Why do I keep getting "half-way audio" when processing

[brassplyer]

I run into this issue - in this case an approx 90 min vhs transfer. Convert the DV file to Huffyuv, no problem. Deinterlace via QTGMC via Avisynth via Virtualdub to a 59.94 frame rate, no problem.

However, when I take that deinterlaced video, apply some cleanup filters and upsize it to 1280 x 720 with Virtualdub, I find that there's only half of the audio. I.e. the last 50% of the audio is missing, which is reflected in the GUI of Vegas. It's not some dysfunctionality with Vegas because the 50% is missing no matter what it's loaded into. The time lengths of the audio and video are identical.

If I recombine the audio and video with Vdub selecting "audio from other source" creating a new file then everything is okay. But it's obviously a PITA extra step that shouldn't have to happen. Apparently there's something that needs to be done differently at the stage of upscaling the video. Simply selecting "audio from other source" during the Vdub upscaling doesn't help.

Any suggestions?

Thanks.

[_Al_]

What software do you use for that clean up?

[brassplyer]

What software do you use for that clean up?

QTGMC via Avisynth via Virtualdub for the deinterlace, Virtualdub for the other filters and the upscale/resizing. I've edited my original post to clarify this.

[smrpix]

Deinterlace with a doubled frame rate

Doubling the frame rate may have something to do with it.

Is your audio all there or cut off? Is the audio or video playing fast or slow?

[brassplyer]

Deinterlace with a doubled frame rate

Doubling the frame rate may have something to do with it.

Is your audio all there or cut off? Is the audio or video playing fast or slow?

What's there is normal speed and in sync but it just doesn't recognize the second half. It just cuts off. I have to recombine the audio as described in the op to have it all be there.

[_Al_]

Why don't you just do all at once in VirtualDub?

For that clean up, do you use Direct stream copy for audio?

[brassplyer]

Why don't you just do all at once in VirtualDub?

For that clean up, do you use Copy Audio stream?

Before I first noticed the problem I used "Source Audio" and "Direct stream copy". Then I tried "Copy audio from other source" which didn't fix the problem.

Hmm, should I have used "full processing" under audio as well as video? I didn't think I would need to since I wasn't making any changes to the audio - EQ etc. - within Vdub.

[smrpix]

Are you dragging the audio along for the whole process? You may be converting it to 59.94 causing only half of it to be recognized.

[_Al_]

But video is too 59.94,
Huffyuv have something that VirtualDub is not reading upon loading?

Interesting question with full processing mode, it is uncompressed in the first place, so it will create uncompressed audio as well, I guess that would be the same copy with full processing mode?

[brassplyer]

Are you dragging the audio along for the whole process? You may be converting it to 59.94 causing only half of it to be recognized.

It recognizes all the audio after the deinterlace & doubling of the frame rate. The problem only manifests itself after the upscaling step.

Why not just break it out early and add it back in vdub during your final upscale. Not exactly what you're looking for but it may save you a step.

If that's the only way to do it, though I was hoping it's just some setting I don't have in place that would fix the issue.

[_Al_]

If you have that clean filter for avisynth you can do the whole thing completely at once, you load script with filters and load it directly into some x264 encoder (meGui, or command line) you set it to give you BD compliant H.264 raw files. You mux video and audio with TsMuxer for BD. No menu though.

[poisondeathray]

Doesn't make any sense . If it's fine before, and you use direct stream copy for audio, nothing is changed in the audio

Exactly what "cleanup and upscaling" in vdub are you doing?

(You can even do the upscaling in avisynth (better quality, less aliasing) , and can probably find equivalent or better cleanup filters)

[fvisagie]

Doesn't make any sense . If it's fine before, and you use direct stream copy for audio, nothing is changed in the audio

Exactly what "cleanup and upscaling" in vdub are you doing?

(You can even do the upscaling in avisynth (better quality, less aliasing) , and can probably find equivalent or better cleanup filters)

Agreed.

If you really must/want to find out where things are going wrong, you have to test those "some cleanup filters and upsize it" steps one by one (at least that is where the problem seems to occur according to your description). Apply the first one only, save to disk from VirtualDub, test audio, if no problem add next one, etc.

[brassplyer]

Doesn't make any sense . If it's fine before, and you use direct stream copy for audio, nothing is changed in the audio

Exactly what "cleanup and upscaling" in vdub are you doing?

Noise reduction, Brightness/Contrast, HSV, sharpening. By upscaling I mean putting the 4:3 image into a 16:9 pillarboxed image and going to 720p.

(You can even do the upscaling in avisynth (better quality, less aliasing) , and can probably find equivalent or better cleanup filters)

I actually don't really have an issue with the basic quality of most of the filters, and with Vdub you have relatively simple, immediate access to viewing the results of adjustments of the parameters. Avisynth is kind of a PITA to use in that respect.

I do wonder if there's a significantly better noise reduction filter for Avisynth or if there's only so much you can do without making the image look artificial.

[poisondeathray]

Exactly what "cleanup and upscaling" in vdub are you doing?

Noise reduction, Brightness/Contrast, HSV, sharpening. By upscaling I mean putting the 4:3 image into a 16:9 pillarboxed image and going to 720p.

I was asking to rule out if a specific filter might be causing the audio issue - none of those operations you listed should account for your observations

(You can even do the upscaling in avisynth (better quality, less aliasing) , and can probably find equivalent or better cleanup filters)

I actually don't really have an issue with the basic quality of most of the filters

haha that's what most people say about yadif and other deinterlacers . It's only when they look at the alternatives like QTGMC that they become "Converts"

Poor deinterlacers do a bad job of interpolating the missing information in interlaced content (that's why you get jaggies, "marching ants" in the first place) , but the same token, an enlargement using a standard resizer will result in the same artifacts.

