thats why billyboy is on my ignore is cause hes always rippin at SPOT or at johncline. jeolousy hurts i guess. i guess thats as close as billyboy can get to someone good at what they do so he pokes at them because he cant do it himself. i never saw SPOT or johncline post about am i posting to much or do you want me to stick around? only billyboy does that. or zippyg
this is a good forum until that stink smells.

RE: We're looking at this the wrong way.

>> And probably SPOTS not explaining it that well either, but then again why does everyone assume he has too?

You think this bunch is going to let misinformation about Vegas stand unchallenged? No one's forcing him, but I"m glad he's willing :)

>> Hope this clears it up a little.

The question we are addressing is whether Vegas can do what has been claimed and if so, how to do it. The difference would not be that startling even with sophisticated image enhancement software - unless the video was shot just right... know what I mean?

EDIT: And on the way we are learning a lot! Skevos, nice catch on that Best (Auto) wierdness when you are at half size. I see the same thing... (it looks like a good match for the resolution difference in Spot's tutorial images - I hope he gets those fixed)

I'll never trust half-size previews again! :D

John Meyer has -- as usual -- made a detailed and specific post about his analysis of the original topic of this overall thread. He asked for detailed responses regarding how to interpret the confusing results of his experiments.

Now there are two messages posted in response to John's message that have nothing to do with his detailed question. This dillutes the sub-thread and makes his on-topic message more likely to be lost. At best, it jumbles up the whole thread and makes everything more confusing.

I see this happen over and over. Is it because some people are viewing the forum without "threaded view"? The forum is designed to keep things in a logical order, if the posters make their replies in a logical order.

Please, folks, when you post, try to have a message opened that has something to do with the theme you are replying to. Don't just make your post while you have the last message in the overall thread opened on your screen. Note the title on the forum posting frame that says "Reply to this message". Try to make sure that "this message" is one that has something to do with what you are about to say.

I realize that my message is also off topic to John's post, but I wanted to put it near another example of this frequent problem.

Please be sure to go up and read John's message that starts "RE: Sony Help! " if you haven't already.

For those who don't like to read, Supersampling is temporal.

It's very good to run tests, but the most telling test is to render the same project twice, once with supersampling turned up to 8 and once with it set to zero (Note: if you're using static images or generators with no motion, your project render times may be remarkably close. For the purposes of this test, don't do that.). After rendering both ways, compare the render times. The supersampled version will take roughly eight times as long.

Knowing from the help that supersampling at least enables temporal supersampling, one may then deduce that the eight-fold increase in render time when rendering with supersampling enabled comes from an eight-fold increase in the number of frames rendered. Were we to spatially supersample, render-times would increase eight-fold or sixty-four-fold again depending on whether we treated the supersampling value as an area or linear multiplier.

Smart Resample, Force Resample, and Disable Resample behave exactly as the help indicates. That is to say that they do not modify the footage spatially.

Chaboud,

Are you speaking for Sony? I didn't know you were a Sony guy. If not why change the title of the message?

You said:
>Smart Resample, Force Resample, and Disable Resample behave exactly as the help indicates. That is to say that they do not modify the footage spatially.

In John's summary of his tests, he said:
>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.

I know there was a lot of talk in the thread about some spatial effect of these options, but I don't see anything in John's message that implies that he saw or expected any spatial effect. As I read it, John asks specifically what "resample" and "supersample" ARE doing and how they might interact.

I don't see a clear answer yet.

John, feel free to chastize or correct me if I got your intent wrong.

chaboud (from Sony) said:

* For those who don't like to read,

Thanks! I think it's all pretty clear now - Supersampling does not affect resolution of individual frames, it's a temporal effect. Though in fairness to the guys like Spot and others that said or used images to imply otherwise, they don't say this because they "don't like to read", I think that the issue with what happens when the footage is viewed at half-project size in the Vegas preview window does give the illusion that something else is going on.

Since you're here Mr Sony person, can I ask why low-res footage looks so bad when viewed in the Vegas preview window at half-project size? I mean, the image quality in this circumstance looks far worse that the original low-res video file itself. Since it doesn't look this way in renders, and since higher resolution footage looks fine when viewed at half-project size, is this a, dare I say it, bug? I've never actually found a Vegas bug before! :-)

Sure, as bugs go it's hardly a show stopper, but at the very least this thread and others like it indicate that this bug causes some confusion, even amongst experienced users, as to what's going on under the Vegas hood. It would be nice to know if this odd behaviour (the munging of low-res footage when the preview window is at half-project size that is) is on a Sony bug list waiting to be squashed.

