DeShaker use

[Tafflad]

Searching through the forum posts .. when using DeShaker .... a common piece of advice is to:
"Crank up the motion smoothing to 50000 to 90000"

There are 4 settings under motion smoothness

a.Horizontal Panning
b.Rotation
c.Vertical Panning
d.Zoom"

Does this advice mean add such a higher value to all 4 of these ? if not any guidance as to what ?

[jagabo]

Generally, all of them. But why don't you just try it yourself since VirtualDub is interactive.

[Tafflad]

Will do .... was just checking in case it was normal to keep same values for example for horizontal & vertical.

[Skiller]

Searching through the forum posts .. when using DeShaker .... a common piece of advice is to:
"Crank up the motion smoothing to 50000 to 90000"

I would not recommend such big values unless the video was shot on a tripod. Let me quote the Deshaker manual

If you enter -1 as smoothness you will get "infinite smoothness", meaning the camera will appear to be stationary at the position of the first frame that was processed in pass 1. This feature usually doesn't work perfectly, though.

if not any guidance as to what ?

For regular handheld shots I recommend 2 to 4 times the frame rate of the video. So for a 50 fps video that would be 100-200 for all 4 motion smoothness settings (although I never deshake zoom and leave it at 0). That's what I use and it's a good tradeoff, because if you go for more you will sometimes have very large borders/lots of zoom and stationary motion blur because the original motion blur of the video is still there but the related motion may be completely stabilized, and that looks just weird.

[jagabo]

Of course, the amount you use depends on the amount of shaking in the video.

Another problem with extreme deshaking is that the shaky frames usually have a lot of motion blur. If you leave no shaking the picture seems to blur for no reason at all.

[Tafflad]

Thanks that was what I was after - a guide.

I have hit an anomaly with some testing
Was going to try some advice seen on this forum instead of Adaptive zoom average + fixed zoom, was to try:
Edge compensation: Adaptive zoom average, some borders

use previous & future frames to fill in borders

leaving in the default of - 'soft edges' ...

This requires you to crop out any borders before hand ... to avoid annoying black lines ...
Hit a problem using crop with VD .... input files is GV lossless .....

I found that if I use null tranform filter ... crop to size I want, then save using fast recompress > GV Lossless .. the saved file does not have the crop applied - even though output window of VD shows it has.

If instead I omit fast recompress and use compression>GV Lossless then the crop is correct on save file

Why is that ?

[pippas]

If instead I omit fast recompress and use compression>GV Lossless then the crop is correct on save file

Why is that ?

You can only use 'fast recompress' with Vdub if you do not apply any filters.
You applied a filter.... and therefore need to re-convert.

[Tafflad]

OK .. I had misunderstood its use.

[jagabo]

The four Video processing modes of VirtualDub:

Direct Stream Copy: Copy compressed data directly from input file to output file

Fast Recompress: Decompress input video, no colorspace change, no filtering, recompress (or not) to output

Normal Recompress: Decompress input video, colorspace change to RGB, no filtering, recompress (or not) to output

Full Processing: Decompress input video, colorspace change to RGB, filtering, recompress (or not) to output

Some of this has changed a bit in recent years with the ability to request a particular colorspace from the decomrpession codec, support for more colorspaces, some filters that work in YUV now, smart rendering, etc.

[Tafflad]

Decided to do some tests .... and getting a strange result ... probably a simple explanation, but not what I expected.

Created a GV Lossless 2 min 'sample A' .......... size 3.02GB
Load 'sample A' into VD, no changes or additions and save as Full processing Compress - GV lossless .... output excatly matches input & same file fize 3.02GB

Load 'sample A' again,no changes or additions, this time fast recompress - GV Lossless ... file size down by 50% to 1.4GB


Looking at Mediainfo I can see that in both original and full compression saved ... bitrate is 429Mbps
In the fast recompress the bitrate is down to 199Mbps

Why is it dropping bitrate by more than half, when I am changing no settings ?

[pippas]

3 things spring to mind...

1)If you're not making any changes why not simply save as a 'Direct stream copy'?... that will be exactly the same size!

