Improving the Look of Graphics and Text

[ This article was first published in the May, 2006, issue of Larry’s Monthly Final Cut Studio Newsletter and  SIGNIFICANTLY updated July & November, 2009 and a minor update Feb. 2013. ]

[ All references to Final Cut Pro are to version 7 or earlier. These image sizes are still correct for Final Cut Pro X (and Premiere Pro, for that matter), but the screens and operational instructions are for FCP 7. ]

Video is really, really good at showing motion and emotion. It’s not so good at displaying text.

One of the discouraging facts of life is that we create all these great graphics on our computers only to see them destroyed when they get transfered to video. (Well, OK, maybe not destroyed, but really, really different.)

Is there anything we can do to change this situation? Well, um, no. But, there are things you can do to improve the look of your graphics and text when they are displayed on video. And that’s what this technique is about.

First, I’ll explain how video and computers are different, then wrap up with a series of specific suggestions you can use to improve the look of your text.

In the beginning

Video was invented about forty years before computers; which means that computers were able to improve on the lessons we learned during the development of video. Because of this, there are six major differences between graphics on video and graphics on the computer. Keeping these in mind will help you improve the look of your text.

These seven differences are:

• Video is fixed resolution, computer resolution varies
• Video displays interlaced images, computers display progressive images
• Video gamma is different from computer gamma
• Video white is grayer than computer white
• Video uses YCrCb colors, which are more restrictive than the computer’s RGB colors
• Video colors are not as precise as computer colors
• Video pixels are rectangular, the computer’s are square

Let’s take a look at each of these and see what problems they cause.

Fixed Resolution

Regardless of the size of our video monitor, all SD video is 720 x 480 pixels (720 x 486 for you broadcast types and 720 x 576 for PAL). If you do the math, this works out to about 0.3 megapixels — a far cry from even a cheap digital still camera’s 4, 6 or 8 megapixels.

This limited number of pixels means that your image quality varies by screen size. The bigger the screen, the grainier the image. Also, because there are so few pixels, lines that are nearly horizontal or vertical will show serious “stair-stepping,” in a way that the same line on the computer will not.

This means that, graphically, we need to avoid using very fine detail or lots of thin or swirly lines. Video just doesn’t have enough pixels to draw the image accurately.

Interlacing

NTSC, PAL, and some HD formats are interlaced. This means that while we see many complete images each second, each image, or field, is not complete in and of itself. Instead, the video monitor displays the image in two parts — first all the odd lines of the image, then all the even lines.

This means that in the US, we are not seeing 30 images per second, but, instead, are seeing 60 half-images per second. These “half-images” are called “fields.” (PAL displays 25 images per second, composed of 50 half-images, or fields.) Essentially, interlacing means that the vertical resolution of our image is instantly cut in half, because we are only seeing 1/2 the image at any given instant.

The result of using fields is that we have even less image data to work with. Thin lines flicker wildly because they are in one field, and not the other. Curved lines look even more stair-stepped. A moving object has weird horizontal lines radiating out from each edge that you can see on the computer, but not on the video monitor.

FCP suppresses displaying interlace lines UNLESS you show your image at 100% size in either the Canvas or Viewer. This picture illustrates what interlacing looks like; the red arrow points to the horizontal interlace lines.

Video gamma is different from computer gamma

This was illustrated in a message from Joe Sphinx:

Hey Larry, this sounds like a stupid problem, but I’m having an FCP 5 problem. I captured hdcam material via decklink card in FCP 5 at dv resolution for offline. then wanted to export stills from the timeline as uncompressed tiffs with quicktime conversion. when having done this, pics come out darker than what is represented on timeline and external monitor. Any ideas around this?

Larry replies: Yup. The computer does not display the same color levels as video. Each use a different gamma (or mid-tone gray) setting. The standard gamma for the Mac is 1.8. The standard gamma for video is 2.2, which is darker than 1.8. You can compensate by load your stills into PhotoShop and setting the mid-tone gray setting in Levels to 1.2. However, don’t change the gamma setting on stills you want to reimport into your FCP project, or they won’t match your existing footage.

TV White is Actually Gray

Video was invented in an analog world — computers are digital. Consequently, they don’t use the same black and white points.

