Shaolin Productions

This guide being about everything you do before you compress. This will help you get the best out of your game and your footage, also a section for Quake3 and for Quake 3 engine games. This applies to ALL movies, Q3 engine is only the best to make movies with, and its what my experience is in.

Many people ask me how I get quality, what settings I use, etc. once a trade secret I think it would benefit me, and many people starting out, to just publish everything I know about quality. I will try and be as in depth as my keyboard permits and add links for extra details. After you get perfect quality, you can check out my compression guide.

Read this through once, then actually try some of the things, then after that if you have questions I will be glad to answer all I can.

Before you even capture a single screen shot.

[capture is a just a term for getting the screenshots, how ever you do it being UT2k4's framedump, q4's avidemo or avigame or whatever ...but my example is quake3 based games with cl_avidemo ]

Note: It helps to have 2 installs of the game (if you also play the game) to make converting from play config to movie config unneeded. And if desired editing of game files (not allowed on most servers) also by installing somewhere new, this will change the capture directory for the screen shots.

Right from the beginning -> Go to the NVIDIA or ATI driver pages. Get the latest update. Then go into driver settings and look for quality, uncheck anything ‘application controlled’ and push all sliders for anisotropic filtering, anti-aliasing, supersampling, filtering to the highest settings. My card goes to 8x/16x/transparent/triple respectively. And anything that says, “performance vs image quality”, go towards quality of course.

Also enable transparent AA and supersampling if you have the options.

I can't comment on ATI configurations cause I have never had a ATI card but for both chip sets there are driver tweaking utilities that a simple www.google.com search can bring up in seconds.

Anti-aliasing: most import and noticeable: makes diagonal lines non-jaggy or smoothed out. Adds pixel data to a image. www.100fps.com

anisotropic filtering: used to make textures more crisp, in particular far away textures.

Conformant texture clamp: Off (for q3 sky boxes)

Quake 3 stuff (skip this if your game is not quake 3)

High quality game config.

Load GQ3.cfg http://www.own-age.com/forum/thread.aspx?id=21193&page=20

Some settings of note (I have so many I have no idea if some are even needed or not) CFG should work decently for Ra3, OSP, defrag(cg_draw2d 0), CPMa, and baseq3.

R_mode = sets resolution, set at 9 for 1600x1200. or

Seta r_mode “–1”

seta r_customaspect "1"

seta r_customwidth "1600"

seta r_customheight "1200"

r_gamma = set at desired level, but a slightly dark setting allows better color control later [niklz], because it saves the color info, if you had a slightly brighter image you lose color data with “washing out”.

cg_fov "107" [100 through 110 recommended] and cg_shadows "1" [1 - 3] ….your choice. High quality shadows need r_stencilbits of 8 or 16, I forget.

Capturing: (all q3 engine games, such as ET, RTCW, etc)

Many ways have been invented to get great quality and motion:

(1.) High final frames per second (fps) (SQF, Lyon connection),

(2.) High capture FPS that is resampled (live to fly, GQ3),

(3.) Combination. (many movies)

(4.) medium or low fps(100) + a non resampled version blended together (iT2, RAN).

Edit: Check out the mod Q3mme, it has in game frameblending with out a limit, you can capture 1600fps and save it as 100fps if you want, come out very nicely also, see the tutorials on the tool for use.

http://q3mme.proboards75.com/index.cgi

Also combining frames improves the color of the images, not just the smoothness of the video.

1.) High final FPS (no resampling) suffers from 2 things, one high fps is harder to store which makes file sizes go way up, and it is over all harder to compress. Also the playback is more CPU intense, because it has to go through 40-50 fps instead of 25 a second, for example. With the added CPU requirements of x264, you need a even newer PC to play the final file back.

