RGB to Studio RGB - confusion reigns....

[Phil Lee]

Hi

Thanks. I found this link gives very good info Color spaces and levels in Sony Vegas 9 and 10

Re-visiting some projects it all makes sense now. In most cases we don't need to apply any level filters, which explains how most people are getting good outputs level wise.

The majority of video most people will use (DV/MPEG2/HDV/AVC) will sit on the timeline as Studio levels, exporting out to the same formats in Vegas requires Studio levels, so all is good. The only time you will need a levels filter is if you add images to the timeline, these are added with Computer levels, so you need to add a Computer to Studio level filter to bring them in-line, and they do look a lot better.

I think Adam was advocating a different approach with a timeline using the Computer RGB level, so all video clips get a Studio to Computer RGB level filter, and any images are fine as is. This has the benefit of making the preview in Sony Vegas correct as it wants Computer RGB. (Not an issue if you have a second monitor as you can tell Vegas to use Studio or RGB levels on a different monitor).

On outputting Computer RGB, you need to then revert back to Studio levels with another filter applied global to the timeline. This works fine, output will look identical, however from reading some more about it most camcorders will record levels and so detail (considered illegal) above 235 and below 16. When you apply a studio to computer level filter it chops everything below 16 and above 235 then stretches it to fill 0-255. Adding this to a clip on the timeline means you have lost some detail that resided below 16 and above 235. Now that detail will get lost somewhere when you convert for Blu-ray/DVD etc anyway as it isn't legal, however having the extra information on the timeline can give you more latitude for exposure changes, or you can add a filter to move the detail into the legal range so you don't lose it completely. So my thoughts are it is best to stick work with Studio RGB levels on the timeline, which I had been doing when I didn't really have a clue what it all meant, it just looked okay when outputted so I didn't question the workflow.

Regards

Phil

[Adam Stanislav]

Originally Posted by Phil Lee

So from what I understand the important bits are to keep all footage on the timeline at the same levels, and the best level is the one that most footage is in by default?

Keep the footage on the timeline, whether 8-bit or 32-bit (but preferably 32-bit), in the computer RGB level. Otherwise some effects will not work to their full potential.

Example: An effect may calculate the square root of the value of a pixel, the basic algorithm of brightening the image while preserving both the shadows and the highlights. When working in the computer level, working in 32-bit mode, white will remain white (square root of 1.0 = 1.0), black will remain black (square root of 0.0 = 0.0), and everything else will be brighter (e.g., square root of 0.5 = 0.71).

If you apply the same filter to studio levels, black (16 / 255 = 0.0627...) will become a lot brighter (0.25..., which converts to 0.25... * 255 = 63.87 = either 63 or 64, depending on how the filter converts floats to integers), which is totally wrong. Your shadows are no longer preserved.

Even worse, in that situation, white (235 / 255 = 0.92...) will become 0.96 which when multiplied back by 255 will become 244 or 245, way outside the 16-235 range. Your highlights will be crushed.

Not all filters are affected this way, or at least not as drastically. So it may seem that just keeping your timeline on the same level is OK even if it is all at the studio level. But sooner or later, a plug-in will appear not to be doing what its author says it is doing, simply because its math is thrown out of whack by an incorrect range on the timeline.

So, it is not enough to keep the timeline at the same level, it is essential to keep it at the computer RGB level. Every single book on computer graphics that I have ever read—too many to count—assumes the data being manipulated is in what Vegas calls computer RGB. And every computer graphics programmer familiar with the studio range thinks those TV engineers were out of their minds when they established that standard (I am sure they had their reasons at the time, but I wish we all just switched back to the full computer scale, both because it messes up too many computer graphics algorithms and because it reduces the number of available colors considerably so you need 10 bits in YCrCb to preserve all those beautiful 8-bit RGB colors, but all modern computers use an 8-bit byte).

So, in case my ravings are overwhelming, just a summary here: Always use computer RGB for your timeline.

