The June thread started with a simple question - What is supersampling - and centered around one of Spot's tutorials on the Sundance site:
The tutorial
We would all like to be able to get that kind of results and some of us were asking questions about it. Even SonyDennis took an early interest in the thread.

Still, questions were left unanswered. That is very rare in this forum. Now the questions have come up again, this time Spot is aware of them, and I'd like for all of us to calm down and wait for the answers. I'm sure Spot can regulate his office/family hours perfectly without our help.
Tor

The main difference as I see it is "Resample" only calculates frames using existing frames from the - and then blends them to create in-between frames to similate a higher frame rate when needed for slo-mo or when using 15fps video in a 30fps project.

Supersampling on the other hand calculates up to 8 frames in between . Since there is no image data between the frames of video, this is only relevant when you are using keyframed events in the project.

For example, quickly panning across a still image using the PAN/CROP in a 24fps project might give you a stuttery look. Motion blur will help, but it will only use existing project frames to generate the blurred frames. So you would have to bump up your frame rate to get smoother looking motion blur. Instead of increasing your project frame rate, you might want to enable Super Sample on top of the Motion blur so Vegas will render several frames that would fall IN BETWEEN project frames and use those images to create something that looks closer to true motion blur. Again, super sampling is only meaningful when smoothing KEYFRAMED animation in Vegas.

You can enable Super Sampling without Motion Blur, but then it will only generate the in-between info from 1 single project frame's worth of time. That is useful when your keyframed animation creates a lot of motion between frames - in other words, very fast animation. Increasing the motion blur makes Vegas use more and more adjacent project frames in the calculation - but if Super Sample is set to 8, it will still generate 8 time slices between each and every project frame encompassed by the motion blur setting - so processing time goes up a lot as you incorporate more and more frames into the calculation. Multiply your motion blur setting by the resample setting to get the number of images used to generate a single blurred frame.

The confusion with the infamous tutorial images ;) are at least partially due to the two screenshots not being compressed / downsampled the same way. When you downsample a still in Photoshop etc... you can choose a method. I think someone screwed up and used different methods for the images. It's easy to see that that the media player is sharper in the bottom image.

But that's not all. The "frame" from the bottom image has far more resolution. Supersampling may have this effect in very specific circumstances - much like the software that uses surrounding video frames to calculate a higher resolution version of a video frame. It only really works if the subject isn't moving and you sort of jiggle the camera around on it. It wouldn't work on normal video.

>>Only a dork looking to create an issue would assume it matters whether it's the same exact frame or not. It's the whole file that's affected.

Us scientific types are usually dorks, but honestly I don't think even NASA could have gotten the results claimed in the tutorial.

Science demands that everything remain the same - from using the same frame to using the same settings to resize the screenshots - everything except for one variable, which would be the thing you are trying to demonstrate to us. Something is different that's for sure. Notice the triangular objects in the center of the frame. It might as well be solid in the top image, but in the bottom image you can see it has holes.

Here's the problem... I followed the tut and tried to duplicate the results but didn't see anything. And much of the information on that page sounds very "enthusiastic" - almost like marketing:

"When used to upsample low resolution/low framerate video such as this original mpeg file seen in Fig 1, Vegas simply outsshines anything."
...
"For this single tool alone, Vegas is a veritable hammer in the toolbox of any video editor. Check out the Vegas 4.0 demo from the Sonic Foundry website."

You have to admit that's a pretty bold claim (marketing or otherwise) considering none of us can duplicate the results.

This line is also a problem for me:

The way this works is that Vegas creates interpretive frames based on the difference between the project framerate and the framerate of the media, or the computer-generated imagery.

What exactly is "interpretive frames"? From what I see, unless you are using generated media with keyframe animation within Vegas, you are just talking about "blending frames" from the source media.

