Best Rendering Setting for 24p (deinterlaced)??

[Jeff Baker]

The only place I see any flicker is on hand held pans after they are de-interlaced (this includes slow pans) that seems to be a trouble spot for deinterlacers unless you can play around with the motion blur parameters (like with Twixtor for AE).

There is a Reduce Interlace Flicker Switch in Vegas, but I don't know if it would apply to progressive timelines even if the original was interlaced.

[David Jimerson]

Jeff, are you converting to 24p when you're seeing these pans?

It's probably not because of the deinterlacing -- it's because of the frame rate conversion. When you shoot 24p, you have to pan sloooooowly or you'll end up with a motion stutter. A quicker pan in 60i may look fine, but when you convert to 24p, you'll end up with the same stutter you would have had if you had shot 24p to start with.

[Jeff Baker]

Right. Only once the footage gets to 24p do pans start get touchy. What I don't get is that pans that are faster and shot on film still seem to look better than what I can get with deinterlacing. The jerky effect is minimized if I use the FieldsKit pluging for AE to convert to 60p (2 frames per field) and then convert to 24p with some motion blur.

[Dale Paterson]

Hi David,

1) No. Never seen that happen. When I render uncompressed and progressive, the file is always progressive.

When you say 'render uncompressed' do you mean that you are choosing the 'Default template (uncompressed)' and then changing the field order to progressive there or are you choosing the 'PAL DV' template and then changing the field order there? If I use the 'PAL DV' template and change the field order to progressive Vegas does recognize the resulting files field order as being progressive BUT if I choose the 'Default template (uncompressed)' and change the field order to progressive in this template - no matter what I do - no matter what my project settings are - or anything else - the resulting file is ALWAYS interlaced, upper field first - at least according to Vegas.

2) I haven't seen that flicker, either. Where do you see it? Playing it on/in what?

Create a new project with the correct settings for a HDV file (or match the new project to a HDV file - it can be either .m2t or a Cineform .AVI file). Load the HDV file into the project and then render it to PAL DV first with interlace method set to none and then render again with interlace method set to blend or interpolate (it makes no difference for this excercise). You will notice that the PAL DV file that you have created from the HDV file with de-interlace method set to none is unwatchable but the file created when the de-interlace method was set to either blend or interpolate is fine. This also happens when you try and convert PAL DV Wide to PAL DV.

I have my own theory on this (which nobody has ever passed an opinion on) and the theory is this:

Vegas de-interlaces a file - using the de-interlace method selected - whenever it feels it needs to and then 're-interlaces' the file to produce the desired output. This may sound strange but think about it - HDV is interlaced, upper field first and PAL DV is interlaced, bottom field first so I can understand that by converting HDV to PAL DV you are changing the field order which is a no-no and I can understand why you would get interlace flicker (particularly when there is movement) BUT why then does the setting of the de-interlace method even come into play in this scenario. In other words (if my theory is wrong) I should get a PAL DV file that is interlaced, bottom field first, with much interlace flicker (because I am effectively changing the field order) if I rendered with the de-interlace method set to none (which is in fact the result) but I should get the SAME result even if de-interlace method is set to blend or interpolate BECAUSE at no point have I instructed Vegas to de-interlace my footage! Any thoughts on this?

Either the above or the Vegas de-interlace method is ALSO used when you for some or the other reason are swapping fields / changing the field order for whatever reason (and again the only explanation for this would be if Vegas first de-interlaces the original footage and then re-interlaces depending on whether or not you are trying to create an interlaced file with a different field order).

Just another bit of useless (useful) information: If you allow the Sony FX1E to convert HDV to PAL DV the output file size is almost identical (as a matter of fact it is slightly larger) to the same .m2t file captured from the FX1E and converted using Vegas to PAL DV with the de-interlace method set to blend fields. Interpolate fields creates a much smaller file. From this I can only assume that you are getting the same or possibly even better quality HDV to DV conversion when the camera does the conversion and of course take only a fraction of the time it takes to convert a HDV tape to DV using Vegas.

Regards,

Dale.

[David Jimerson]

Originally Posted by Jeff Baker

Right. Only once the footage gets to 24p do pans start get touchy. What I don't get is that pans that are faster and shot on film still seem to look better than what I can get with deinterlacing. The jerky effect is minimized if I use the FieldsKit pluging for AE to convert to 60p (2 frames per field) and then convert to 24p with some motion blur.

In Vegas, which deinterlace method are you choosing?

Are you disabling resample?

A trick to getting faster pans to look more smooth when shooting 24 fps is to follow an object -- like a walking person -- keeping the object constant in the frame. The background still strobes, but you don't notice because you're following the object.