[Phil Lee]

Hi

I think I understand but...

If you have Studio RGB on the timeline and convert to computer RGB, the assumption is there are no values less than 16 or greater than 235 to preserve, so 0-15 is discarded and 236-255 is discarded and then all the values are remapped, so 16 becomes zero and 235 becomes 255. This causes an inaccuracy though surely, because where do you map value 17 for example in this new scale, it doesn't precisely fit in anywhere.

If you have Studio RGB on the time line it still sits in the range 0-255 with the advantage is you have some latitude now. For example you might have a filter that moves shadow details lower than 16, but another adjustment moves some shadow details back above 16. Now if you've converted to computer RGB of 0-255 so black is now 0 and white 235 becomes white, the first filter drops some detail as it can't move that shadow detail any lower than 0, it's clip it off, so the second filter has no shadow detail to move back into range, i.e. you've lost detail. That is how I understand it which makes me want to keep Studio RGB on the time-line. While it's not ideal not using the full 0-255 range, we might as well get some advantage for it by having more latitude to shift things around without bits falling of the ends?

I see what you mean by your maths examples, but this brings me to another issue in that most HD camcorders will record detail less than 16 and higher than 235. As I understand it the 16-235 is not designed to be a hard boundary anyway, and the higher and lower ends being spaced away from the maximum and minimum are purposely designed to allow a bit of over or undershoot in recording values and they can be made "legal" later. Adding an immediate Studio to Computer RGB conversion drops all that information which might have been of use? For example reducing exposure with 16-235 on the time line might bring in highlight detail that was at "superwhite" level abover 235, I certainly have one example of this where details appear in the sky. If you have an immediate conversion to Computer RGB though haven't you removed the superwhites and anything blacker than black, and reducing exposure wouldn't bring in any extra highlight detail at all where you might have had some? Same applies for shadow detail.

Yes some filters may not work absolutely correctly at the extreme of black and white, but that is only the case perhaps where there is absolutely no picture information below 16 or above 235? Or if a filter isn't working correctly than just on that clip apply a Studio to Computer RGB Conversion immediately before the filter, than after convert it back to Studio RGB, wouldn't that be technically better?

I'm not saying your method isn't correct, there is not necessarily a wrong or right way with this stuff, just trying to understand the rational by having two level conversions which is your suggestion, with the Studio to Computer RGB one being potentially lossy on modern HD footage, that's really where I'm coming from now I've played about and understand what is going on some more.

I found one clip that reducing exposure magically makes some telegraph lines and more cloud detail appear in the sky, when I convert that to Computer RGB then try and reduce exposure, these details do not appear as clearly they are recorded above 235 by my camera which is cut off by the Studio to Computer RGB conversion, which is why I'm not 100% convinced converting everything to Computer RGB on the time line is absolutely the right thing to do, certainly where modern HD footage is concerned, as that extra detail might just be useful and brought back into legal range by some filters or exposure adjustments.

The other pointer is that Vegas by default uses Studio RGB levels on the majority of it's codecs, certainly the modern HD ones, if it was better to convert and always work with Studio RGB converted to Computer RGB, wouldn't that be the default for anything added to the time line?

Regards

Phil

[Adam Stanislav]

Originally Posted by Phil Lee

This causes an inaccuracy though surely, because where do you map value 17 for example in this new scale, it doesn't precisely fit in anywhere.

If you use 32 bits, yes it does (fit, that is). Those are floating point numbers, i.e., real numbers, not integers.

As for some camcorders not sticking to the exact 16 - 235 range, they are supposed to. But if they do not, convert whatever range they have to the proper linear range.

Note (in the enclosed image) how Sony Vegas Color Corrector converts Studio to Computer: It increases the saturation by the factor of 1.164, multiplies everything by 1.164 (gain of 1.164), then subtracts 18.6 (offset of -18.6), since 16 * 1.164 = 18.62. So, just change the gain and the offset (and apparently saturation by the same factor as your gain, I’m not quite sure about that, but that seems to be the algorithm Vegas uses) to whatever your footage requires to stretch your values. And keep it in 32 bits so you do not lose anything to round off errors. And make sure to save it as a preset so you can use it on other footage from the same camcorder.