By the way, super sampling is an AWESOME feature when used as it was intended. If you want to see what it can do, generate a "Checker board" pattern using generated media, and create an animation of it rotating very fast. Ahhh hell I'll post some screenshots! :P

http://ben.orona.com/video/supersample/supersample.htm

Notice the last image - when used on media in which the spin has already been rendered the effect is almost useless.

- Ben

Ben,
thank you, thank you!
I've been churining this over in the back of my head for the last few days and even after I read your post it still didn't quite add up!
I couldn't for the life of me think why creating more intermediate frames than were needed made any sense apart from blowing out render times.
Then I looked at your screenshots and a penny started to drop. A few weeks ago I'd done a keyframed zoom to get an object flying in and the motion blur just didn;t look that realistic, you could clearly see the previous frames mixed in, more like strobe effect than a motion blur. Adding the tween frames gives the motion blur more to work with i.e. more temporal resolution resulting in what looks like a comet trail.

This all makes sense, and again a lot of head scratching could have been avoided with better documentation.
Still leaves the mystery of the apparent increase in spatial resolution that SPOT seems to have achieved. Although that may not be beyond the bounds of possibility now that I understand what is happening although I'd have to agree there does seem to be way too much marketing hype and not enough detail in what was being said.

Certainly looking at your screenshots, comparing the last and second last images I'm thinking I can see something like what SPOT was showing happening under very specific circumstances.

BillyBoy, I've used video super sampling and it's a great tool that, like so many tools in Vegas, is underappreciated and underused. Sure, it hits the render times pretty hard, but if you need it for a particular issue there are very few apps out there that can do a similarly good job - and this one is built right into Vegas waiting for you to use. Fan-bloody-tastic. And I know it seems counter-intuitive, but what Spot said about adding a little blur to make a low-res image look sharper when upscaled is paradoxically true, though I personally would prefer to phrase it that adding a little blur makes the image look "better" rather than "sharper", as it doesn't confuse people as much that way.

Having said all that, I took one look at this supersampling tutorial web page and saw the before and after images at the top of the page and I knew right away that there had been a mix up somewhere. Spot, I don't know how long it's been since you took a close look at that page, but you need to take another look at it, as it's clear to me that there has been some kind of goof up in the images that were uploaded to illustrate that tutorial. You need to change either the before or the after image (or both).

Spot, before you reply to me please note that I'm not referring to any seminars or streaming tutorials you may have done - I can't refer to them as I haven't attended any. I'm just referring to this web page and nothing else. Also, when you reply please remember that I have the highest regard for you, your editing skills, and your Vegas knowledge. I know we disagreed recently on a small element of the complex copyright issue and that you seemed to take offense at what I said (you thought I was calling you an "idiot" when I truly wasn't), and I'm loathe to cause you any further discomfort or annoyance, but the web page I'm looking at on my screen, the one showing before and after supersampling, is clearly false.

I'm sure it was an honest mistake, and perhaps it was a mistake made by whoever put the page together, but there is no way that the second image could have been derived from the first (I'm not referring to the "it's a different frame" issue, I'm referring to the increased resolution of the second image). I know the Vegas supersampling feature is good, but it isn't that good! :-)

If I had to guess how the error on that page occurred, I would suggest this: I know, from experience, that it can be tricky getting screen grabs of the contents of the Windows Media Player window, since on most people's systems Windows Media Player uses the Video Overlay feature of their video card/chip to display the video images. This usually results in most screen grabbing software giving you a black nothing where the video should be (since the video is rendered on a different "layer" to the rest of the desktop). I suspect what has happened here is that an image editing app has been used to paste the before and after images into the WMP windows in the screen grabs. The before image seems to be a "simulation" of what the image looked like before supersampling. All I'm suggesting is that whoever made the simulated "before" image slightly over-egged the pudding - perhaps in an attempt to clearly show the difference between the low-res screen grabs.

