Cropping...

[manono]

Add:

Info()

to the bottom of the script (without any colorspace conversion) and it'll give you the colorspace and much more information. If it isn't YV12 to begin with, and it'll become YV12 afterwards, it's better to let AviSynth do the conversion.

[poisondeathray]

I think most DV is YV12. However whatever you are doing in Premiere seems to be converting it (probably to RGB). You can check with avisynth on any video, you would use your source filter and then info()

e.g.
AVISource("DV.avi")
Info()

I mentioned earlier it is possible (at least with CS4) to keep YV12 all the way through the workflow with the one I described earlier (even with multiple filters, effects) to minimize losses. Not sure what is happening in your case

Some programs do care, and some filters require a certain colorspace.

[jagabo]

What color space is DV

Internally, NTSC DV uses 4:1:1 YUV encoding. You have a 720x480 luma channel and two 180x480 chroma channels. The uncompressed video that comes out of the DV decoder will not be in that format. It will be RGB (4:4:4), YUY2 (4:2:2), or YV12 (4:2:0).

RGB will either be 24 bit with 8 bits each of R, G and B. Or 32 bit with 8 bits each of R, G, and B and 8 bits unused.

YUY2 will have a 720x480 luma channel and two 360x480 chroma channels.

YV12 will have a 720x480 luma channel and two 360x240 chroma channels.

[Cornucopia]

Most good video editors (incl. Premiere) treat DV slightly differently than other codecs.

1. They fully support non-square pixels, and use it as default.
2. They fully support DV's 4:1:1 (or 4:2:0) sampling (IOW, YV12)
3. They use SMART RENDERING by default.

If you DON'T use a DV edit session preset, it'll often default to square pixels, and RGB. That's probably what Premiere is doing.

Therefore, I would STRONGLY suggest that you do your edit in native DV, without any cropping, deinterlace, etc. Then once your master clip is created (either by rendering or frameserving), you do the deint, crop, smooth, resize, etc. in AVISynth.

Less generations, less manipulations, less LOSS, probably less time used overall.

Scott

[bsuska]

Thanks Scott...

I do usually do my editing with a Matrox DV preset, but both the Huff and Lag codecs did seems to default to RGB in their codec settings.

This seems like a good workflow...

Anyone know of any code for AVISynth that will tell me if a clip is interlaced or not?

[jagabo]

I do usually do my editing with a Matrox DV preset, but both the Huff and Lag codecs did seems to default to RGB in their codec settings.

It's Premiere that's defaulting to RGB output or requesting RGB input (unless you have the "force RGB" output options checked in the codecs).

Anyone know of any code for AVISynth that will tell me if a clip is interlaced or not?

Look at the video in a program that doesn't deinterlace (ie, not a media player). If you see comb artifacts it's interlaced. If you don't see comb artifacts its not interlaced. This is the only reliable method.

[poisondeathray]

^Just to add to jagabo's comments, an example of non-deinterlacing player would be vdub. So if you opened your file directly in vdub, and step through it should give you an indication of you see combing

since you have 2 project goals from 1 workflow, you are going to have to make some trade offs. You never answered my questions on "what type of editing" you were planning on doing. Here is why it's important, and why I STRONGLY suggest you deinterlace before importing if your goal was progressive output:

Edges get ratty and artifacts are worse if you deinterlace after. Even simple still titles get artifacts. It's much worse when you begin to add more effects, overlays , animations, motion graphics, or more advanced stuff from after effects. Interlaced footage is just plain bad if you plan to do any of the above. Not even by using tempgaussmc_beta1 can you escape the ratty edges

The following examples are just stills taken from lossless huffyuv exports (ignore the aspect ratio (1.5) as I took screenshots in AvsP so it assumes 1:1 pixels, all were ConvertToYUY2(interlaced=true) in avisynth if exported as interlaced, and ConvertToYUY2() if exported as progressive (it didn't matter in this case if ConvertToYV12 was used instead). It's not shown in the stills, but the ratty edges "vibrate" and look even worse in motion (I animated the title).

1) Yadif before importing


2) Yadif after exporting


3) TempGaussMC_beta1 before importing


4) TempGaussMC_beta1 after exporting


For kicks, the following is a video demonstrating different bob deinterlacers. Remenant artifacts ("jaggies") combine on subsequent frames to produce shimmering. Remenant jaggies from bad deinterlacing also "eat up" more bitrate, which is especially important in low bitrate streaming situations like yours. You really want to get the cleanest, smoothest image to optimize bitrate and bandwitdh usage, so jagabo's recommendation of tempgaussmc_beta1() is a very good one in terms of quality (just takes a lot of CPU power)

https://forum.videohelp.com/images/guides/p1934885/stockholma_0-520_q3_yadif_mvbobmod_t...mca4_tdtmm.avi

[jagabo]

^Just to add to jagabo's comments, an example of non-deinterlacing player would be vdub. So if you opened your file directly in vdub, and step through it should give you an indication of you see combing

Be careful with VirtualDub and the MPEG import filter. It performs 3:2 pulldown on 23.976 fps progressive sources with pulldown flags. This leaves you with 29.97 fps interlaced frames. VirtualDubMod on the other hand does not perform the pulldown and will retain the original 23.976 fps progressive frames.

[poisondeathray]

Be careful with VirtualDub and the MPEG import filter. It performs 3:2 pulldown on 23.976 fps progressive sources with pulldown flags. This leaves you with 29.97 fps interlaced frames. VirtualDubMod on the other hand does not perform the pulldown and will retain the original 23.976 fps progressive frames.

Good point jagabo. I have to admit I don't use vdub for looking at fields, I usually use avsp for everything and this method (http://neuron2.net/faq.html#analysis); I just thought vdub would be easier for bsuska, instead of using a script. I think he should be ok using DV source and vdub

[jagabo]

I just thought vdub would be easier for bsuska, instead of using a script. I think he should be ok using DV source and vdub

Yes, the soft pulldown problem doesn't exist in DV.

[bsuska]

Is there away to run a bunch of files thru Virtual Dub at once (batch deinterlace a bunch of files)?