[Bill Koehler]

Thank you so much for your explanations Mr. Stanislav, especially Post #11 & #17.
I have been going nuts trying to get a handle on what's going on.

[Adam Stanislav]

I’m glad you found it useful. :)

[Gerald Webb]

I'm reading this over and over, and must admit I'm still a bit lost.
So, if I have multiple tracks/cams and some shoot in 0-255 and others 16-235,
( I know this because when dropping the raw files onto a 32bit timeline with video gamma set to 2.2,
some cams have blacks and whites already clipped at both ends of the RGB parade,
and others have no blacks below the 16 line, some whites are higher than 235, but generally the same scene is much more compressed on the scope.)

Up until now, Ive always brought all the ranges down to sit between the 16-235 with a levels and/or color correction filter with the view that doing this as I convert to Cineform will make grading after cutting more uniform and blacks and whites will be less crushed and blown out. Is this right or wrong?

From what I'm reading, I should be making my srgb cams into crgb ( so using input levels to crush blacks and brighten whites) on the time line or when converting to Cineform, and then applying a global crgb to srgb filter on export, is this right?

I am confused because if editing at crgb levels 0-255, when you use Magic bullet or the Cineform First light Looks, it just destroys your footage, anything grey becomes black, and virtually any lighter colors are blown.

Sorry for being the slow kid in the class :) but if someone (looking at you Adam) could explain a real world workflow from cam to DVD or Bluray it may help my simple brain understand.

[Adam Stanislav]

Interesting. I don’t use Magic Bullets, I do all color grading with tools built into Vegas, so I cannot really comment on MB. It is possible that the programmers of Magic Bullets have figured that too many people do not convert to linear computer RGB and are doing the conversion internally within the plug-in. In that case you need to do whatever any particular plug-in expects.

Ultimately, do whatever looks right.

[Phil Lee]

Hi

Originally Posted by Gerald Webb

I'm reading this over and over, and must admit I'm still a bit lost.
So, if I have multiple tracks/cams and some shoot in 0-255 and others 16-235,
( I know this because when dropping the raw files onto a 32bit timeline with video gamma set to 2.2,
some cams have blacks and whites already clipped at both ends of the RGB parade,
and others have no blacks below the 16 line, some whites are higher than 235, but generally the same scene is much more compressed on the scope.)

Up until now, Ive always brought all the ranges down to sit between the 16-235 with a levels and/or color correction filter with the view that doing this as I convert to Cineform will make grading after cutting more uniform and blacks and whites will be less crushed and blown out. Is this right or wrong?

From what I'm reading, I should be making my srgb cams into crgb ( so using input levels to crush blacks and brighten whites) on the time line or when converting to Cineform, and then applying a global crgb to srgb filter on export, is this right?

I am confused because if editing at crgb levels 0-255, when you use Magic bullet or the Cineform First light Looks, it just destroys your footage, anything grey becomes black, and virtually any lighter colors are blown.

Sorry for being the slow kid in the class :) but if someone (looking at you Adam) could explain a real world workflow from cam to DVD or Bluray it may help my simple brain understand.

I think do what works for you, if it looks right then that is all that matters. If something isn't right with levels you will see it, if it all looks good then it most likely is. Just compare the raw file playing in your chosen media player, then put it through your entire workflow (obviously with no colour grading) then compare that with the original. If there is no shift in levels or colour it's all good.

By luck I've had Studio levels on the timeline and when encoding out using the built in codecs they want Study levels so there isn't anything to really do, I should think this is by design. I've followed Adam's advice and tried converting studio to computer, then computer back to studio on output and no matter how much I pixel creep (and a little alt-tab trick you can quickly switch between screen grabs) there is absolutely no difference on the final output to colours and levels. This makes me go for the less is more approach, and rather than two shifts of levels I'd rather do none, it's safer as the tendency might be to forget the shift back to Studio, and for performance I'd rather not stick the level shifts on there while editing anyway.