I know the images aren't true because:

* the before resolution resolution is too low to support the claimed results in the after image, no matter what software was used. Outside of Hollywood, software cannot create detail where there simply isn't any - check out the creases behind the knee of the foreground character for example,

* the before image is simply "pixellated", whereas Windows Media player uses a softer algorythm when upsizing low-res video in real time - WMP does not look this "chunky" or "square" when playing low res video, and

* there's also the fact that the supersampling feature of Vegas does not affect the resolution of video at all, only generated media and keyframed motion, etc. The text on the tutorial page itself confirms this. It says "The way this works is that Vegas creates interpretive frames based on the difference between the project framerate and the framerate of the media, or the computer-generated imagery. This also creates smoother flow and edges for generated behaviors such as Pan/Crop, Track Motion, transitions, and other 'new' media created in the timeline during the editing stages."

There is no way that adding a little blur and supersampling could ever produce the after image if given that before image. One or both of those images needs to be replaced.

Like I said, I'm sure it was a simple honest mistake made when preparing the screen grabs for the page, but I urge you to take another look at them and replace them with truer representations of the feature (perhaps animated gifs, since the effect is principally a temporal one?). The last thing we want is people buying Vegas under false pretenses (not that anyone would buy Vegas for just one feature - but still).

Spot, please take this carefully worded post in the friendly spirit it was intended.

All the best,



Skevos Mavros
http://www.mavart.com

Just wanted to weigh in on a positive note as I can see from where I sit that things aren't always easy, especially when you're constantly being leaned on as The Expert and expected to perform by others. There's a few shades of difference between asking a question of you out of necessity, and one feeling entitled to an answer -- quick or otherwise -- and it's not clear that most people see the difference.

I hope you know that some of us get exactly your intentions, your points, etc. that you're trying to make clear for other here, and we're thankful that you're here.

Not that anyone is above criticism or disagreement -- I've had my share of disagreements with you in the past, and have myself made posts online here and elsewhere that -- given the crap that Life In General has been slinging at me at a given time -- I regret later and feel poorly represents my otherwise good intentions. I expect you go through some of that yourself.

But the above posts couldn't be further from the kind of turbulence expected over a given time -- I'm just putting this up to let you know that all your hard work is appreciated by many, that you presence is quite literally invaluable.

- jim

Vitamin D said:

* all your hard work is appreciated by many,
* that you presence is quite literally invaluable.

I agree! I actually had zero opinion on this particular topic until I saw those images on that web site. Spot is definetely "the man" when it comes to Vegas!

That's why I want to see those images modified.

All the best,



Skevos Mavros
http://www.mavart.com

Which is all I'm saying Skevos. The current tutorial, the one that's the subject of the older thread and what I reported for clarification isn't telling the whole story. SPOT has a habit of attacking the messenger rather than admitting he messed up.

The issue isn't supersampling itself, rather HOW it was presented in the tutorial which can't be duplicated following the tutorial's instructions and using its source material. For SPOT to imply oh you don't know how, come watch me to do it is again self-serving bullshit suggesting those questioning HIS TUTORIAL don't know how to use supersampling which is a weak defense to massage a bruised ego.

Why the author of the tutorial gets huffy simply because many of us pointed this out is probably more to do with personality. Some people can't handle construcitve reviews and engage in name called (billybob) or get condescending, come to my show I'll show you.

If anyone notices errors in any of my tutorials tell me. I'll gladly fix them and admit I was wrong. No brusied ego here. I'm just trying to help. Not trying to peddle a book.