[David Jimerson]

Originally Posted by Dale Paterson

When you say 'render uncompressed' do you mean that you are choosing the 'Default template (uncompressed)' and then changing the field order to progressive there or are you choosing the 'PAL DV' template and then changing the field order there? If I use the 'PAL DV' template and change the field order to progressive Vegas does recognize the resulting files field order as being progressive BUT if I choose the 'Default template (uncompressed)' and change the field order to progressive in this template - no matter what I do - no matter what my project settings are - or anything else - the resulting file is ALWAYS interlaced, upper field first - at least according to Vegas.

I mean "default template (uncompressed)." I do a lot of compositing, and I need to render with alpha channels, and you can only do that with uncompressed. I've never had a problem making a (recognized) progressive file from it -- but I've never done it in PAL.

Create a new project with the correct settings for a HDV file (or match the new project to a HDV file - it can be either .m2t or a Cineform .AVI file). Load the HDV file into the project and then render it to PAL DV first with interlace method set to none and then render again with interlace method set to blend or interpolate (it makes no difference for this excercise). You will notice that the PAL DV file that you have created from the HDV file with de-interlace method set to none is unwatchable but the file created when the de-interlace method was set to either blend or interpolate is fine. This also happens when you try and convert PAL DV Wide to PAL DV.

I have my own theory on this (which nobody has ever passed an opinion on) and the theory is this:

Vegas de-interlaces a file - using the de-interlace method selected - whenever it feels it needs to and then 're-interlaces' the file to produce the desired output. This may sound strange but think about it - HDV is interlaced, upper field first and PAL DV is interlaced, bottom field first so I can understand that by converting HDV to PAL DV you are changing the field order which is a no-no and I can understand why you would get interlace flicker (particularly when there is movement) BUT why then does the setting of the de-interlace method even come into play in this scenario. In other words (if my theory is wrong) I should get a PAL DV file that is interlaced, bottom field first, with much interlace flicker (because I am effectively changing the field order) if I rendered with the de-interlace method set to none (which is in fact the result) but I should get the SAME result even if de-interlace method is set to blend or interpolate BECAUSE at no point have I instructed Vegas to de-interlace my footage! Any thoughts on this?

If you're working in an interlaced project and wanting to output an interlaced file, I'm not sure why deinterlacing will ever come into play. Maybe I'm not following you here.

Either the above or the Vegas de-interlace method is ALSO used when you for some or the other reason are swapping fields / changing the field order for whatever reason (and again the only explanation for this would be if Vegas first de-interlaces the original footage and then re-interlaces depending on whether or not you are trying to create an interlaced file with a different field order).

I don't see why this would happen.

Just another bit of useless (useful) information: If you allow the Sony FX1E to convert HDV to PAL DV the output file size is almost identical (as a matter of fact it is slightly larger) to the same .m2t file captured from the FX1E and converted using Vegas to PAL DV with the de-interlace method set to blend fields. Interpolate fields creates a much smaller file. From this I can only assume that you are getting the same or possibly even better quality HDV to DV conversion when the camera does the conversion and of course take only a fraction of the time it takes to convert a HDV tape to DV using Vegas.

I wouldn't say the difference in file size means you're necessarily getting a better-quality file; it could have to do with compression efficiencies, etc.. But I've never worked with the FX1, so I don't have any first-hand experience with this.

[Jeff Baker]

Originally Posted by David Jimerson

In Vegas, which deinterlace method are you choosing?

Are you disabling resample?

I use smart resample. There seems to be some connfusion about resample settings elsewhere in the forum so I stick to the defualt for Vegas 6. Can you apply the resample switch on a clip by clip basis (in other words, turn it off of panning clips).

Keeping the center of attention anchored in a pan is a cool tip, thanks. I do a lot of landscape work which is where I often run into trouble. That and handheld shots.

Is there away to affect motion blur outside of the project settings in Vegas?

Oh, this probably goes without saying for deinterlace projects, but I alway shoot with a shutter speed of 60 (ntsc footage).

[David Jimerson]

Smart resample is fine, but be sure that your project deinterlace method is "interpolate."

Also, which pulldown scheme are you adding? If you're adding 2-3-3-2 pulldown and try to watch it, it's going to have a more pronounced stutter than 2-3 pulldown. Keep an eye on that.

[Jeff Baker]

I render with the DVD Mpeg2 template for 24p that inserts the 2-3 pulldown.

I had the best luck when I deinterlaced with fields kit to 60p and then brought the resulting footage into a Vegas 24p project for final rendering with smart resample set to blend and motion blur set to guassian.

Have not tried the interpolate yet, but I will on my next pass which will be use Magic Bullet for the deinterlace.