Thanks for your reply.,


Skevos Mavros
http://www.mavart.com

Geez, get on a plane and the thread goes wild....
Chaboud is one of Sony's engineers. And a very good one at that. Matt, you still playing vid games late at night? :-)
I just logged in, found the original file, and will look for the old veg on the archives tomorrow morning. Then I'll post so you can play with it if you want.

I actually meant "for people who don't wish to read my message."

Sorry for the confusion.

Are you working with footage that is, for example, 360x240 in a 720x480 project and working at a preview resolution?

If not, you may be falling victim to working at preview or draft quality. The spatial resampling algorithms (scalers in the video world) used in the faster quality settings are optimized for speed, rather than appearance. They don't produce erroneous data, but they don't interpolate new samples, either.

Of course, I'll take a look to make sure that nothing unexpected is happening. On your end of things, try upping the display quality to Good or Best.

>Chaboud is one of Sony's engineers. And a very good one at that.

Ok. I meant no disrespect. I didn't know that, and I usually see SONY (something) in posters from Sony.

Chaboud, my appologies for doubting you.

If by "video games" you mean fixing a somewhat obscure bug, then yes...

sigh...

Doubt away. I've been absurdly wrong before, and I'm far too thick-skinned (or thick-headed) to have been offended.

I didn't mean to imply that John had run bad tests or done anything wrong. It's easier to make a statement like "We do it this way" when the code is in front of you.

Note: I think that you should see a little icon next to the names of Sony employees who post on the forums.

>Note: I think that you should see a little icon next to the names of Sony employees who post on the forums.

Ahhh. The clueless become enlightened. I will try to remember that in the future.

Thanks for helping and I'm sorry I clouded the issues.

Hi there,

farss said:




Agreed.

* But the software isn't just working

Possibly - but not with Vegas!

* We see this all the time, we just

This effect does indeed exist, but it takes place within the brain as it "processes", for want a a better term, moving images. You are correct that a single 720x576 frame usually looks worse to a human viewer than a series of them running at 25 or 29.97 frames per second. It's an interesting phenomenon.

This is something we all discover when we do our first freeze-frame or screen grab from video. It can be mystifying at first as to why the single image seems so much softer or less detailed than the same image when displayed for a fraction of a second surrounded by similar frames, but the human brain is a remarkable thing. It pieces together, in real time, glimses of details from multiple frames and constructs a mental impression of something more detailed than any individual frame. Spooky.

* A low res video at low frame rate

Correct. So far no one has come up with software that can integrate multiple low-res images from a video file into a single more detailed image. Vegas certainly can't do this. The brain is a remarkable thing. If you know of any such software - link please! Gimme gimme! :-)

* Up the frame rate using a

Only when viewed in sequence - not when you look at individual frames. Which takes us back to your original observation.

* Using software to integrate a

What software are you thinking of? Vegas can't do this!

* Using very

I've heard of this idea (combining multiple low-res images to obtain a single more-detailed image), but the results are, I believe, aimed at the image analysis field (for examining intelligence photos etc) and the results look kinda odd and need to be interpreted by experts. It's particularly good for mentally filtering out artifacts caused by the picture-taking process, and for spotting changes that have occured between when the photos were taken. But I haven't seen anything available or suitable for video editors to use to generate good stills from multiple low-res images. Have you? I'm not challenging you - I'm genuinely interested. I'd love to get my hands on that kind of software, it would be very handy, not to mention fun!

I've also seen motion vector math used for "intelligent deinterlacing", to very good effect. As we all know, de-interlacing effectively halves the vertical resolution of interlaced video (though strictly speaking it depends on the method used - if the fields are interpolated the results are effectively kinda sorta better). There is software out there that uses motion vectors to only deinterlace those parts of the video that are in motion (in motion compared to previous and subsequent frames), leaving the motionless areas interlaced and hence at full resolution. This is remarkably effective, since the moving parts of the image are usually motion-blurred anyway, and hence the lowered resolution caused by de-interlacing is barely noticed. I believe there's a plugin for VirtualDub that uses this technique. It would be nice to have this in Vegas, but I'm on a tangent now as this technique does not increase detail, only preserves it in those areas of the image not in motion.