The arguments Adam puts forward may well indeed be sound, but nothing I've seen with having computer RGB on the timeline then shifting back is any different to keeping it at studio levels on the time line.

If you are mixing studio and computer levels, you need to decide what you want on the timeline, and I would go for keep if the majority of clips are studio levels, or computer if they are computer, this is as per the advice in the article I linked to. As for Cineform according to the http://www.glennchan.info/articles/v...s-9-levels.htm article and codec list, it wants to see Studio levels.

Regards

Phil

[Phil Hamilton]

I am working on an animation project with individually rendered frames at 1280 x 720 from a 3D Modeling software package.

I'm using Vegas Pro 2012 and the timeline I am using is HD 1080-24p (1920x1080, 23.976 fps) with the intended delivery being the internet and also to Bluray using DVDA 6.0.

When I render to WMV the colors, contrast, and brightness look fine. And the way I want it. When I render to MPEG2 - Blu-ray 1920x1080-24p, 25 Mbps video stream for DVDA that will NOT require recompression - the images come out too dark and contrasty. So, based upon the recommendations here, I am using the SONY LEVELS "Computer RGB to Studio RGB" and on the preview it washes out but when I burn to Bluray and play it from my Sony PS4, the levels seem to be almost exactly like what I am seeing on WMV.

My question is: Is there another setting or adjustment I should make in order to make the video on the Bluray look as close to the WMV as possible? What I am noticing is the Bluray looks like the video has some DNR applied - softness and the colors seem a bit less saturated.

Thoughts and suggestions? The image on the RIGHT is an M2V capture and you can see it's a little less sharp to me and seems to be missing a bit of RED in the face.

[Rainer Listing]

Wow. Old thread. Very basically, if it looks right on you and your client's final display, it is right. But it has to be viewed on the (properly calibrated) final display. Normally, original 3D is in full range RGB (computer RGB) and you can leave it that way. View in Vegas, upload .mp4 (or wmv) to YouTube and most browsers will display it that way. If at the end you're seeing crushed blacks and blown out highlights not present on the original footage, any of your display, broadcast, codec, browser, whatever, is clipping computer RGB to studio RGB levels. Add the computer RGB to studio RGB effect to the output - and it then should look right. You can check with your scopes (Wave form and RGB parade) or a simple utility like Instant Eyedropper. To make things look right in the Vegas preview, a really useful utility is the free SeMW Extensions Preview level, which simulates the signal levels of the output device, available from SeMW Extensions

(edit: Have a look at your scopes, the m2v is slightly more saturated, could be compression effect, don't know about the DNR, seems unlikely)

[Phil Hamilton]

I don't have scopes. Doing all of this by "eye".

I'll check out the extensions link you gave me I appreciate the response. So I did. I installed it but I am not seeing anything in the Application Extensions folder which is where it is supposed to show up. I can't find out the name of the extension so I can't search for it. It says it installs but.....


Thank you.

[Rainer Listing]

Hi Phil
I may be a bit confused by which version of Vegas Pro you are using, I assumed version 12, since there is no version 2012. The scopes I was referring to are the inbuilt Vegas Video scopes which you should have if you are using version 12, Menu: View>Window>Video scopes, or Cntrl Alt 2, Google for how to use if you don't know. Don't know about SeMW extensions, they worked for me in version 12 and still are working in version 14.

[Phil Hamilton]

yes thanks I have version 12.0 build 563. I see the SCOPES. I just need to see how they work.

Regarding the SeMW extension - if I knew what the extension name was I could search for it to see WHERE it is installing it. But it is NOT going into the Application Extensions folder as I understand it should be. I could drop it there if I can find out where it's going.