2) Even though codecs like Grass Valley Lossless are just that .. lossless (i.e. you can recover all the original quality) .. they still apply compression (IIRC Grass Valley Lossless is about 2:1?).
If you further compress that already compressed file, then you will both reduce the size - and the quality... You can only create one generation of a compressed lossless encode as 'lossless'

3) Why not take a small sample file from what you are working on, and encode it both with Grass Valley Lossless and also Grass Valley HQX. I'll bet you can't see any difference in visual quality. But the HQX file will be much smaller.

HQX files encoded with the 'standard' setting (rather than the default 'fine' setting) occupy about half the space of an equivalent DV file.
And for clips derived from standard definition video will be more than adequate.

But don't take my word for it. Encode a short clip for yourself - see if you can see any difference between lossless and HQX...

[Tafflad]

I'm only using GV lossless as intermediate format for future processing in VD which was what was recommended........... output file will probably be X264vfw


The test shows that is I apply full processing mode - compression - GV Losless again ... there is no reduction in file size (what I expected), but if I use fast recompress then the files size is reduced by 50% and bitrate dropped by 50%

The 2nd part ties in with your 2:1 description - but not the first.



On you other comment .. I hadn't thought about using direct stream copy ... don't know why .,.. used it previously .... looking at media info the net result is the same ... but I will use it in future.

[pippas]

Looking at jagabo's list above in post#9, you'll see the differences between the fast re-compress and full re-compress.
It appears that the first step is to decompress the video in both cases, so my comment about further compressing a compressed file probably doesn't apply to lossless. Sorry about that.

The subsequent steps are different for the two procedures, and I'm guessing that is where the differences in encoding is occurring, resulting in the different file size.

If it is still lossless than it doesn't make any sense, as you say. Looks like fast re-compress may not be lossless then?...

Best for a Vdub expert to confirm that..

I'm only using GV lossless as intermediate format for future processing in VD which was what was recommended........... output file will probably be X264vfw

As I say, I personally think that HQX would be more than adequate for what you're doing -- still very resilient to multi-generational re-encodes, and much smaller working files than lossless. x264 is an excellent choice for your outputs, and easy to place in an mp4 container for better compatibility....

You can read more about HQX here: http://www.grassvalley.com/docs/WhitePapers/professional/GV-4097M_HQX_Whitepaper.pdf

It's a very versatile codec!

[jagabo]

Decided to do some tests .... and getting a strange result ... probably a simple explanation, but not what I expected.

Created a GV Lossless 2 min 'sample A' .......... size 3.02GB
Load 'sample A' into VD, no changes or additions and save as Full processing Compress - GV lossless .... output excatly matches input & same file fize 3.02GB

Load 'sample A' again,no changes or additions, this time fast recompress - GV Lossless ... file size down by 50% to 1.4GB


Looking at Mediainfo I can see that in both original and full compression saved ... bitrate is 429Mbps
In the fast recompress the bitrate is down to 199Mbps

Why is it dropping bitrate by more than half, when I am changing no settings ?

The difference is about what you would expect between RGB and YV12 colorspaces. Or between YV24 and YV12. Does MediaInfo say anything about colorspace? I've never used GV codecs so I'm not familiar with their particulars. But look at the configuration dialogs for colorspace settings.

[Tafflad]

There is no colourspace info on the Media Info files that I can see ..... 3 files attached
The only difference between the files is on fast recompress bitrate reduces to 199Mbps, while on the original test and full compress test version bitrate is 429Mbs .. and on compress there is an extra line of information .... it gives a "Format Profile : Open DML"
No such line on the fast compress file.

There are no settings on the fast recompress codec .. in the VD window no options or settings to change.
The 3 mediainfo files are attached.

I have my way fwd - don't use fast recompress ...... just wondering why this does not work.

[jagabo]

I was afraid MediaInfo wouldn't say anything useful. Try Opening each of the videos with an AviSynth script like:

AviSoure("filename.avi")
Info()

That will overlay information about the video onto the picture. What does it say about the colorspace?

[Tafflad]

OK .... just to run the sequence
Original file DV capture - YV12
Deinterlaced and colour space change to YUY2, compressed with GV-Lossless (Bitrate shows as 429Mbps) - saved as test
open test and save again as GV Lossless or direct stream copy - YUY2 (Bitrate shows as 429Mbps )
open test and save using fast recompress - YUY2 (Bitrate shows as 199Mbps)


So not showing any colour space change only bitrate

[jagabo]

How are you verifying the colorspace used in each file?