Since there is such a buzz about this tutorial of over a year ago, I've now gone in and rebuilt the project. Hope it's worth the several hour download/upload and the $$ cost. I grabbed a smaller vidfile this time. Rather than showing it in WMP, all screen shots are taken from Vegas.
The page is located at http://www.sundancemediagroup.com/tutorials/supersample_2.htm
Hopefully this will clear this up for some of you. Skevos, again, I've done this process dozens upon dozens of times in front of seminar attendees. you might feel it couldn't possibly come from the same file, but it does. I suppose that we'll need to upload both the original and finished MPEG, which is what's shown in the tutorial, hence the use of WMP.
BB, I'm sorry you don't feel I "share" and that I don't know what I'm talking about. All I can respond with there is that while I may not post as much as you do, I believe I give what time I have pretty freely between the 11 forums I moderate plus the many others I participate in. Somewhere in all this, I do have to generate an income, have a life, and enjoy a few minutes to myself. As far as me not knowing what I'm doing, for a dummy, I've been modestly successful with my work in both video and audio. Care to share anything you've done for broadcast, DVD distribution, corporate spot, or other publically accessible project? The test isn't in what you know, but rather one of properly applying what little you may or may not know to create a great work of art regardless of your skill or training.

Very impressive!! Thanks Spot!

Ben, thanks for the clarification and the pics. This is probably the clearest example yet of the practical aspect supersampling and motion blur have on video. You've cut the proverbial Gordian knot in half. :)

1. I don't need to 'admit I messed up' because I didn't mess up. I may not have explained it in a way that is clear to each and every reader, and since the tutorial didn't cost you anything except the time to read it, you don't really have a valid gripe.
2. I'm not 'trying to peddle a book' at all. It would be oh so easy for me to say "go buy the book if you want to know the answer" but you won't find a single, solitary post out of my thousands of posts that even suggests that. Marketing the book isn't my job. I'm proud it's out there, proud of the new edition coming soon, and very happy for another author out there that's doing his version of a book on Vegas. It's all information. No one is obligated to share with you, it's done out of love for the craft. For some reason, you seem intent on pissing on anyone who's been successful with Vegas or any other tool, BB. Why is that? I say "come watch me" because then we can have a dialog about it if I say something that's not clear or if I go too fast. I apologize if my no-cost tutorials have errors either by omission or typos. I think in fairness, anyone in these forums, save it be yourself, would agree that when I'm wrong, I apologize and make it correct when possible. I've owned up to errors here and in other forums several times. No one attacked you, whereas you tend to refer to me as "the SPOT" in your posts, coupled with your constant condescension and need for attention does indeed sometimes garner a caustic response. "Am I posting too much?"

First, my apologies as the intention of my post wasnt to cause strife on this forum. It does seem like the question has been answered.

To Spot, when I asked about the delay of answering, it wasnt taking a shot at you. It was merely asking why it was never answered in the thread from last year but as you said, you didnt see it. I respect all on this forum (minus Zippy) so please dont take it as being called out.

It does bother me a bit to see some of the more knowledgeable people on this forum bickering back and forth. In the words of Rodney King, " Cant we all just get along?"

I don't take kindly to someone deliberately calling me 'billybob' suggesting I'm some rube in barefeet and bluejeans. It isn't the first time SPOT hasn't been able to control his temper anytime somebody disagees with his opinion. Nor is it the first time he needs to resort to look at me, what have you done defense.

I'll I said was the ORIGINAL tutorial was incomplete, misleading and IMPOSSIBLE to duplicate the results of. All the more experienced users know that's true. Instead of SPOT saying oops, sorry guys he goes into another rant, its free take it or leave it defense.

Now who is really is being petty?

That is pretty good it worked for me

and I thought it was professor plum in the conservatory with the candlestick :-)

Paul

BillyBoy,
Breathe fully in... hold for two seconds... Breathe out slowly. Repeat ten times. You seem all riled up about something, and even though I found similar issues with the tutorial you don't see me using some of the terms you've used. Please stay calm, relax, it's only software man! :-)

Spot,
Thanks very much for preparing the second tute/clarification, though in the end it confirms my concerns about the first. Either I didn't make myself clear the first time or we're looking at different things. It's 4am here and I hope I do a better job explaining myself this time, or else I'll try again tomorrow.