Digital black is displayed at 0% on FCP’s Waveform monitor. Digital white is displayed at 109%. (Here’s a good article on how to read scopes.)

However, digital black to white is too great a range for analog video, which include broadcast and DVD duplication. When you are creating graphics or text for video, set your black level to zero on the Waveform monitor, but keep your white level to 92% (235 in PhotoShop’s color picker).

This is called keeping your white levels “broadcast safe.” You can clamp your white levels to the correct value by applying the Broadcast Safe filter (Effects > Video Filters > Color Correction > Broadcast Safe).

You can tell if your white levels are too hot by selecting either the Canvas or the Viewer and choosing View > Range Check > Excess Luma. If you see a yellow warning triangle, your white levels are too hot. A green check-mark means they are OK.

You can see this, as well, by looking at your video on the Waveform monitor. If any white pixels are above the 100% White line, your white levels are too hot.

Excess white levels will cause white text to shimmer, tear, or break-up. It can also cause a buzz in the audio. It also means that your video will be rejected for broadcast, as well as by most cable outlets and duplication facilities. You’ll need to reduce your white levels and re-output.

Shooting, editing, and outputting on DV — since that is all digital — does not require clamping, or restricting, your white levels to 100%. However, if you then compress your file for DVD, you’ll find that the process of compression converts your images into a broadcast safe format, which may, or may not, give you the results you expected.

Video can’t display as many colors as your computer

Computers use an RGB color space, digital video uses YCrCb (analog video uses YUV). Both video color spaces are more restrictive than the computer. In other words, you can easily create colors on the computer that can’t be displayed in video. Saturated yellows and blues come instantly to mind.

You can tell if your chroma levels are too hot by selecting either the Canvas or the Viewer and choosing View > Range Check > Excess Chroma. If you see a yellow warning triangle, your white levels are too hot. A green check-mark means they are OK. (These symbols look remarkably similar to the white level symbols I just discussed — that’s because they are.)

You can see this, as well, by looking at your video on the Vectorscope. If you connect the tops of the six targets, representing the six primary and secondary colors, if any white pixels exceed the boundaries of that rectangle, your chroma levels are too hot.

Excess chroma levels, like excess white levels, will cause colors to shimmer, tear, or break-up. It can also cause a buzz in the audio. It also means that your video will be rejected for broadcast, as well as by most cable outlets and duplication facilities. You’ll need to reduce your white levels and re-output.

Video colors are not as precise as computer colors

Through a process called “color-sampling,” the colors of adjacent video pixels are averaged to reduce file size. The problem with color sampling is that the quality of your image is often degraded. Worse, it gets harder to pull a clean chroma-key, or to do good color correction.

Here are two articles on my website that describe this problem in more detail:

• DV, Color Space and Chroma-keying, by Larry Jordan
• What is Color Sampling, by Graeme Nattress

The key issue with color sampling is that we don’t have the same precision with color in video that we do with the computer. This is, generally, not an issue when we are shooting pictures, but becomes a significant concern when doing effects compositing, color correction, chroma-key, or adding text.

UPDATE – July, 2009

Recently, in working on my upcoming book for Focal Press, I discovered a significant mistake in one of my video tutorials – #12 – Working with Stills. (I immediately updated the webinar so the current version online is correct.)

This came about because, as part of writing this book, I took a detailed look at how to prepare still images in Photoshop that look great in Final Cut Pro. This is a surprisingly difficult question to answer and, while I got a part of it right, I also got a part of it wrong.

I’ve written about this before, but now I can explain this better.

As has been discussed many times, computers use square pixels to describe their images, while video uses rectangles. So, for instance, while the video frame might have a 4:3 shape, if each pixel inside it is tall and thin, you’ll have more pixels horizontally inside it than if each pixel were short and fat. Or, said another way, if every pixel were big, you’d need fewer of them than if every pixel was tiny.

The overall shape of the frame doesn’t change. However, the number of pixels needed to fill that frame changes depending upon the size and shape of the pixel.

TWO TYPES OF STILL IMAGES

I’m indebted to Tom Wolsky for some additional insight on this issue. There are two types of still images that we can create for use in video: images without transparency and images with transparency.

