Learn More Contact Now

Alex: August 2008 Archives

STP-Spectrum-View.pngAs a quick aside from my audio sermons, I'd like to take a quick look at something a bit more practical.  In Soundtrack Pro, we're all used to seeing waveforms when we open up clips.  But there's another perspective on the sound that, while it's a little more scary at first, can be almost as useful once you figure it out.  You can separate some sounds from others, spot areas with funny spikes that don't show up in the waveform, and even copy and paste specific sounds from within a flat clip.  With practice, you can even begin to recognize the "fingerprints" of individual words.  They call this view ... Spectrum View.
During Editing
You've laid down your sequence, you've got your clips sounding consistent relative to each other - now you should check for peaks.  In Final Cut, this is as simple as picking Mark Menu -> Audio Peaks -> Mark.  Markers will appear above the sequence anywhere there's a peak.  Now, you could just pot down the gain on those clips, but if you do that you'll lose the consistency between your clips' audio levels.  So instead, you can hack out the individual peaks themselves.

You can do it by hand, or you can do it with a single step in Soundtrack Pro.  Read on ...

B2B-Audio-PeakMeter.png During Shooting

The first step comes while you're producing your footage in the first place: if you happen to be a one-man show and you're doing your own taping, be absolutely sure to check your audio levels on your camera as you're shooting.  Most cameras can be set up to show some kind of audio meters on their display; as a last resort, though, almost all will let your plug in headphones and listen for gross distortion.

Ever seen your audio meters go into the red?  Marked those unsightly peaks in your audio?  Heard the raging distortion that happens when somebody shouts into a mic?


Oftentimes, despite your best efforts in production, you'll find yourself editing together clips shot with audio that sounds different.  Most importantly for our purposes today, some clips' audio might be louder or quieter than other clips'.  One way to bring every clip into the same volume range is called normalization - and, while audiophiles have good reason to turn up their noses, it's probably the best way we have to fix this kind of editing problem.

What does normalization do?  Read on now.  How do you actually do it?  Tune in again tomorrow for the cheat sheet ...


Just as a computer thinks of video as a set of images, it thinks as audio as a whole lot of "snapshots" of this sound wave - tens of thousands of snapshots every second.  Each of those tiny little snapshots reflects the energy of the sound wave at that point.  Now, the human ear is capable of hearing frequencies that repeat up to about 20,000 times per second (20,000 Hertz).  Because of some fancy math, we have a rule of thumb: if the computer taps in to the sound wave twice as many times as the highest frequency we want to represent, it will describe the sound fully for us.  That's where the most common full-quality sample rates - 44,100Hz and 48,000Hz - come from.

I'm going to shift gears into discussing ways we can enhance the audio on our videos.  In order to do this, we should take a whirlwind tour of how computers think about audio.

Application: Fonts and Vector Images
Antialiasing also has something to do with why fonts and vector images always look good, even if they're really big or really small.  Actually, fonts are vector images, so that's a bit of a strange thing to say - but at any rate, vector images are simply a set of mathematical equations that describe the curves that make up a shape.  So, like the real world, the computer can "look at" these images with infinite precision.  Therefore, it knows enough about the shapes to behave like our eyes do, and approximate the way we would see the fonts and such if they didn't have to be on a quilt of pixels.

Application: Photoshop
One place where this knowledge might help you think is in Photoshop, especially if you're designing for compositing software or DVD production.  Say you're doing a glass bug for your video: did you notice the anti-alias checkbox when you used the Magic Wand tool?  That checkbox means that the computer will actually select fractions of pixels on the edge of the selection it shows - "blurring" the selection boundary in the same way that it did to the red channel in the example reproduced below.

We know antialiasing best as "smoothing" the jagged edges that any digital editor will eventually happen across.  Why do we need to antialias, and where do aliasing artifacts come from in the first place?  The answer is a little mathy, but we'll bring it down to size.


Masks vs. Alphas

Alpha channels are our most sophisticated way to handle transparency, but they're by no means the only way.  We commonly use the term "Mask" to refer to something similar to the alpha channel above - that is to say, a full range of "see-through-ness" for each point.  Photoshop, for example, uses the term "Mask" to mean just that.  But more primitive incarnations of "image masks" in applications like DVDs and some still image files like GIFs use a different approach.