At first glance the first pair of images in your new tute look exactly as I would expect them to look, based on my use of Vegas in general and the supersampling feature in particular. We are agreed are we not that supersampling is a temporal effect and doesn't actually affect the content of any individual video frame, yes? It's Vegas' built-in resizing algorythm and the small amount of blur that make the resize so nice. i know to others it can seem odd that softening can help a blow up, but I too know this is true (and screenshots on progressive scan monitors don't really do it justice - somehow these things look even better on an interlaced video monitor - not sure why, perhaps it's the contrast/gamma difference?). So all is well with the first pair of images (interesting footage by the way).

It's the second pair of images that concern me, the titles. At first I laughed out loud, convinced that you were playing a late April Fools joke on us all. There's just no way that the first image could become in the second image under any circumstances, particularly not with supersampling. Just look at that first text image, at how soft and blurred the text is, then compare it to the second image, the defined curves in the capital S's for example. Impossible.

Sure enough, I duplicated your efforts within Vegas and noticed no improvement at all. But my Vegas preview window was at full project size, and I noticed that your screen grabs were of project windows at half size. And then I realised what was happening. It wasn't the supersampling that was improving the image quality, it was the preview window being at half-size that was softening the image!

Don't take my word for it - perform the same before and after tests but with the preview window at full size and you'll notice zero difference in image quality - especially on the text. Or try leaving the preview window at half size and rendering out full DV files of just the titles - one DV file with supersampling on and another with it off - no differences in video quality.

Download this screenshot from my Vegas and Quicktime player (333KB) and you'll see that while the NTSC Vegas preview window at half size is slightly larger than the same video file in the Quicktime player, the Vegas preview window contains much less information. The Vegas preview window seems to greatly soften the image when viewed at half size. This gives the impression that supersampling is creating "detail" out of nothing, but it is an illusion. Look at the Quicktime player - all the information is already in the file, it just looks bad in the Vegas preview window when it's at half size. All this has no effect on renders however, and neither does supersampling.

Mystery solved! :-)

As for the before and after images in the first tute, although they appear to be within windows media player, I assume they were actually taken from Vegas and pasted into those screen shots? This would explain the differences, though perhaps there are other reasons. I still recommend you change them as they, and the second pair of images in your second tute, give a misleading impresion of what supersampling does. Supersampling does not affect the images within a video stream per se, only generated and keyframe effects.

Glad we could clear this up.

All the best,


Skevos Mavros
http://www.mavart.com

Referencing the example shown at www.sundancemediagroup.com/tutorials/supersample_2.htm

Now I understand the statement about softening resulting in the appearance of a sharper picture. Very impressive.

there is a big difference when using preview auto for the reasons you mentioned.

I have rendered the clip and there is still an improvement viewed full resolution and on an external monitor preview

Not as marked as the quirk with the resolution of the preview window but it still is there.

Regards, Paul

Oops, if the preview resolution is set to best full I can't see a difference

Sorry about that

Paul

Skevos: "We are agreed are we not that supersampling is a temporal effect and doesn't actually affect the content of any individual video frame, yes? It's Vegas' built-in resizing algorythm and the small amount of blur that make the resize so nice."

I got the same results. I can't see any difference in spatial resolution using the supersampling envelope on Spot's MOV, or any other video I've tried it on.

For those who may have skipped my long post, here are some screenshots that show what supersampling is for:
http://ben.orona.com/video/supersample/supersample.htm

Feel free to post it someplace where it might be more useful.

Just remember to ALWAYS keep your preview window set to Best (Full) when playing with motion blur, deinterlacing, or super-sub-up-down-or-re-sampling ;)

edit: forgot "under"

Good thread going here. I have nothing important to add, but one comment by Steve caught my attention:

>i know to others it can seem odd that softening can help a blow up, but I too know this is true

Here's a trick I discovered one day. Next time you are watching TV and you see shots where a person's face is masked with moderate pixelation, try this. Squint your eyes so that your vision blurs a bit. Often this will give you a pretty good idea of what the person really looks like.

This is a real world example of one effect Steve is describing in this thread. I have to agree, though, that I don't see supersampling in Vegas having much to do with this effect.