Images without transparency are, generally, digital photos or scans of photographs. In these images, the entire frame is filled with pixels and there are no transparent areas. They are always a single layer, and most often saved as a TIFF or PNG. (Tom prefers PNGs, I prefer TIFFs. Both are high quality, uncompressed images.)

Images with transparency include single layer logos, where areas around the logo are transparent, as well as multi-layer graphical cornucopias of imaging wizardry. These are almost always saved as PSD files.

The reason for this distinction is that Final Cut treats these two groups of images differently. Which means we need to size these images differently, in order for them to look correct. If we don’t, when we import an image, it gets squished, with circles becoming eggs and squares becoming rectangles — not something you want to see happen with a sponsor’s logo, for instance.

This is the classic illustration of what happens when you don’t compensate for the differences in pixel shape (also called “aspect ratio”). The black circle was drawn in Photoshop and imported into Final Cut. The red circle was created in Final Cut. They should match, but they don’t. It’s caused by those darn pixels acting up.

NOTE: Chris Meyer has written an excellent blog that goes into lots of detail on how we got ourselves into this mess in the first place. If you are curious, as I was, you’ll enjoy reading this:

 

http://provideocoalition.com/index.php/cmg_keyframes/story/par_for_the_course/

NON-TRANSPARENT IMAGES

To design images that behave properly upon import, we need to create them using specific dimensions that adjust for these differences. While the math can take another page or two to explain, if you are creating non-transparent images for import into Final Cut, this table gives you the sizes you need to know.

 

Video Format	Aspect Ratio	100% Size	2.5x Moves
DV NTSC	4:3	720 x 540	1800 x 1350
	16:9	853 x 480	2133 x 1200
SD NTSC	4:3	720 x 547	1800 x 1368
	16:9	853 x 486	2133 x 1215
PAL	4:3	768 x 576	1920 x 1140
	16:9	1024 x 576	2560 x 1140
HD 720	16:9	1280 x 720	3200 x 1800
HD 1080	16:9	1920 x 1080	4800 x 2700

If you don’t want to create moves around the image, size your image in Photoshop using the pixel dimensions in the “100% Image” column. Because images always look better when they are scaled at 100% or smaller, if you want to move around (the “Ken Burns” effect) inside an image, create it at the size of the “2.5x Moves” column.

Images are always created at 72 dpi, because DPI is relevant when printing. For video, the key number is the total number of pixels in each dimension. Also, Final Cut Pro 6 or earlier does not like images greater than 4,000 pixels on a size; it tends to cause the application to crash.

Recent versions of Photoshop have included the ability to change the pixel aspect ratio (or shape) to match the video format of your sequence. The problem is that Adobe and Apple calculate these ratios differently. Which means that the Adobe sizes don’t work perfectly for Final Cut, though they do work perfectly for Premiere Pro. (sigh…)

So, for non-transparent images, if you are in a hurry, the Photoshop presets will get you close. If you want to be perfectly accurate, use the numbers above in the table.

IMAGES WITH TRANSPARENCY

When we create images with transparency, however, things become more confused. We still have all the issues of pixel aspect ratio to worry about, but Final Cut treats these images differently. Non-transparent images are imported as graphic files. Transparent images (that is, PSD files) are imported as sequences.

So, graphics files get adjusted for differences in pixel aspect ratio. Sequences do not. And here’s where I made my mistake. I decided that the best place to correct for this was in Final Cut. While this works, it is cumbersome and VERY confusing. A better place to adjust for this is in Photoshop. And that is what I want to explain now.

Here’s the executive summary: like a non-transparent graphic, you need to create your image in Photoshop at specific sizes, depending upon video format. Then, when design is complete, you need to alter its size in Photoshop to match the video format you are using. Finally, before importing into Final Cut, you need to be sure your Easy Setup matches the video format of your sequence, because FCP adjusts the size of imported sequences to match the current settings of Easy Setup.

Complicating matters – as if it weren’t already bad enough – the size you alter your image to varies by video format, because different video formats use differently shaped pixels. (Remember, all images destined for video get created at 72 dpi.)