* Hope this clears it up a little.

Not really! :-) Since the capabilities you refer to are not possible in Vegas, and hence those screen shots on both tutorial pages are a tad misleading. Which is all I was trying to say in my first post on this thread. Supersampling is a great feature, but lets not put up images that make claims about it that aren't true. It's a temporal effect, and a bloody cool one, but it's not a generate-spatial-detail-where-there-was-none-before effect.

But if you have links to software that can do the sort of thing you mentioned, please share! :-)

All the best,


Skevos Mavros
mavart@mavart.com
http://www.mavart.com

Hi chaboud,

Thanks for your reply. you said:

* I actually meant "for people who

Okay, sorry I misunderstood you - I thought you were taking a swipe at people. My bad. :-)

* Are you working with footage that

I can't speak for the first tute as it's unclear on this sort of detail, but the second tute is an NTSC project working with footage at 160x120x24fps. My screenshot from Vegas plus the Quicktime player (333KB) shows what this footage looks like when the Vegas preview window is at half-project size. As you can see, the scaling of the video is terrible, even at Best quality. Just compare it to the actual video file playing in the Quicktime player.

Sure - it's not a big bug, but I suspect this bug is the actual cause of the confusion around the supersampling feature. Yes yes, we should all read and trust the Vegas docs and help, but if the software looks like it's doing something else, you can't blame people too hard for sharing that information.

I just noticed another quirk. If, in the above example from the screen shot, I resize the Vegas preview window smaller by clicking and dragging the corner of the Preview window, the resize of the video looks great - until I let go of the mouse.

Bug?

* If not, you may be falling victim to

Good point, but I checked - this issue happens at all quality settings.

* The spatial resampling algorithms

And I love them. They allow me to have a client sitting next to me requesting changes to multiple layers of effects and I can show them the changes in real time, no rendering, at close to full frame rate. 99% of clients are happy to watch a low-res preview if it means not waiting for renders. Sure, there are some effects that just have to be rendered in order to fully "see" them, but in general I love the Vegas quality settings as they give me much of the funtionality of a real-time system without the cost! It was one of the things that made me switch to Vegas.

But the resizing of low-res video in the preview window at anything other than project size seems to be broken.

* They don't produce

Agreed - how could they?

* Of course, I'll take a look to make

See the screenshot above - as you can see I was at Best.

All the "best"!



Skevos Mavros
http://www.mavart.com

One could write a sampler to produce erroneous data (rounding aside), but it seems like a great deal of effort to produce the wrong result.

Your example image's content is well outside of the scope of the degredation that could result from point-sampling. I was unable to duplicate your example on my running copy of Vegas and was about to reply to that effect when I thought to make sure that it didn't happen in Vegas 4.0. It does.

There is definitely undesired output there, and it's fixed internally. Look for the fix in the next version.

I was unable to duplicate your example on my running copy of Vegas and was about to reply to that effect when I thought to make sure that it didn't happen in Vegas 4.0. It does.

Heh - so I guess that means that your "running copy" of Vegas is v5? Lucky you!

There is definitely undesired output there, and it's fixed internally. Look for the fix in the next version.

Ah well, another reason to upgrade! Not that I needed any extra ones... :-)

Thanks!

Skevos Mavros
http://www.mavart.com

chaboud> "Supersampling is

Thanks for clearing this up! Very similar to Kelly's initial response... 600 posts later LOL

Marquat>Skevos, I agree with farss' statement: "Using software to integrate a series of frames and you can extract a better looking frame". Look at Ben's Examples 3 & 4. Notice how smooth Example 4 is. IMHO the difference in the two pictures sums up the supersampling envelope.

Just to be clear, remember that in the Examples the reason #4 is so smooth is Vegas rendered 80 images internally (10 frames from the motion blur multiplied by a super sampling setting of 8) and then blended them together to create that single frame. This was only possible because it was keyframed animation.