Alpha Channels

As we first start thinking about transparency in a computer sense, let's go back to the beginning of how computers think about our NTSC broadcast images.  You might remember that, for our purposes, a computer thinks about videos as a sequence of many individual images (frames).  It thinks about each frame as a big rectangular "quilt" of pixels, or individual dots of color - for our purposes, 720 dots wide and 480 dots tall.  And it thinks of each of those dots of color as the amount of red, green, and blue in the color.

Whether you're using titles, creating supers, or doing any other sort of compositing, you're trying to tell the computer to show some parts of an image but to hide other parts.  

I've been surprised by the number of folks who were never really taught how that process works from the computer's point of view.  After the jump, an easy little primer that may help you understand --


As you know, fonts normally come in a single color. Essentially, each letter is a shape - and the computer needs to describe your logo as a shape as well. Therefore, you need to reduce your logo into one or more black-and-white images that correspond to the regions you want the computer to fill when you're using your font.

Incidentally, the very basic techniques we'll use in Photoshop might help you with other matte and compositing work as well.

Installing the free software you need to create your fonts can be a little bit tricky.  Read on to find out how to set up the font software and the auto-tracing plugin on Mac or Windows.


Most modern titling software allows you to import logos or graphics to incorporate with your text.  Avid's Marquee Title Tool does not, at least in any sophisticated way - which is a shame.  Wouldn't it be nice to be able to do the easy 3D manipulations to logos just like you can to letters?

That's what motivated this little tutorial series, but, depending on your workflow, the implications may be even broader.  If you have a set of many logos or shapes that you use often, for example, you can put all of them just a keystroke away.  You can use them as an easy way to make mattes, as some of our other tutorials describe.  For that matter, you could create characters (glyphs) for things like your signature.

Also, the kinds of fonts that we'll be using are vector-based - meaning that they're a set of equations that completely describe each character.  In English, that means that you can zoom the characters up to the size of a skyscraper without worrying about the ugly, fuzzy, pixelated edges that a normal image would show.


* - If you're just doing one-off work with logos (or anything else for that matter), the process we're about to go into is probably overkill.  Interested in a tutorial on vectorizing these more "expendable" images?  Drop us a line, and we'll get one in the pipeline ...

This time, we'll write a script to detect whether the DVD player is showing in widescreen (16:9 anamorphic) mode or standard (4:3) mode - and automatically play the video in the appropriate format.

I hope that this tutorial will be a practical one for those of you who are putting both regular NTSC-DV content and anamorphic on the same disc, but more importantly I hope that the introduction to bit math might be useful if you need to extract other characteristics from the bitwise SPRMs, like SPRM14 (Video Config), SPRM15 (Audio Config), and SPRM11 (Karaoke).

You'll have to excuse me for getting all technical up in here; "bitwise" just means that the single SPRM stores a lot of pieces of information in one number. Read on for what I promise will be a painless introduction ...

Let's say your title is a bit ... racy. And maybe, because you're a good citizen, you don't want it to play on sets where parents have enabled parental control.

Using DVD Script, the solution is surprisingly simple. First, create a script - call it, say, "Verify Parental Controls" ...

Finally, a practical example! And we won't even need all the SPRMs or most of the GPRMs from last time as we create a video that plays its tracks in random order.

In fact, we'll only use a single GPRM - those are the variables that are ours-all-ours, remember? - plus a jump or two, a little math, and the conveniently-named Set GPRM Random command. Let's say you have four tracks of video, and you want them to cycle randomly ad infinitum - maybe you're at an exhibition or something, who knows.

Ever think that those old algebra classes would come in handy producing video? Well, I'll ignore the fact that algebra is a major part of almost all of the editing process, and pretend that I'm BLOWING YOUR MIND.
Receive FREE Tutorials by email:


    Avid Media Composer Training
  • Enrollment Cost: $50.00
  • 84 Media Composer Lectures
  • Includes Practice Media
  • Interactive Quizzes
  • Official Certificate of Completion
  • 30 Day Money Back Guarantee
  • Click to Enroll for 10% off!
    Final cut Pro X Training
  • Enrollment Cost: $20.00
  • 60 Final Cut Pro X Lectures
  • Includes Practice Media
  • Interactive Quizzes
  • Official Certificate of Completion
  • 30 Day Money Back Guarantee
  • Click to Enroll


This page is a archive of recent entries written by Alex in August 2008.

Alex: July 2008 is the previous archive.

Alex: September 2008 is the next archive.

Find recent content on the main index or look in the archives to find all content.