Here’s the table with the numbers you need.

Video Format	Aspect Ratio	Master Image	Squished Image
DV NTSC	4:3	720 x 540	720 x 480
	16:9	853 x 480	720 x 480
SD NTSC	4:3	720 x 547	720 x 486
	16:9	853 x 486	720 x 486
PAL	4:3	768 x 576	720 x 576
	16:9	1024 x 576	720 x 576
HDV 720	16:9	1280 x 720	1280 x 720
HDV 1080	16:9	1920 x 1080	1440 x 1080
DVCPRO HD (P2) 720	16:9	1280 x 720	960 x 720
DVCPRO HD (P2) 1080	16:9	1920 x 1080	1280 x 720
AVC 720	16:9	1280 x 720	1280 x 720
AVC 1080	16:9	1920 x 1080	1920 x 1080
RED 1080	16:9	1920 x 1080	1920 x 1080
XDCAM HD 720	16:9	1280 x 720	1280 x 720
XDCAM HD 1080	16:9	1920 x 1080	1440 x 1080
XDCAM EX 720	16:9	1280 x 720	1280 x 720
XDCAM EX 1080	16:9	1920 x 1080	1920 x 1080

Video Format. The video format of your Final Cut sequence.

 

Aspect Ratio. The aspect ratio of your Final Cut sequence.

 

Master image. The size, in pixels, of your original design. All design work and alterations are done at this size. This becomes the master file from which all size adjustments are made. The resolution of this image is always 72 dpi.

 

Squished image. The size, in pixels, of your complete design prepared for import into Final Cut. The resolution of this image is also 72 dpi.

THE PROCESS

Create a new image according to the Master size column in the table above. Do all your design work in this file. Make it perfect. Get it approved. When your image is complete, you will need to resize it according to the Squish column.

To do this, go to Image > Image size.

Turn off constrain proportions – this allows you to set the size of the width and height independently.

According to what I’ve read on the National Association of Photoshop Professionals website, when changing the size of an image, you will most often get the best results by changing the bottom pop-up to Bicubic Sharper. This improves image fidelity when reducing the size of an image.

Finally, change the size of either the width or the height to match the settings in the Squished column above and click Save.

You now have two versions of the file: the master version, which you’ll use if you need to make changes, and the file prepared for import into Final Cut.

Next, in Final Cut be SURE the Easy Setups match the video format you are editing. This is a critical step, because Final Cut configures your imported sequence to match these settings.

Finally, import your graphic. Because this is a PSD file, Final Cut imports it as a sequence. Double-click the sequence to open it into the Timeline, at which point, you can copy and paste elements from one sequence to the next. In this screen shot, the circle and the two smaller squares came in perfectly.

It took a long time to get here, but we finally have a reliable way to prepare images for Final Cut that gives us accurate imports whether the image contains transparency or not. Having to worry about two versions of the same file — master and squished — is not ideal. But, the alternative of constantly fighting to get your images to look right is far worse.

Whew!

---

TIPS TO IMPROVE YOUR TEXT

So, given all these constraints, it’s a wonder we can create anything on our computers that displays properly on video. But, we can. And here are some tips that can help.

1. Avoid lines thinner than 4 pixels, they will flicker
2. Avoid lines that are almost horizontal or almost vertical, they will stair-step
3. Avoid fancy, curvy fonts like Harrington, Chancery, or Savoye
4. Avoid fonts with very thin bars or serifs, such as Engravers, Baskerville, or Modern
5. Avoid point sizes smaller than 24 points
6. Avoid saturated colors, use the Vectorscope if you are unsure
7. Avoid whites brighter than 92% (235 in PhotoShop)
8. Always use drop shadows on any text you want your viewers to read
9. Always keep any text you want viewers to read inside Title Safe
10. Keep text on screen long enough for you to read it three times.

Remember that text always looks better on your computer screen than it does on a TV set. So, if you can’t read it inside Final Cut, your viewers won’t be able to read it either.

---

UPDATE – AUGUST, 2009

Of all the subjects I’ve wrestled with, nothing is harder to pin down than an efficient system for working with still images. I’ve been wrestling with this subject for years.