[David Jimerson]

For best results, what you really ought to do, Jeff, is convert your footage to 24p before you edit (using 2-3-3-2 pulldown), then edit the converted files on a 24p timeline.

[Jeff Baker]

yeah next time I will. But this is a feature length project that I have been editing for the past 2 years. Anyway, I showed a 24p version from DVD at the San Francisco Film Arts Foundation on Saturday and I was very pleased with the results. Some people thought it looked like 16mm which I took as a completment. I hope they were not expecting 35 from dv...

Thanks for everyones help here, the look really came out nice.

[Dave Haynie]

Originally Posted by Dale Paterson

Hi David,
I have my own theory on this (which nobody has ever passed an opinion on) and the theory is this:

Vegas de-interlaces a file - using the de-interlace method selected - whenever it feels it needs to and then 're-interlaces' the file to produce the desired output. This may sound strange but think about it - HDV is interlaced, upper field first and PAL DV is interlaced, bottom field first so I can understand that by converting HDV to PAL DV you are changing the field order which is a no-no and I can understand why you would get interlace flicker (particularly when there is movement)

I think it might be even more fundamental than that. In the computer, particularly in an application like Vegas, things tend to work out well when being frame oriented, not field oriented. So it's very likely that Vegas would treat a 60i video track as, essentially, 30p plus an "interlace" bit, for all practical purposes. That "progressive" representation is the natural field order for that media type, eg, the upper and lower field that belong in a frame together, based on the chronology. This isn't real de-interlacing, in any sense of blending.

Originally Posted by Dale Paterson

BUT why then does the setting of the de-interlace method even come into play in this scenario. In other words (if my theory is wrong) I should get a PAL DV file that is interlaced, bottom field first, with much interlace flicker (because I am effectively changing the field order) if I rendered with the de-interlace method set to none (which is in fact the result) but I should get the SAME result even if de-interlace method is set to blend or interpolate BECAUSE at no point have I instructed Vegas to de-interlace my footage! Any thoughts on this?

I think basically, it amounts to a choice here... if you were to render "straight" to, say, DVD-class MPEG-2/60i from your HDV, same field order, the video would look fine rendered directly from that internal model. If you rendered to DVD-class MPEG-2/30p, same thing. But if you render to a form, like DV, where the interlacing is reversed, then the temporal ordering of your frames is wrong, given than internal model, and must be changed. I think the "de-interlace method" set to "NONE" is essentially saying "never change from this internal model" in any intelligent way.

This isn't just Vegas, either.. other programs I've used seem to behave in a way that fits the model I'm suggesting.

Originally Posted by Dale Paterson

Either the above or the Vegas de-interlace method is ALSO used when you for some or the other reason are swapping fields / changing the field order for whatever reason (and again the only explanation for this would be if Vegas first de-interlaces the original footage and then re-interlaces depending on whether or not you are trying to create an interlaced file with a different field order).

I think it really amounts to "do something" vs. "do nothing", it's just that interlacing is the biggie, when it comes to bad-looking video. If you render from HDV to MPEG-2 60i MP@ML resolution, without any de-interlacing method, you just get the resizing. If you have de-interlacing checked, it should be essentially the same thing (well, assuming it's smart enough to do field-based resizing; if not, the former could be messed up). If you render to 30p, the video still looks reasonable, though with de-interlacing checked, you may well get "smart" field blending, rather than something a bit more brute-force... but it's way more nunanced. When changing the field order, it's an seriously obvious case when this isn't handled properly (back in the days of SuperVideoCD, I made this mistake occasionally).

Originally Posted by Dale Paterson

Just another bit of useless (useful) information: If you allow the Sony FX1E to convert HDV to PAL DV the output file size is almost identical (as a matter of fact it is slightly larger) to the same .m2t file captured from the FX1E and converted using Vegas to PAL DV with the de-interlace method set to blend fields.

That makes perfect sense: DV isn't a variable-rate format. I've never actually tried on-camera DV conversion (I have the HVR-A1), simply because I would naturally assume that Vegas would do it better -- it's got many, many times the resources, even on a modest PC... and there's no demand to do it in realtime, though I would hope in my case (Athlon64x2 4200+), it's happening much faster.
[/quote]

Originally Posted by Dale Paterson

Interpolate fields creates a much smaller file.

When rendered to DV? That's really weird; it's not supposed to work that way.

Originally Posted by Dale Paterson

From this I can only assume that you are getting the same or possibly even better quality HDV to DV conversion when the camera does the conversion and of course take only a fraction of the time it takes to convert a HDV tape to DV using Vegas.

I think you can bank on Vegas, or other PC-based tools, doing at least as good a job on the downconversion. They should do better, if there's any real margin in the conversion process.

-Dave