Rendering Question
 

I am doing a very long project in Vegas right now (a business seminar).
It is about 1h 40min long. It has a lot of transitions in it between the speaker and the powerpoint file (which I converted to avi using camstudio).
It is taking a tremedous amount of time to render to MPEG2 format. (left it on overnight and woke up this morning and it says 8 hours estimated time left after roughly 7 hours of rendering last night) Quality settings was set to Good.

Is there anyway I can speed this process up? (pre-render some transitions maybe?)

Any suggestions? or is this normal for a project like this?

Thanks

It really depends on your system specs and the type of effects you used.

The length of your edit isn't abnormaly long. With my old Athlon 900mz system, it would take 12 to 20 + hours to cook. If you have any global effects (effects that you apply over large expanses of the video, blurs, lighting etc. . .) really take a lot of time.

Another time saver is to drop the transitions - remember, the cut is the most dynamic transition of all.

Other than beef up your system, try saving your rendered output file to a disk that's different from the source material. Also, defrag your disks. Also also, sometimes, virus scanning software
will check programs and hog CPU time, so consider turning it off for a bit. Also, check your memory usage to see if you're close to full with Windows Task Monitor.

Another thing you may try that may help speed things up is to render FIRST to DV-AVI, load that new file in a new project, and then render that to MPEG2. Amazingly enough, that combination can actually be faster than a straight to MPEG2 render.

Pre-renders will not help a render to MPEG2.

<<<-- Originally posted by Edward Troxel : Another thing you may try that may help speed things up is to render FIRST to DV-AVI, load that new file in a new project, and then render that to MPEG2. Amazingly enough, that combination can actually be faster than a straight to MPEG2 render. -->>>

Will I lose any quality with this approach?

<<<-- Originally posted by Jeff Toogood : Will I lose any quality with this approach? -->>>

No you will not.

Ok, I tried rendering to DV-AVI and started the process last night before going to bed (11:00pm) and when I woke up this morning (7:00am), it was at 40% complete estimating 11 hours remaining!!

My system specs are as follows:

P4 2.4GHZ 533 FSB
1024 PC2100 DDR Ram
Source HD: 160 GB 7200RPM Maxtor
Target HD: 120 GB 7200RPM Seagate
O/S HD: 80 GB 7200 RPM Western Digital

The Maxtor drive is on a PCI controller card (not sure if that matters) and the Western Digital & Seagate both have DMA enabled.

I don't have any processes running in the background and I double checked that my anti-virus was disabled.

Is this normal? or do you think my computer has a problem somewhere?

The reason I am concerned is because in the past I have rendered 2 hour long clips to MPEG2 in about 6 hours or so (albeit with hardly any transistions). Are the transistion the key to the long rendering process?

Taking that long just to go to DV AVI?! Your footage must be *FULL* of effects and transitions- that's the only way it would normally take that long.
With a 1 hour wedding video (including lots of effects, soft focus, slow motion, etc) it goes right from timeline to MP2 in roughly 2:1 time. In other words a 1 hour program takes 2 hours to render out to MP2.

My Specs:
P4 3.0ghz 800mhz FSB
XP home
1 gig of dual channel PC3200 ram
3 120gig Maxtor Diamond plus9 SATA drives
Asus P4C800E mobo

Usually footage going straight from DV AVI to MP2 without any transitions or effects is realtime if not slightly faster.
Check your track headers make sure there are no accidental opacity changes. Sometimes you can bump the opacity on the track inadvertantly and move it to 99% which is not visible yet causes major renduring throughout the ENTIRE project. Worth a shot- good luck.

Jeff, can you send me your VEG file? I'd like to see if you have something set that shouldn't be. Transitions will affect times some but not to the extent you are seeing.

Edward,

Isn't DV-AVI a compressed format at 25 mbits/sec ? If so, quality would be lost in the additional jpeg-like compression.

Yeah but don't you capture from the camera DV AVI from the start.

Jeff,

You said you converted the PowerPoint to AVI with Cam studio.
What is the resolution of this file? If it is bigger tha 720 x 480, then THAT is what is taking so long to render.

Gary

Yes it was captured at a resolution of 800x600

I don't think I can capture at 720x480

Edward, I will email you my veg file

Yes, DV-AVI is compressed. However, the Sony codec is VERY good. Risk of a very slight quality hit? Maybe. But usually not enough to be seen.

Jeff, I got the file and will look at it probably in the morning. However, you're rendering a Camstudio capture? I missed that part at the beginning.

I used a screen capture program to capture the video for the demo video for Tsunami www.jetdv.com/tsunami. Take a look at that 10 minute video. It took THIRTY HOURS to render. That's the nature of resizing every frame.

In this case I would say that it would probably be faster to render to DV-AVI first and then convert it to MPEG2.

Why aren't these video editors using 3D cards to accelerate such functions? With standard interfaces such as DirectX9, resizing would just flow like nothing using all of those
antialiased texture functions.

Vegas doesn't use any additional hardware acceleration. Resizing every frame takes a LOT of CPU horsepower.