In my recent article, and video tutorial, on working with stills I divided still images into two categories: those without transparency, such as JPEGs, PNGs, and TIFFs; and those with transparency, such as PSDs.

This provided a simple and consistent way to create stills in Photoshop and import them into Final Cut with all their geometry and image quality intact. The system works, reliably, and consistently. So I published it.

At which point, Mark Spencer – who is the most knowledgeable person on Apple Motion that I’ve ever met – sent me the following:

I’m sure you’ll get a lot of email on this one.

 

“Transparent images (that is, PSD files)”

 

What about TIFFs and PNGs (and PICTs) that have an alpha channel? How should those be treated?

Larry replies: Sigh… I didn’t even know PNGs and PICTs had alpha channels.

My honest answer is “I don’t know.” If FCP opens a TIFF with transparency as a sequence then it needs to be sized as you would a PSD.

On the other hand, most people (I’m guessing) use PSDs for transparency, while PNGs and TIFFs are used for non-transparent images, in which case my article works.

If you have a better way to classify these images, I am COMPLETELY happy to hear it. For now, though, I think these classifications work, in general, for most people. I hope.

Mark then wrote back:

I don’t think transparency has any impact on how FCP interprets the file – I think all that matters is whether it comes in as a single file or a sequence – if it comes in as a sequence, you need to resize (“squish”) first.

Larry adds: On this I agree. The bigger question, though, is what criteria Final Cut uses to determine whether to open an image as a graphic or as a sequence. My suspicion is that transparency is the determining factor.

UPDATE – Aug. 31, 2009

Ryan Mast writes:

A PNG with or without transparency will be treated as an image, not a sequence, in a Final Cut timeline (at least as of Studio 2).

 

Unless I’m doing animations or effects with layers, I’ll usually export a Photoshop file into PNG if I’m bringing it into Final Cut or Motion. I’m probably doing it wrong, but I can’t reliably get the TIFF’s alpha channel to be recognized by Final Cut.

Lee Berger adds:

I use PNG as it supports transparency and imports into FCP as a graphic and not a sequence.

 

I often use Motion for titles instead of Title 3D because it’s better at soft shadows. The one thing I don’t like is the long renders at the high quality setting. To avoid this I export the title from Motion as a PNG. When imported and added as a key the PNG file requires no rendering. The downside is that you cannot use the “Open in Editor” function if you need to update the title, but it’s worth it to avoid the long Motion file render.

Larry replies: Thanks!

---

UPDATE – NOV, 2009 — SELECTING IMAGE QUALITY

John Martellucci writes:

I have watched your tutorial on Lynda.com covering moving stills… resizing to 1800 x 1350 x 72, Bicubic options, sRGB gamma setting, etc… fantastic info!

 

The project I am working on will incorporate both video and photos. I have 490 scanned photos (but in jpeg format).

 

I’m planning to bring all the photos into Photoshop and make the adjustments you explained in your tutorial… and per your recommendation, was also planning to save them as a Tiff… with the LZW compression.

 

However… since the existing format of the photos I am bringing into Photoshop is jpeg, will the quality be improved by changing it to Tiff… or are my efforts pointless and futile?

Larry replies: John, thanks for writing!

Converting a JPEG to TIFF won’t improve quality. It just changes the format. Think of pouring a cup of water into a five gallon bucket. You’ve changed the size of the container, but not the content.

Since you are going to standard def video, as I assume from the image sizes you mentioned, JPEG will “probably” be OK – do a test and see if it works for you. Things that may get damaged are thin lines, fine detail, subtle textures – which SD video tends to not display anyway.

For HD video, it will probably be better to re-scan as TIFF — BUT!!!!! do some tests first. If you can’t see a difference on a good monitor, there probably isn’t a difference. TIFF or PNG are better, but when compared to a really high-quality JPEG, it may not be enough to notice.

If you are creating this project for digital cinema, rescan the images to TIFF. When projecting to large screens, anything you can do to improve image quality is a good idea.

In general, I am not a fan of JPEG, compared to PNG or TIFF, due to the lower quality of JPEG. However, if JPEG is all you have to work with, Final Cut can easily edit them.

---

Bookmark the permalink.