[Tafflad]

with the script you posted in #16 .....
can post screen grabs if you want ... but other than colorspace all other info was in Mediainfo files

[poisondeathray]

I just did some quick tests and Grass Valley "lossless" appears to upscale to 4:2:2. It looks like 4:2:0 isn't supported. A larger problem for interlaced material is it probably treats the chroma plane upscaling as progressive. So it's not "lossless" for you with a PAL DV 4:2:0 source

I tested both official GV decoder, and the ffmpeg variant

[Tafflad]

So all the good suggestions here to use Grass valley are no good for me on my DV captures ?
Is this issue on both compress & fast compress ?

Checking as I have spent a lot of Hrs de-interlacing a load of captures .... do I need to start again and save as something else ............
Lagarith ?

Want to use interframe lossless format as will be carrying out extra steps, e.g. stabilizing.

[jagabo]

If you're deinterlacing first you don't need to worry about interlaced 4:2:0 to 4:2:2 upsampling by GV Lossless -- after deinterlacing the 4:2:0 video is progressive. I don't know what the deal is with GV Lossless and your file sizes. I don't really want to install the codec to test it out myself.

[pippas]

Want to use interframe lossless format as will be carrying out extra steps, e.g. stabilizing.

I think you probably mean 'intraframe' rather than 'interframe' ?

As I suggested before, why not try HQX instead of lossless?... It's a high quality intraframe intermediate, more than capable of coping with processing your converted DV files without any visual quality loss (that's what intermediates are for!).

[Tafflad]

If you're deinterlacing first you don't need to worry about interlaced 4:2:0 to 4:2:2 upsampling by GV Lossless -- after deinterlacing the 4:2:0 video is progressive. I don't know what the deal is with GV Lossless and your file sizes. I don't really want to install the codec to test it out myself.

All of my DV captures will be deinterlaced with QTGMC, and as per advice on Doom9 correcting for Youtube/Vimeo, resizing, and setting levels to [16,235] Then changing color space to YUV "PC->TV"

So I should be OK with GV Lossless (good means 10's of Hrs work not wated)

I'll just ignore fast recompress.

Thanks for your input

[poisondeathray]

Sorry my fault, I wasn't clear on your workflow

Yes, if you are deinterlacing first, those GV issues with interlaced material not a problem (If you keep it interlaced - you will see chroma "notching" artifacts and "ghosting" from treating the chroma as progressive. Those are "serious" problems)

But if that progressive GV intermediate is for use in vdub for deshaker; deshaker works in RGB. Technically it's still worse because you go from YV12 => YUY2 => RGB , instead of YV12=>RGB. But that is not a "serious" problem and you probably can't even tell, especially on your type of source material . I guess you can convert to RGB before and use an RGB intermediate, but filesizes will be larger. For any given codec, YV12 will be the smallest in filesize compared to YUY2 or RGB, because the chroma planes are the smallest

[Tafflad]

Want to use interframe lossless format as will be carrying out extra steps, e.g. stabilizing.

I think you probably mean 'intraframe' rather than 'interframe' ?

As I suggested before, why not try HQX instead of lossless?... It's a high quality intraframe intermediate, more than capable of coping with processing your converted DV files without any visual quality loss (that's what intermediates are for!).

Yes I did mean intraframe ... you are right.

I have stayed away from HQX as had been having help from several people and they advise I should save as an intraframe lossless intermediate, when I have further processing steps to undertake. (never used GV at all so no personal experience, previously used Lagarith for this))

I had been intending after processing complete to encode as X264vfw .......... but found out I need this in an MP4 container not AVI.

Ultimately taking a load of these MP4 files into an NLE (Vegas) ........... when I discussed X264vfw format there - was quickly told by several people that it is not a good step.

".... there is no reason to deliver MPEG-4 video in anything but an MP4 container. The fact that x264vfw uses the out-dated Video for Windows format in an AVI container is big problem for Vegas. Use either the Sony AVC or MainConcept AVC MPEG-4 encoders. It's OK to render to X264vfw if you have a compelling reason to deliver H.264 as an AVI file but do not try and edit any of it on the Vegas Pro timeline because it will just cause problems."

