I read here that Vegas 6's 24p conversion from 60i looks really good. Does it look close to the 24p cams (DVX-100/XL-2). Or does DVFilmaker do a better job?

James, let's get technical. Does anyone know what method Vegas uses to convert to 24p? This thread (http://www.dvinfo.net/conf/archive/index.php/t-8980.html) would indicate that it's throwing away 60% of the fields and interpolating based on the remaining fields. That sounds extreme, I'm not sure I can believe that.

Another thread said it does a basic inverse telecine (IVTC). That means it throws out 20% of the fields and possibly deinterlaces the remaining fields, either through interpolation or field blending (sounds like it's configurable).

Does anyone have a better reference for how it works?

IVTC will look fine for NTSC playback, but it's not for film out.

Josh

Whatever it's doing, it does do damn good job of it.

I have both DVFilm and Vegas 6. Up until Vegas 6 I would have said DVFilm did a better job, but Vegas 6 does what to my eyes looks like a perfect 60i to 24p. They've changed the frame rate conversion algorythm. I've posted questions on the Sony site several times asking how it works, but the question has been ignored each time. I suspect that they're treating 60i like 60p then dropping frames from this in something based on a 3:2:2:3 pattern but I'm not sure. Whatever it is, it really looks good.

By the way, PAL conversion is equally good too now. So does uprezzing interlaced video.

*removed* doh!

Thanks for posting the clip. If you look at the upright poles about 60% through the clip, and advance frame by frame, you will see that there is a double image, a double image, then a triple image, a triple image, etc. So, it's definitely doing some frame blending, I would say. Slight double images can occur in DVFilm Maker, but I've never seen a triple image. Interesting. If it was interpolating, there would be a single image. MB would give you a single image. It would also probably strobe badly on this clip. The blending tends to smooth the strobing out.

It's an interesting pattern, though, I can't figure it out exactly. There should be a test clip for 60i to 24p conversion that will help you figure out what's going on.

Josh

I downloaded the clip and stepping through it frame by frame, I can see what you're talking about. Then I went through one of my own conversions frame by frame and it was perfect.

I don't know why our results are different.

Lawrence, I believe there are at least two modes, from the previous threads I read: interpolation and blending. He's probably using blending, you're probably using interpolation. There are circumstances where one is better than another.

There were two modes (interpolation or blending) which affected 24p renders on Vegas 5 and before. In Vegas 6, neither one is used on a 60i to 24p render. In looking at his file again, It looks like he might have rendered to 24p instead of 23.97 which meant that it couldn't use a 3:2:2:3 pulldown perfectly.

*edit*

checking something...

Ok, couldn't be sure if I originally exported that last clip under Vegas 5 or 6 (just got my upgrade last week and that ghetto footage is from last week and I don't remember).


Anyway, here's a new clip, just captured it off tape now, just exported it now (straight from Vegas to Quicktime). 23.976, Interpolate fields, etc etc.

Click (http://www.twodogfilms.com/Patrick/24p.mov)

Hope that's a better indication of 24p under Vegas. I still see a 2-3 ish type of cadence when looking at the light poles, but all in all, it's pretty good to my eyes. No stuttering. Nothing looks wrong, etc.

Patrick, yeh, definitely blending. It won't produce accurate and crisp stills, but it'll look pretty smooth on NTSC playback. Due to the multiple (and changing effects on some objects, it's probably not suitable for film out, but who of us is actually ever going to do a film out?

Would you be game to grab the demo version of DVFilm Maker, and convert the same clip? I'd be interested to see a side by side.

Vegas 6 (http://www.twodogfilms.com/Patrick/24p-vegas6.mov)

DVFilm Maker (http://www.twodogfilms.com/Patrick/24p-dvfilm.mov)

For DVFilm, I used the "24P Editing Options -> Convert 60i to 24P" option. Reimported back into Vegas and saved back to Quicktime.

I dunno, to me, I actually think the Vegas 6 footage looks better, or to my eyes, what I would expect to see. The blurring between frames looks a lot better and more even over time, like actual motion blur (sort of) between frames - DVFilm is very staccato, then blurred, then staccato, etc - feels a lot more mathmatical than naturalalistic.

