filtergraph. What's wrong?

ffdshow_rev1280_20070612_clsid_sse_icl9.exe
Avisynth_257.exe (070518 alpha tested as well)
DGIndex v1.4.9
Vista32

[Leak]

Mpeg2Source("movie.d2v")
LoadCPlugin("C:\AviSynth\Plugins\ffavisynth.dll")
ffdshow("default")
LanczosResize(512,384)
converttoyv12
FFDshow doesn't connect to the filtergraph. What's wrong?
I just used ffdshow("default") successfully to add the mother of all snowstorms over an AVI file with ffdshow's noise filter, so I'm pretty sure ffavisynth.dll isn't broken.

I do believe you wrote in the ffdshow thread (BTW - don't crosspost, the mods don't like that) that you wanted to use ffdshow's SPP post processor this way.

That's why it does nothing - the post processing filters can't be used via ffavisynth as they're then missing the crucial quantizer info they need to work. Enable the post processing you want in ffdshow's VfW configuration and have it decode the file via AVISource instead. Or, in your MPEG2 case, via DirectShowSource...

np: Robert Babicz - Imperial Star (A Cheerful Temper)

[NoX1911]

Thx for the explanation.

I deleted the other post. I didn't know where to put it.

[henryho_hk]

Enable the post processing you want in ffdshow's VfW configuration and have it decode the file via AVISource instead

This is not really a good workaround because the post-processing parameters should be tweaked for different files. In the past, the ffdshow filter was working exactly as NoX1911 mentioned. We could even define different presets for different clips and that was really convenient for batch encoding. Unfortunately, it seems not working in the recent ffdshow tryout builds anymore.

[Leak]

This is not really a good workaround because the post-processing parameters should be tweaked for different files.
I agree that it seems to be not working at all currently, but what makes you think blind postprocessing (with the postprocessor not having any information about the frame's quantizers) is better than postprocessing with that information?

And using ffdshow(...) is always blind postprocessing - I'm quite sure you can't even set a fixed quantizer that way.

[Fizick]

BTW,
today I try ffavisinth at first time and found that "option" parameter does not work with examples provided with avisynth offline doc. I discover that small problem was a space around of comma delimiter.
ffdshow(options="isLevels=1, levelsMode=0, levelsGamma=1500") # does not work

ffdshow(options="isLevels=1,levelsMode=0,levelsGamma=1500") # works fine

[henryho_hk]

...but what makes you think blind postprocessing...

Actually, we had the "preset auto loading" feature to enjoy the best of both worlds. Unfortunately, it seems to be broken too in the recent tryouts.

[Leak]

Actually, we had the "preset auto loading" feature to enjoy the best of both worlds. Unfortunately, it seems to be broken too in the recent tryouts.
I guess we're talking about two different things here:

I was saying that using ffdshow("presetname") for post-processing a video opened with Mpeg2Source (if it were working) is going to be inferior to postprocessing when decoding the same video with ffdshow directly as the quantizer information for each macroblock (and/or frame) just isn't available to ffdshow when it isn't decoding the video - it only has the decoded image, but no meta-information to work on. And you need to know the quantizer to make a good decision on how much postprocessing to apply on each macroblock.

You were saying that different files could be treated differently with preset auto-loading (which is true, of course) and that preset auto-loading seems to be broken recently (which I can't really comment on, since I don't use it). But preset auto-loading has nothing to do with using the ffdshow("...") filter in an AviSynth script since you have to specify a preset name or filter options yourself anyway. The ffdshow("...") filter can't even know the file's name, since there's no filename associated with a clip in AviSynth.

[henryho_hk]

Having dedicated presets is essential for me because I often do batch encodes and I need to apply different settings for individual files.

Actually, we have 4 options:

Option 1: FFDShow VFW + AVISource + "default"
Option 2: XviD VFW + AVISource + ffdshow(preset="xxx")
Option 3: FFDShow VFW + AVISource/OpenDMLSource + "xxx" (auto selection based on filename, etc.)
Option 4: FFDshow DS + DirectShowSource + "xxx" (auto selection)

I recall that Options 3 was somehow working in the past but not anymore. And I've just tested that Option 4 works for clsid's rev1309 build (maybe other builds too) using filename wildcard match.

[NoX1911]

FFAvisynth and VirtualDub plugin still don't work with DirectShowSource (ffdshow_rev3124_20091103_clsid_icl10.exe). 'Levels' does work though. VirtualDub Plugin even reads 'Milan Cutka'. Long time no hands on it i think.

Any ideas or suggestions how to post-process videos? I'm somewhat outta ideas at the moment.

[Boulder]

I'd rather use Deblock_QED for deblocking, if that's what you are looking for - or MCTemporalDenoise with deblocking turned on if you need denoising and deblocking.