I think farrs was talking about improving existing video. If video was shot at 120fps (or maybe 3d animation rendered at 120fps), then you might want to use Supersampling to improve the look at 30fps, but otherwise it's only useful for smoothing the look of keyframed animation.

- Ben

Some time again I did find a thesis paper on this but a quick google and I couldn't find it again. The technique involved starting with a sequnce of frames from a video camera that was slowly panning across a scene. Motion vectors were derived to re-align the frames and then a LOT of maths that went right over my head that did some form of frame integration to produce a much higher resolution image.
The thesis did contain the source code so it might be interesting if I could find the thing again. Then again maybe it's already in use in various skunk works.
I did see that frame integration seems to be a fairly common technique used in still cameras to reduce noise. That's a bit different to using it to enhance resolution though.
I was only speculating though and it would now seem that there's a much simpler explaination for what's going on.

O.K., Supersampling is temporal. Based on cahboud's statement: "the eight-fold increase in render time when rendering with supersampling enabled comes from an eight-fold increase in the number of frames rendered," if I crank up the number on supersampling, it is going to create a multiple of the number of frames. Instead of thirty (nominally) frames in a one second clip, I will have 240 (using the Supersample set at 8 example).

Confusion #1: Since my final one-second video clip will only have 30 frames, what happens to those other 210 frames? If I slow down the video, I can see the need for more frames, but if I use a playback rate of 0.250, this will require 4x the number of frames. Therefore, is there any point using Supersampling of more than 3 (or 4)?

Confusion #2: What is the difference between Resample and Supersample? I thought I had it figured out when I though that one was spatial, and the other temporal. However -- and I will probably reveal my confusion/ignorance in this question -- it sounds like BOTH are temporal, i.e., they both create intermediate or additional frames based on information from one or more adjacent frames.

Obviously the reason for these questions is the desire to know when to use -- and when NOT to use -- these features, and to have some idea of what we can expect from each of them. The first step is trying to understand technically what is going on. However, most people don't care about the technology -- they just want to use the darn feature. To this end, it might be useful for the documentation people at Sony to some up with a matrix that shows various scenarios where these two features might be useful (or might not be). It might look something like this (which I am quite certain is wrong):

			Resample	Supersample

DV Video			No	No

Low Res video -> DV Video	Yes	No

Slow motion		Yes	Yes

Keyframe video		No	No

Keyframe generated media	No	Yes

Still photos		Yes	No

Keyframed photos		Yes	Yes

Here's the difference, as far as i can see both from reading the manual and experimentation:

Resample functions during the rendering process on ANY material which does not match the project/rendering frame rate. It creates enough new inbetween frames by blending adjacent source frames to match the final output frame rate. So, if your source is 15fps and your output is 30fps then it will create output frame 1 from input frame 1, output frame 2 from a mix of input frames 1 and 2, output frame 3 from input frame 2, output frame 4 from a mix of input frames 2 and 3 .... etc. This is why you can get doubled/ghosted images when slowing video.

Supersample generates inbetween frames ONLY on frames that move due to pan/crop, track motion, or other effects before rendering. It has NO effect on motion that is in the source video. Considering Ben's excellent example of a checkerboard being rotated with track motion. If you want to blur the image then Vegas will show a small amount of the previous frame mixed with the current frame, and a smaller amount of the frame before that, and an even smaller amount of the frame before that ... summing them all together to create the blur. However, at any given frame rate the amount of motion between one frame and the next may be great enough that the result shows distinct multiple images rather than a smooth blend as seen in the third image. Supersampling will create more frames of smaller movement so that it can blend these together more smoothly, as seen in his 4th image.

Resample won't produce the extra frames necessary to create the smooth blur; it will only produce enough frames to match the output frame rate. Supersampling will go beyond this to create more frames for a smooth effect even if most of them end up being thrown away after the blending process.

Resample works on ALL motion. Supersample ONLY affects motion generated by pan/crop, track motion, and other movement from the editing process.

>>Supersample ONLY affects motion generated by pan/crop, track motion, and other movement from the editing process.<<
So does this mean I was not crazy or blind when I could see absolutely no difference in comparing slo-mo that had and had not been supersampled?
Randy

Randy, right. Ordinary resample is sufficient for this task. Supersample won't do anything because the motion is in the video itself rather than being created by Vegas.