$.02

Patrick, awesome, thanks for posting those!

Yeh, with DVFilm Maker, it samples every 2.5 fields. So odd number frames are based on those half fields, where it sample backwards and forwards, and there is a slight double image. You get single/double/single/double.

Vegas you're getting double/triple/double/triple. Vegas must be blending, not just two fields, but three in some cases. The extra blending actually gives surfaces a very smooth appearance. Grain gets averaged out and its very smooth.

I'm a little bothered by the double/triple cadence. The lampposts, in particular, look very strange, because when they triple, they looks very fat.

I guess it goes without saying that the same scene, shot on film would have a single blurred lamppost in each frame.

Next step, Magic Bullet. You'll find with that each frame is interpolated and there won't be double images either. A single lamppost. It's actually so sharp in my tests that it gets very stroby.

Wierd, I used to see what you're talking about in Vegas 5 and below. In Vegas 6 I see no double images at all. I wonder what we're doing differently.

anyone want to post the specific instructions on how to do this in vegas? thanks

You just open a regular NTSC project in Vegas 6 and render to NTSC P24 3:2:2:3 pulldown avi or NTSC 24P mpeg 2. It couldn't be easier. If you're going to AVI, selecting the 3:2:2:3 pulldown method is important. You'll see double images on certain frames if you don't choose this option.

I'm kind of losing the idea of what the prefered look of 60i->24p should be. Are there any good examples of really good 60i->24p that don't involve some sort of 2:3 work for looking good on a crt?

Or perhaps my brain has shut down and I'm blurring all these things together.

Here's a 2-3-3-2 export (then exported to quicktime) of the same clip. The original Vegas 6 clip as 2-3.

While the frame/field patterns noticably vary between the three samples I've posted, I have no clue anymore which one is superior to the other nor why it is... I mean, if it looks right in motion, that's the most important part...

*head thump*


Vegas 6 - 24p w/ 2-3 (http://www.twodogfilms.com/Patrick/24p-vegas6.mov)

DVFilm Maker 24p (http://www.twodogfilms.com/Patrick/24p-dvfilm.mov)

Vegas 6 - 24p w/ 2-3-3-2 (http://twodogfilms.com/Patrick/24p-2-3-3-2.mov)

Patrick, you know, that is absolutely bizzare. The 2:3 vs. 2:3:3:2 cadence correlates exactly to the double/triple images in the files.

What's really weird is that 2:3 and 2:3:3:2 refer to interlacing methods when storing 24p in 60i files. It's a question on storage. 2:3:3:2 is the ideal storage format not for playback, while pure 2:3 is smoother on playback but isn't intended for storage.

Neither should impact what the image looks like at 24p. 24p is 24p. The frames should be the same, and the image shouldn't change based on the storage format.

It seems like Vegas is defining these terms in a new way. If you use 2:3, it blends 2 fields into a frame (double image), then combines the next three into a frame (triple image), and so on. 2:3:3:2 just changes the number and sequence of frames blended. That's completely backwards.

Like I say, it really smooths things out, but it's unlike anything I have seen.

I don't really have enough experience to know what to look for either. I do know that the Vegas 6 24p conversion looks smooth and pleasing to my eyes, and that on my own stuff I don't see any double images.

I will probably never transfer a project to film. My interest in 24p is really just for four reasons:

1: I think it scales better between 16:9 and 4:3 TVs. Interlaced 16:9 footage played back on a 4:3 TV goes through some kind of extra hardware deinterlace stage done by your DVD player as it's dropping every 4th line. To my eyes at least, 24p widescreen looks better as it is seen letterboxed on a standard TV.

2: 24P compresses better, both when you want to compress it down for the Internet or if you want to stick more than an hour on a DVD.

3: I like the way still picture and title animations look at 24p.

4: It is the look people are used to on their DVDs.

Lawrence, I'm with you on all those reasons and more.

The reason we are seeing the double and triple images is because the image contains tall, thin, vertical objects like the lampposts. On larger objects you won't see a "double image," but rather blurring around the edges of objects moving horizontally across the screen.