Positive: It does create a closer to in game look, you play at a high fps, not a 'motion blurred' one, that’s just how game graphics work, your eye creates the blur. (higher the better.. up to around 200+ fps, there is no 30fps see-able limit, it's different with movies, don’t believe the uneducated)

2.) Super high fps (400, 500, 1000) main draw backs are it takes 3 or 5 times longer to capture the clips themselves, which is real time used that can be used for the editing of the movie. The resampling method from normal game footage to create motion blur is flawed. It does not create a real/clean motion blur image. Also some people ever say that the high motion blur is to much (although I disagree)

Positive: The speeding up of clips, with resampling, (or time remapping), makes for a very smooth picture. Real (Hollywood) movies use the motion blur principle; so they only need 24 frames to create smooth motion. [With apps like gaz’s tga_hook you can create those blurred frames, but that’s not covered here.]

3.) Combo: Also you can do a combination of those methods (resampling + a high final FPS, like q3fuel2) That being a new and supposedly superior final quality and look, it also combines the problems of the above methods. More storage space needed on your hard drive, more time capturing frames, higher final file size, and higher decoding requirements. If you have a very powerful computer, and lots of free time (or 2 computers even) I suggest giving this a try, it does look smooth.

4.) Multi-layer: (explained with Vegas video) Capture 100fps and bring that footage into Vegas and edit as normal, this would be a 'light' or 'weak' resampling. Now when you render your movie, turn off resampling either on the clips(right click, switches) or export the movie as “preview” instead of “best” and you will get a non-resampled version, Import that which you just rendered, and place that on the time line, in sync with the rest, above your other footage. (of course you can do 2 renders, one “best” one “preview”, then make a new project file to speed rendering, yet you might not gain that much time in the end.) Now the top layer, you lower its opacity(40% maybe) then change the transfer mode / compositing mode. Modes such as “overlay”, “add” and “lighten” look nice. Play with different opacities and modes.

My method is a bit different, designed (with help by many) and tested over many movies. In short: Capture at 200 FPS, save in virtual dub as 100 FPS avis then import into a Vegas project with a final FPS of 25. I suggest 25fps as the final FPS for the movie/project. It’s a PAL standard and it takes less file size, decoding requirements then 30fps, but still looks pretty smooth.

Notes on resampling:

Resampling is best done in multiples. Everything works better in multiples or even divisions. Like 25fps x 4 = 100fps, or 100fps resampled evenly 4 times is 25. Making your capture an even number allows the frames all to get the same amount of resampling, and the same amount of frame data, when combining.(This possibly being just a theoretical warring :D )

Vegas Video, After Effects, combustion, Premiere and most major NLEs should have a combining of frames opinion or setting. For Vegas it is ON by default, AE it is as simple as a button. Some useful vocabulary follows.

Resampling, Frame blending: combining of frames.

From Vegas' Help index:

Select a radio button to determine how video frames will be resampled when the frame rate of a media file is lower than the project’s frame rate. This can occur either when the event has a velocity envelope or when the frame rate of the original media is different than the Frame rate setting on the Video tab of the Project Properties dialog.

With resampling, the intervening frames are interpolated from the source frames, much like a crossfade effect between the original frames. This may solve some interlacing problems and other jittery output problems.

This also works for higher to lower frame rates of course, I'm sure Vegas dev team doesn't know about game vids :D

Supersampling: another method of increasing image quality. Supersampling takes surrounding pixel data, and depending on the algorithm, and creates a better image.

In Vegas this can be used on frames and video. Video supersampling can improve the appearance of computer-generated animation by calculating intermediate frames between the project's frame rate, allowing you to create smoother motion blurring or motion from sources such as track motion, event pan/crop, transitions, or keyframable effects.

Motion blur: takes surrounding frame data and combines them. Its similar to Vegas' supersampling. In AE, motion blur is used on text and motion graphics to smooth edges.

Now you captured a frag/trick at 200fps and the screenshots are 1600x1200.

Note: that you may not need to capture that high, 1024x786 with a newer card looks just fine. IT2 was captured at this, only higher for zooming or panning shots. Resizing in general is used to get rid of 'jaggies' [alias] from an image, same(debatable) as turning up the anti-aliasing on your video card, so if your system supports 8x, you may not need the higher res. With capturing high and resizing down there is an increase in the color depth and 3D look to the image. I still like 1600x1200.

