32-bit floating point questions
 

Never seen the forum so quiet! I assume everyone is busy playing with their new v8!

First impressions are generally positive. My main reason for upgrading was the ProType Titler and while I wasn't exactly overwhelmed with excitement, it certainly is a great improvement on the old text generator. Shame that the credit roll generator didn't get a re-work as well.

1. Anyway, to my questions:

Under what circumstances will I benefit from 32-bit floating point video processing over 8-bit?

I am working exclusively with SD at this time and I have the PC horsepower to to do more intensive processing (2.66 GHz quadcore) but will I see an overall improvement in quality?

2. I understand that certain plug-ins and media generators do not support floating point processing - does that mean they will switch automatically to 8-bit or will they just not work in a 32-bit project?

3. Does 32-bit processing require Vista or will it work under XP SP2?

$. Finally, under what circuamstances might it NOT advised to use 32-bit processing.

Thanks in advance, as always, for any advice.

Ian . . .

>>>>>Shame that the credit roll generator didn't get a re-work as well.

What the????

I will probably still upgrade, but thinking the improved titler included the credit roll generator was going to be a significant part of the justification.

I guess its back to the old keyframe scroll.

Brian

I don't understand why so many people are complaining about the credit roll generator. Who actually uses it anyway? I've always just used the regular text media and added the scrolling via keyframes! You can do the same with the new proTitler - just add movement via keyframes.

"Who uses it anyway"? Probably very few people because it is so poor. It certainly is NOT the 'professional video effect' it's described as being in the Sony blurb for v8.

In my opinion, the text generator or ProType Titler are not acceptable substitutes.

I want to be able to add logos or other graphics in my credits, have three or four columns of text, connect dual items with characters of my choosing and in different colours, create reusable customised style presets for elements, fade text in and out at the top and bottom of the screen (without using a separate mask). I want to be able to roll the credits based on a selection of typical presets (forward, backward, left to right etc etc) without having to keyframe. Basically, I want a decent purpose built credit roll generator!

Neither the text generator or the ProType Titler help me create the professional credit rolls I want which is why I will continue to (reluctantly) create my credit rolls in Photoshop.

So, 32-bit floating point processing. What's all that about then? ;-)

I don't have Vegas 8 installed yet but work extensively in 32-bit color space.

1. 32-bit will vastly improve any render/processing where banding/quantization/stepping is created by 8-bit processing. If 8-bit processing does not produce artifacts, 32-bit will offer no improvement of itself. It's also garbage-in, garbage-out. It won't magically fix bad footage but will prevent new artifacts from 8-bit processing being introduced.

2. AFAIK, they are flagged and will work (assuming they work with Vegas 8) but process at 8-bit with any resulting artifacting.

3. It will work under any OS the application supports. 32-bit is application internal. Note that to proof correctly in 32-bit since most display devices are 8-bit requires some tweaking and I'm not sure how this is addressed yet in Vegas. Also, outputting back to lower bit codecs can re-introduce artifacts and testing/experience will be needed to see what works best.

4. When there is no benefit. It's going to be slower for all operations and more tricky to proof. A cuts only project shot 8-bit with no CC, FX etc rendered out 8-bit will probably see no visible improvement. But a project with mixed 8-bit footage, some 8-bit plugs and but some native 32-bit processing will see benefit though it will be slower. Time to upgrade to quad cores :)

Very informative, Stephen, thanks.

OK, already got the quad core, but if there is unlikely to be much visible improvement (which I suspect will be so in my case, based on what you've said) I really don't want to be going back to pre-quad core performance!

Thanks again.

Ian . . .

Please see my other response at:

http://www.dvinfo.net/conf/showpost....1&postcount=37

There's a lot of confusing about this 8-bit vs 32-bit stuff. The 8-bit simply refers to the bit depth of your video format (e.g., SD DV). The 32-bit floating point refers to how the software does calculations. No NLE worth its salt has ever done true 8-bit calculations.

Behind the marketing hype, all Sony have done as far as this is concerned is to make use of features available on more recent CPUs. It is OS-independent (i.e., you don't need Vista etc). For older CPUs, the calculations are performed using either 16- or 32-bit integers. This is still more than adequate for 8-bit video.