A bit of blurring/blending can be advantageous since 60i's 1/60 shutter speed is slightly faster than films 1/48 shutter speed. However, too much blurring/blending, you lose defintion and detail.

If it looks good, use it.

Hello,

Interesting thread.. Had not seen it before today.. i have FX1..And I am also interested in getting that nice 24p look for my DVD's.. Does everyone recomend just shooting in plain vanilla 60i to start out?? No picture profile settings , like cinema tone etc...???

Mike M.

Mike, congrats on getting an FX1.

If you can get the look you want in-camera, go for it. It will certainly be much faster than correcting in post. 60i is the way to go, though, for any 24p conversion or output to film.

for some reason when i render from vegas to 24p it keeps it 29.97/interlaced and im using the defaults for that profile. I have a NTSC wide project and I go to render as and tried both of the NTSC 24p Wide settings, tried both pulldowns and they both ended up giving me a 29.97i video

Hey Riley,

Just curious are you rendering to a file, or are you printing to tape??


Mike Moncrief

rendering a file

Hello,

I had a little bit more time tonight to experiment with what you were saying.. And I got confusing results.. I took some footage shot with my FX1 captured in Vegas 6.. Set my project size for DV widescreen 24p.. Then rendered out the footage to a file using the Ntsc DV Widescreen 24P (2-3-3-2 pulldown)
After the render i load the new file I just rendered out into Vegas as some media.. and Vegas 6 shows it being DV widescreen footage 23.976 fps..
Ok all looks well..

But if I load that same piece of media in After Effects 6.5.. It list the footage as DV at 29.97 fps....
Also if I use windows explorer and right click on the file to see its properties,, I see 29 frames per second listed???

Anyone have any insight at what is true here??

Mike M.

If you're using the DV format (regardless of 24p, or the pixel aspect), it'll always be 29.97. 24p encapsulated inside 29.97 just pads frames (all the 2-3 or 2-3-3-2 stuff in this thread) to fudge out 23.976 to 29.97.

If you want a true 24p file (23.976) without fields, render to another format - like a YUV quicktime file or a YUV avi, etc.

Patrick, you know, that is absolutely bizzare. The 2:3 vs. 2:3:3:2 cadence correlates exactly to the double/triple images in the files.

What's really weird is that 2:3 and 2:3:3:2 refer to interlacing methods when storing 24p in 60i files. It's a question on storage. 2:3:3:2 is the ideal storage format not for playback, while pure 2:3 is smoother on playback but isn't intended for storage.

Neither should impact what the image looks like at 24p. 24p is 24p. The frames should be the same, and the image shouldn't change based on the storage format.

It seems like Vegas is defining these terms in a new way. If you use 2:3, it blends 2 fields into a frame (double image), then combines the next three into a frame (triple image), and so on. 2:3:3:2 just changes the number and sequence of frames blended. That's completely backwards.

Like I say, it really smooths things out, but it's unlike anything I have seen.



I finally figured it out!

Here's a clip of 24p (from 60i) that's supposed to actually look how it's supposed to. The car shot from before isn't a good example using the 'proper method'. It's too unsteady (doh!) and it'll make the footage look jerky from the 24p method when really it's the bad camera movement.

Hopefully this is better. I've tried a few different things and I haven't noticed any odd reconstructed blended frames, nor stacattoing images that look downright odd.

24p finally (http://twodogfilms.com/Patrick/24p-finally.mov)

What Vegas is doing in the earlier examples actually makes sense now. Though I'm new to Vegas, I'm imagineing that this way it uses is actually superior to how it used to(?) and other ways of doing.

Do a 24p project and turn off Smart resampling for a clip. The deinterlacing is still in effect to clean up and reconstruct the image, but in 24p mode using resampling is blurring the sequence of frames (that was figured out in earlier posts) to make less obvious the differences between the shutter speeds. 60i converted to 24p gets that realtime-movement-yet-hyperreal shutter look to it and the resampling effectively eliminates that. The actual 60i -> 24p process also generates smooth (non 2:3 stuttery - but it is normal progressive stuttery) motion.