Think of it this way - when you bob deinterace you're essentially enlarging each field to become a full frame . But if you just enlarge it without all those other processing steps, QTGMC will be as poor as another deinterlacer . Adding sharpening , other post processing afterwards will just sharpen the aliasing artifacts . At least with nnedi_rpow2 - you have a cleaner upscale and fewer aliasing artifacts - so it will respond better to sharpening and other post processing steps

, and with Vdub you have relatively simple, immediate access to viewing the results of adjustments of the parameters. Avisynth is kind of a PITA to use in that respect.

Yes I agree with you for the most part - but in some ways the "ease of use" is actually worse in vdub. For example , can you compare different filter combinations /strengths in vdub simultaneously or very easily ? You adjust something, but you can't compare to an adjustment you made a minute ago. You have to take a screenshot or something. In avspmod - you can try out different filter combinations and flip back forth between tabs using the number keys in avspmod instantaneously .

I do wonder if there's a significantly better noise reduction filter for Avisynth or if there's only so much you can do without making the image look artificial.

I would say there is a wider range of filters and strategies with avisynth denoising filters. There is no question you can do a better job than using stock vdub filters. BUT the bottom line is - with a VHS source, there is only so much you can do.


But I have no idea what is causing the original problem. If you take that exported video and open it in vdub is there full audio ? It would seem more like incompatibility with vegas despite what your are saying . If the it works fine in everything else but vegas....wouldn't the problem be vegas ? Maybe I'm misinterpreting what you are saying ?

It's not some dysfunctionality with Vegas because the 50% is missing no matter what it's loaded into. The time lengths of the audio and video are identical.

Also, why are you converting to huffyuv then deinterlacing with qtgmc ? Why not load the original DV ? and do all the other processing in 1 step as suggested above? Seems like you're doing a few extra unnecessary steps? The more convoluted your workflow - not only wastes your time and HDD space - it increases the risk of something inadvertently going wrong .

[budwzr]

You keep mentioning "PITA". Is that the flatbread, or the animal rights group?

And just for fun, why don't you demux the audio before processing? And see what's there. Maybe it's no-good from jump.

[brassplyer]

, and with Vdub you have relatively simple, immediate access to viewing the results of adjustments of the parameters. Avisynth is kind of a PITA to use in that respect.

Yes I agree with you for the most part - but in some ways the "ease of use" is actually worse in vdub. For example , can you compare different filter combinations /strengths in vdub simultaneously or very easily ? You adjust something, but you can't compare to an adjustment you made a minute ago. You have to take a screenshot or something. In avspmod - you can try out different filter combinations and flip back forth between tabs using the number keys in avspmod instantaneously .

What I've been doing to A/B is saving filter chains - "save processing settings" and just loading them. Takes a couple of clicks. Happens pretty much instantly. A set of preset buttons would be even faster but it's not too bad.

The other thing is besides the individual filter adjustments, the order of the filters can make a big difference.

If you take that exported video and open it in vdub is there full audio ? It would seem more like incompatibility with vegas despite what your are saying . If the it works fine in everything else but vegas....wouldn't the problem be vegas ? Maybe I'm misinterpreting what you are saying ?

No, the audio is missing no matter what I load the video in.

Also, why are you converting to huffyuv then deinterlacing with qtgmc ? Why not load the original DV ? and do all the other processing in 1 step as suggested above? Seems like you're doing a few extra unnecessary steps? The more convoluted your workflow - not only wastes your time and HDD space - it increases the risk of something inadvertently going wrong .

Doesn't seem to work the same. Seems less problematic to take the video that's been given the base preparation of converting to Huff, deinterlaced and going from there.

[poisondeathray]

No, the audio is missing no matter what I load the video in.

ok, can you use gspot or mediainfo(view=>text) , what do they say about the audio duration ? how does the report compare with the other earlier steps

Also, why are you converting to huffyuv then deinterlacing with qtgmc ? Why not load the original DV ? and do all the other processing in 1 step as suggested above? Seems like you're doing a few extra unnecessary steps? The more convoluted your workflow - not only wastes your time and HDD space - it increases the risk of something inadvertently going wrong .

Doesn't seem to work the same. Seems less problematic to take the video that's been given the base preparation of converting to Huff, deinterlaced and going from there.

Ok, but you can still skip the 1st step. Do you see that's it produces identical results, fewer steps, less time and space ? Avisynth frameserves uncompressed video.

DV=>QTGMC=>HUFFYUV => farther processing

DV=>HUFFYUV=>QTGMC=>HUFFYUV => farther processing

Now there are times and valid reasons, when you might want to break out steps . The 1st step isn't one of those times.

[budwzr]

Seems to me that Vegas is "deinterlacing" the audio too. Did you match the project settings to the source? Is "Adjust Media To Match Project" selected?

The problem with using utilities is they will allow you to screw things up, and remain silent. So you have been hoist by your own petard.

[brassplyer]

DV=>QTGMC=>HUFFYUV => farther processing

DV=>HUFFYUV=>QTGMC=>HUFFYUV => farther processing

My reason for going to Huff in the first place is of course to avoid generational degradation. Do you feel a deinterlace done on the DV file before converting to Huff as per the first example above will produce identical results to a deinterlace done on a file converted to Huff first?

[_Al_]

Think of it this way, Avisynth loads DV.avi directly , video has to be decoded to uncompressed YUV, then QTGMC filter comes . So at the begging using that Huffyuv is useless step. That deinterlace within Avisynth is happening on uncompressed level anyway.

AviSource("C:\DV.avi")
AssumeBFF()
QTGMC( Preset="slow" )
.
.
.

[mike20021969]

You keep mentioning "PITA".

Twice is hardly keep mentioning.