The best Vegas can do is blended frames. It will do this automatically if you turn on resampling for that clip. No other futzing necessary.

The best Vegas can do is blended frames. It will do this automatically if you turn on resampling for that clip. No other futzing necessary.

Kelly, thanks for that input. Even though I didn't mention in my original post, this was the first thing I tried, but it didn't work. However, after you mentioned it again, I thought for a moment, and then pounded my head -- duh! I hadn't done a Shift-M render, and of course Vegas doesn't actually blend the frames during playback; only during the render. Unlike fX, there is not approximation done for re-sampling; you actually have to perform the resample, and then do the preview.

Problem solved. I knew it was simple. Thanks!

Interesting! Glad i could help. I'll mention though that i see the blended frames in preview before rendering. *shrug* dunno.

OK, now you've got me curious. Recently I did a bunch of experiments with resizing video and supersampling. After these experiments I decided that I could see no real advantage of supersampling. I did all of these experiments with a shift-B RAM render. Do you mean to say that this doesn't really do the proper resampling, and that I actually need to render to a file to see the benefit rather than just a quick temporary RAM render?

Supersampling does absolutely nothing to existing video. It only affects generated media and motion caused by Vegas (such as a combination of motion blur and track motion). So, most of the time you won't see any change by using supersampling other than rendering taking much longer.

In any case, supersampling and resizing have nothing to do with each other. Supersampling is purely a temporal effect, not a spatial effect. It only alters time resolution, not pixel resolution.

Vegas does resample properly. You can easily see this by slowing a clip down to half speed, then stepping through the frames one at a time. With resampling on, every other frame will now be a blend between adjancent frames (this will be visible while previewing; no need to render). With resampling off you will see the same original frame for two frames, then an instant jump to the next original frame with no blending.

That's what I thought. So the fact that I just "shift-B" temporary rendered my tests shouldn't have affected my results, at least not as far as supersampling goes.

The resampling generated frames will just be blends of the previous and following frames then.

Is a program like Motion Perfect from Dynapel (at dynapel.com) any better than Vegas at this? I've played with the demo and it seems to do some kind of morphing to generate intermediate frames rather than merely blending frames. It seems to me this should be better, though I wasn't that impressed by the 15 to 30 fps test I did recently.

Is a program like Motion Perfect from Dynapel (at dynapel.com) any better than Vegas at this?

It can be, although it can also be much worse. I am actually ending up using a combination of MotionPerfect and the Vegas-generated blending suggested by Kelly (Chienworks).

MotionPerfect uses a totally different approach to generating the "missing" frames when going from 10 fps to 30 fps (or any other similar problem). It uses "motion estimation" algorithms, similar to those used in MPEG-2 encoding, to predict where each pixel in one frame is going to be in the next frame. Thus, the missing frames, instead of being a blend of adjacent frames and therefore quite blurry and indistinct, are totally sharp because they are the result of predictions of pixel movement. On certain types of motion, the results are absolutely stunning. On other types of motion, the results are bizarre and disturbing, A common nasty effect is to see someone walking in front of the camera and for a moment, their legs appear to break because the pixel prediction for the faster movement at the foot does not keep up with the pixel prediction for the slower movement at the hip.

I have long believed that this technology could be made to work much better, especially given the thousands of man-hours of development put into the prediction algorithms for MPEG-2 encoding. However, Dynapel (who owns the MotionPerfect code) hasn't enhanced the product in almost four years.

I have tried a freeware product called MVTools (it works inside of AVISynth) and have gotten remarkable results, but the author of that product has gone off in different directions, so that probably will never go anywhere. I have tried to get Sony interested in this, but so far have never received a response.

Sony, if you're listening, here are reasons you should develop this technology:

1. Killer slow motion (virtually 100% of all video editors use slow motion as some point. Having several ways of generating slow motion would be a huge competitive selling edge).

2. Ability to change frame rates. This is becoming a huge deal. Not only do still cameras take low-frame rate movies, but so do cell phones. Thus, all of us are going to need to be able to incorporate these clips into our work, and we will want to translate them to 25 or 29.97 fps, if possible. While I may not go out and shoot a wedding with a cell phone, if the best man gives me a clip he took while standing next to the groom at the alter, I'd sure like to use it and make it look really great.

3. Fantastic denoising. This is beyond the scope of this post, but if you have looked into what it takes to produce really fantastic temporal denoising, you really need to include motion estimation.

4. Dropped frame replacement. Do you want to make the Vegas customer base drool, and the video reviewers rave? Then provide a facility that gives us replacement of a glitched frame simply by pushing one button. How many people reading this post have ever encountered a single bad frame (dropped or missing frame, noise glitch, photo flash, etc.). What if you could position the cursor over the bad frame, push a single button, and have that frame replaced with a predicted frame that uses motion estimation information from several frames on both sides?

Wow, that would be hot.

But, as usual, I digress.

If you have to get the best results possible today, you need to use Twixtor Pro in conjunction with Combustion or After Effects (or DF, PP, FP, AS or any of several other platforms).

The review at the above link shows with very clear footage examples exactly what happens and what you can do.