Skevos,

I am glad you took the time to clear up the confusion. All the sniping that BB started really clouded the issue.

In an attempt to further look into this, I performed the following test:

Using my photo editing program, I created a 720x480 still image of a black circle on a white background. I put this into Vegas, set the pixel aspect ratio to 0.9091, keyframed the circle to move across the screen, disabled resample for the keyframed event, and then rendered this to a DV AVI file.

This gave me a test file with a black circle moving across a white background.

I then created a new Vegas project and put this AVI file on the timeline. Just to check, I moved through the file, one frame at a time, and made sure, both on my computer monitor and on an external monitor, that the spatial quality of each frame looked exactly the same, one frame to the next (i.e., I hadn't screwed up somewhere and ended up with interpolated frames). Each frame looked the same, other than the fact that ball was moving diagonally one frame to the next, from upper left to lower right.

Next, I set the playback rate for this video clip to 0.250. This forces Vegas to do something to create new frames. I then rendered about one second of this clip to four different AVI files. I used the four possible permutations of Resample and Supersampling, as follows.

File 1: Disable Resample, No Supersampling
File 2: Force Resample, No Supersampling
File 3: Disable Resample, Supersampling set to 3
File 4: Force Resample, Supersampling set to 3

I then created a new project and aligned these four files on four tracks above one another. I next soloed the file #1 and moved one frame at a time. With resampling disabled, and playback rate set to 0.250, the ball does not move for three frames, and then moves on the fourth frame. This pattern repeats.

I moved the cursor until it was on the middle of the three frames that are identical.

I then soloed the first track. I used the snapshot script to take a full 720x480 snapshot and brought this snapshot into a photo editing program. I un-soloed the first track, then soloed the second track, and repeated the snapshot. I did this for all four tracks.

Once I had the snapshots from each track in my photo editing program, I zoomed to 3x, and compared the results.

Three of the snapshots were identical. The only different one was the one with Resample disabled, and no Supersampling applied. For this frame, resampling, supersampling, and the combination of the two all produced the identical result.

Hmmm.

I then repeated the test for each of the other three frames.

When I did the test on the last of the three "new" frames, the two "Force Resample" renders were identical, regardless of the Supersample setting, but the "Disable Resample" combined with Supersampling was different from Disable Resample by itself, not only in spatial quality, but also in location. Disable Resample/No Supersample was the furthest to the left (the circle was traveling to the right). Disable Resample/Supersample was furthest to the right (farther to the right than the two identical Force Resample tests). It was also more blurred than the two Force Resample tests.

I got the identical result when I did the test on the first of the three "new" frames.

Finally, I did the test on the first of the four frames, i.e., the frame that, when you move one frame at a time with the "disable resample/no supersample" track soloed, is the frame where the circle moves to a new location, followed by the three identical frames. In this test, the only track that gave me different results was the one where I had disabled resample, but had superampling enabled. The edges of the circle were smoothed, and the circle shifted slightly to the right. The other three were identical.

Conclusions

I cannot say for sure what Vegas is doing, or what order these two effects are applied. I would love to have Sony comment on this. However, my suspicion is that resample is creating additional frames by blending adjacent frames, whereas supersampling is actually up-sampling the resolution before fX are applied, and then downsampling back to 720x480 when finished.

			Frame 1	Frame 2	Frame 3	Frame 4	Frame 1	Frame 2	Frame 3	Frame 4

Resamp: No/Supersamp: No	Left	Left	Left	Left	Coarse	Coarse	Coarse	Coarse

Resamp: Yes/Supersamp: No	Left	Center	Right	Center	Coarse	Smooth	Smooth	Smooth

Resamp: No/Supersamp: Yes	Right	Right	Right	Right	Smooth	Smoothest	Smooth	Smoothest

Resamp: Yes/Supersamp: Yes	Left	Center	Right	Center	Coarse	Smooth	Smooth	Smooth