[jagabo]

All of my DV captures will be deinterlaced with QTGMC, and as per advice on Doom9 correcting for Youtube/Vimeo, resizing, and setting levels to [16,235] Then changing color space to YUV "PC->TV"

So I should be OK with GV Lossless

I guess. I don't use the codec so I don't know exactly what's going on with the different file sizes. Since AviSource() is giving you YUY2 for all sources try comparing them with:

v1 = AviSource("video1.avi")
v2 = AviSource("video2.avi")
Subtract(v1,v2)
Levels(120,1,136,0,255)

If the two videos are identical you will see a flat grey image. If they are different you will see the differences as variations from grey.

Your DV source is YV12 (4:2:0), and the output of your AviSynth script is probably YV12. Upscaling to YUY2 (4:2:2) may blur colors a tiny bit but it's not worth worrying about with Hi8 or VHS sources (which have very little color resolution).

But your plan to adjust levels to 16-235 and follow up with PC->TV is wrong. 16-235 is already TV levels.

[pippas]

I have stayed away from HQX as had been having help from several people and they advise I should save as an intraframe lossless intermediate, when I have further processing steps to undertake. (never used GV at all so no personal experience, previously used Lagarith for this))

.......I had been intending after processing complete to encode as X264vfw .......... but found out I need this in an MP4 container not AVI.

Ok with the HQX... IMHO, by using Lossless rather than HQX you are simply creating files that are much larger than they need to be. HQX is an intraframe intermediate as well....it's just not actually 'lossless'... but as you've already decided and created lossless files, there's no point in going back.

(For your next project though, do try a few clips with HQX.... you really won't see any difference from lossless... even if you make half a dozen or more processing procedures!... )

Vdub exports x.264 as a .avi file. The simplest way of creating an mp4 without recoding the video is to use the free program MkvToMp4. (It will re-encode the audio, but not the video).

The preferred audio format for Mp4 is AAC. You can't export from Vdub with an AAC audio track, but you can export audio encoded as AC3..
MkvToMp4 will recognise that format, and convert the audio automatically to a compliant AAC audio file.
- I've not been able to get MkvToMp4 to accept linear pcm audio tracks (which is the default format you're working with).

There are of course other methods of creating mp4s, but simply 'dragging and dropping' your x264 .avi from Vdub into MkvToMp4 is the simplest way I've found.

You've no need to install Mkv2Mp4 (it's portable) and the only change I make from the 'default' settings is to change the output from m4v to mp4 in the video tab in the 'settings'...

[Tafflad]

All of my DV captures will be deinterlaced with QTGMC, and as per advice on Doom9 correcting for Youtube/Vimeo, resizing, and setting levels to [16,235] Then changing color space to YUV "PC->TV"

So I should be OK with GV Lossless

I guess. I don't use the codec so I don't know exactly what's going on with the different file sizes. Since AviSource() is giving you YUY2 for all sources try comparing them with:

v1 = AviSource("video1.avi")
v2 = AviSource("video2.avi")
Subtract(v1,v2)
Levels(120,1,136,0,255)

If the two videos are identical you will see a flat grey image. If they are different you will see the differences as variations from grey.

Your DV source is YV12 (4:2:0), and the output of your AviSynth script is probably YV12. Upscaling to YUY2 (4:2:2) may blur colors a tiny bit but it's not worth worrying about with Hi8 or VHS sources (which have very little color resolution).

But your plan to adjust levels to 16-235 and follow up with PC->TV is wrong. 16-235 is already TV levels.

No, what I meant is that I'm using PC->TV to set the 16-235 levels to avoid detail loss on Youtube.

I'll try the compares out

[Tafflad]

(For your next project though, do try a few clips with HQX.... you really won't see any difference from lossless... even if you make half a dozen or more processing procedures!... )

Vdub exports x.264 as a .avi file. The simplest way of creating an mp4 without recoding the video is to use the free program MkvToMp4. (It will re-encode the ...

Thanks for lead to Mkv2Mp4

I will try out HQX next time ...... was just so surprised in massive file difference assumed I would be losing a lot. (13x smaller)
Quality wise is HQX better quality than X264 ? again files sizes are much smaller.

I took a short clip ... original was from a DV capture, deinterlaced, resized, levelled and encoded in GV Lossless, then as HQX and then as X264
To reflect end goal they are combined and encoded as SonyAVC/MP4

Be interested in your comments on the relative quality of the 3 files.