Images to Video

There are many ways to convert your footage, but I believe the best process is using virtual dub and resizing to final res (whatever you choose, 640x480 is good) with the lanczos3 resize method. [or http://www.avisynth.org/Resize ] There is really no need to use Photoshop and pjbmp2avi (that was my old method also). Virtual dub can resize to any res with better quality, works faster, and you can convert to avi in one nice batch process. PS uses bicubic resize method which works best for resizes to half res [bicubic resize method should only be resized in: ½, ¼, 2x, etc..]

In Virtualdub

1. File / Open video file.. and find your images, double click on the first screen shot and the rest will be loaded in. It will import till there is a major change in the screen shots, such as a name change or resolution difference.

2. To create a script in virtual dub with a resize filter [Video/ Filters, Add / resize filter] In the boxes, select lanczos3 and type in your res.

3. Then go to Video / Compression set as Huffyuv [a ‘lossless’ codec, halves needed storage space, and speeds up reads and writes of the file, but keeps the original quality (unlike xvid or other ‘lossy’ codecs)].

4. In Video / Frame rate, save the clip as 100fps. Source rate adjustment-> Change to: “100” [default is 30 I think] Vdub has a limit of 199 fps, if it was 200fps we could just import the clips into Vegas and they would be auto resampled to 'normal speed'.

5. Then File / Save processing settings -> give it a name. Whenever you load up Virtual dub for converting screen shots, this allows you to just load that .vcf settings file and load up your script whenever you want (works great for compression settings also) You can of course set cropping, compression and tons of other filters on your footage.

Capture a shot, then move those screen shots into a folder [or use Q3MME with automatic screen shot and folder naming, q4 also has this feature], then capture another, etc, till you do a few or fill up your Hard Drive partition like I do. Then do the virtual dub steps, and do File / Save Avi as for each clip, but check the “add to job control” box to queue it for batch processing, so you can convert all your screen shots with the push of a button. Then to start the batch: File / Job control then start.

Editing

Now whatever program you use to edit there are some rules to follow.

Always work with square pixels and never interlace your footage. NTSC and PAL, the standards for TV/ DVDs around the world are displayed as non-square and interlaced(frames are displayed in fields or half frames. For Computer Monitor playback you never want these things, PC displays are progressive(full frame) and square pixels. Go through all your programs and when you are setting up projects, that everything is progressive [or non-interlaced, no interlacing, or field order: progressive ] and that the “pixel aspect ratio” is “1.0”. [the “frame aspect ratio” can be different.] Just make sure the pixels are square.

Also do not compress things unless it’s your FINAL movie file. Always work with a lossless (no quality lost) codec such as huffyuv / mjpeg, or export uncompressed. This also goes for audio, if you got it from a .mp3, convert it to .wav, then only work with .wav till the FINAL compress. With that, make sure you grab the highest quality version of the audio tracks that you can find (rip from CD if possible).

Now you do all your editing and things, making sure all your clips are treated as progressive and square, you then render out uncompressed from your editing program.

If you render “uncompressed” -> should have a perfect (quality wise) .avi.

Now actually go test this, never hurts to do some test clips, because my word is not law, nor always truth. :D

Good loss-less codec for storing clips -> Huffyuv - [skip the install of all the other codecs if you don't know what your doing, I only suggest this installer, and only install the codecs you need. More codecs on a system the more conflict you have]

http://prdownloads.sourceforge.net/gordianknot/Gordian.Knot.Codec.Pack.1.9.Setup.exe?download 

Credits go to: Vegas help contents, jrb, mrks, own-age, robo-kill, niklz, joka, gaz, hooker, tck, doom9.org, max and own-age forms.

My first teachers:

http://www.own-age.com/forum/thread.aspx?id=3982&page=1

http://www.own-age.com/forum/thread.aspx?id=5883&page=1

Probably a ton of others I cant think of at the moment, thanks.

3-3-06

updated few links and added q3mme