Not to change the subject but.....isn't the file you capture right out of the camera via firewire a DV AVI? How would rendering your footage out to DV AVI after editing cause even *possible* visual degregation?

Also is this correct- the video codec only goes into play when you "change" the footage in one way or another?

*Trying to sort these things out for myself* lol

<<<-- Originally posted by Glen Elliott : Also is this correct- the video codec only goes into play when you "change" the footage in one way or another? -->>>

Yes, that is correct. Unmodified sections will simply be copied.

The degredation might occur if sections need rendering (with
effects or transitions for example).

I would not even want it to use my 3D card for it. Why? Because
a 3D card is build for speed, not quality. Yes, a lot of it looks good,
but that's with COMPUTER GENERATED stuff, not real world
footage! Also a lot of drivers overrule program settings when
the end user changes something in the configuration panel. This
would be a big no-no when doing video rendering through such
a board.

The last issue would be how to get the signal back from the card.
That might prove quite difficult. I do know some 3D rendering
applications are looking at the possability to have the 3D card
aid them in some of the 3D calculations (not actual image
rendering I believe).

Jeff, I can't find the VEG file this morning. Could you send it again? (I swear I saw it last night - might be on the other computer) :-(

Won't be able to send it until I get home tonight form work.

I am really thinking it has to do with the cam studio captured stuff.

I think I am going to just split it into 10-15min segments and render those to DV-AVI and then piece about 10 clips together and render to MPEG2.

Question: If I go this route, the render to MPEG2 shouldn't take extra long becuase as far as it's concerned it is just converting DV-AVI to MPEG2? Right? All the complicated rendering was done when rendered to DV-AVI??

I think....

The render to MPEG2 should be pretty fast. I think you will find the MPEG2 render takes MUCH less time than the original render(s). Coming from the odd sized source takes a tremendous time to render - just the nature of the beast.

Based on this info, I'm sure the reason for the slowness is the source file. Standard DV-AVI sources should fly! If you want to resend the file, that's fine. However, I'm sure it's the source causing your slow speed.

>I would not even want it to use my 3D card for it. Why? Because
>a 3D card is build for speed, not quality. Yes, a lot of it looks good,
>but that's with COMPUTER GENERATED stuff, not real world
>footage!

Rob,

Because texture processing is big in games, the 3D cards have accelerator chips for many image processing functions used for
video. They can resize/resample images better than Photoshop,
all in real time. Starting with GeForce2, nVidia included various deinterlacing algorithms and others were included to accelerate DVD playback functions, specifically MPEG2 decode. Imagine
Vegas offering better deinterlacing than Blend or Interpolate,
all with no additional load on your CPU.

Future processors will accelerate MPEG2 encode for use with the Personal Video Recorder software. The last few rounds of 3D cards have programmable "shaders", which would also allow any savvy image processing coder to write the BEST image resizer. Common video funtions such as film grain generation, color correction and what could be processed by these 3D cards.

To give you an idea, the GeForce4 chop is actually bigger than Pentium4. The higher end cards sit on 550 MHz DDR memory with
a 256 bit data path. Video enthusiasts should be asking for
NLE vendors to use these functions.

I know what these cards and chips can do. But I doubt they'll
have higher quality algorithms then either Photoshop or a good
effects/NLE package. Why? Because they are all about speed.
Lets just say that if they can get better quality in realtime then
why are we waiting on rendering such effects on 2 - 3 GHZ
machines which are way faster then those chips? Yes I know they
have the algorithms in hardware, but I don't believe they are
that much faster and have a better quality.

Even if they did. There is one other problem. How are you going
to get the results of those back and not on the screen? If I'm
not mistaken it will only process what is up on the screen. Which
might introduce problems since there is also other stuff on the
screen (like the user interface).

Then my point about the user settings also still stands. The users
can tell the drivers what quality to use (often) and for example
always do an anti-alias pass. You don't want the user to have
control over such settings, especially out of the application that
is rendering.

Now there might be an interface that I don't know about (please
point me to such information on the web) or such a system might
be coming in the future. But I doubt it would be feasible today.
a thing might be coming in the future

Well, I decided to just let the computer go and leave it alone and it rendered the full 1hr 40min clip straight to MPEG2 in about 22 hours.

I should have just let it go in the first place :)

Anyways guys, thanks for all your help and suggestions

In reference to utilizing video cards for rendering. . . too many potential problems between the computer, video, programs, card. I spent the better part of three years trying to get card dependent systems to produce acceptable video output. I finaly discovered 1) Canopus for my analog stuff. 2) Vegas for my NLE. All of my problems went away. (except for the ones I create)

Rob, the last two generations of nVidia cards have programmable
pixel shaders. These cards are have generic image DSP abilities. So, what you were asking about is already here. As a plug-in, any software developer can insert any kind of image processing algorithm. When they are done, the frame buffer can be read by the video application. These graphics cards will easily exceed the performance of the latest Intel or AMD CPUs. As a plus, these 3D cards are coprocessors that free up the CPU for other tasks, such as disk I/O and data compression.

hmm... i think for converting powerpoint to video the http://www.geovid.com/Presentation_to_Video_Converter/ is better than camtasia...