Frame Rates are Tricky Beasts

My goal in this article is to discuss the challenges in converting frame rates.

If everything you shoot, edit and output is a single frame rate, then don’t change anything. This is the ideal way to work. However, as you start to integrate elements that originate at different frame rates, frame rate conversion rears its very ugly head.

DEFINITIONS

Think of a video clip as a series of wooden children’s blocks connected by a piece of string. Each block represents a frame of video. As we pull the string, tugging the blocks along in a line, the frame “rate” represents the number of blocks (or images or frames) that pass an observer each second. Frame rate is measured in frames per second; “fps.”

Changing the speed of a clip is NOT the same as changing the frame rate.

• Changing the speed of a clip always speeds up or slows down the action displayed in the clip by repeating or removing frames.
• Changing the frame rate of a clip changes the number of frames passing an observer without changing the perceived speed of the action displayed by the clip.

This difference is significant. The first is easy, the second is hard.

We change the speed of a clip to create a visual effect. We change the frame rate of a clip to match the settings of our clip to the project. If you don’t need to match settings, don’t mess with changing frame rates.

SETTING BOUNDARIES

There are two sides to a frame rate discussion:

• Aesthetic
• Practical

There is a lot of debate as to which is the “best” frame rate. Some feel that 24 fps is more “cinematic,” while 60 fps is more “real.” As you should know by now, there is no “best.” Just as there is no “best” car, camera, or restaurant; there are simply choices.

Converting to a 24 fps frame rate will NOT make your movie look “filmic.” It will, generally, just make it look worse. The “cinematic look” is a combination of: lenses, lighting, depth of field, shutter speed, shutter angle, motion blur and frame rate. Changing the frame rate only affects the frame rate, not the look.

There are no right answers, just louder voices.

Also, to keep this article to a manageable length, I will ignore:

• Over-cranking
• Under-cranking
• Frame blending
• Optical Flow
• Field-based editing

These special cases don’t alter the basic rules of frame rates, though they can complicate understanding.

THE BASIC RULES

Whether you use Adobe, Apple, Avid, or any other video editing software on Macs, PCs or mobile devices, the basic rules of frame rates remain the same:

• Where possible, always shoot, edit and output the same frame rate.
• Camera-native frame rates always look better than converted frame rates.
• Frame rates are irrelevant on the web. Streaming and downloadable files can play at any frame rate.
• Frames are indivisible. (Think about those children’s building blocks. They can’t be split, stretched, squeezed or discombobulated.)
• The frame rate of your project/sequence takes precedence over the frame rate of your source media.
• The duration of a frame is one frame. Not more, not less.
• Partial-frame durations do not exist for video, though they do for audio. (You can’t display images for half-a-frame, or a quarter-frame. If your project/sequence is 25 fps, then the shortest duration you can display an image is 1/25th of a second.)
• Video editing software automatically “conforms” (converts) clips with different frame rates to match the frame rate of the project/sequence.
• Projects/sequences can only have one frame rate, though they can contain clips that use different frame rates.
• It is “easier” and “better” to convert from a faster frame rate (i.e. 50 fps) to a slower one (i.e. 25 fps), than from a slower frame rate to a faster one.

HOW WE GOT TO TODAY

In the early days of film, say 1890 – 1915, all cameras were hand-cranked. During this time, frame rates wandered from 8 fps to 30 fps, often in the same scene. In those days, the value of a camera operator was not based on their composition skills, but on the consistency of their cranking.

NOTE: This is one of the reasons comedies were so prevalent in the early days of film. Speed changes are inherently comedic and physical comedy does not require dialog.

As films grew in popularity and profitability, standards developed allowing cameras to be cranked by a motor, rather than by hand. Also, at this time, the industry settled on a frame rate of 18 fps.

Why? Because film was expensive and producers were, um, cheap. 18 fps provided the illusion of smooth movement without wasting a lot of film and money.

This standard continued up until the advent of talkies, which exploded on the scene in 1927 with The Jazz Singer. The problem was that 18 fps was not fast enough to support high quality audio. This frame rate yielded audio roughly equivalent to a telephone call.

So, a new frame rate standard needed to be developed – and the industry chose 24 fps.

Why? Because film was expensive and producers were still, um, cheap. 24 fps provided the illusion of smooth movement with relatively high-quality sound without wasting a lot of film, and money.

NOTE: Sound quality continued to improve over time, not by increasing the frame rate, but by shifting audio from an optical track to a magnetic track.

When video arrived, in the 1930’s, we had a major timing problem. How to get the TV receiver to “pulse” in sync with the transmitter? The solution was AC power. All across the US, power “pulsed” at 60 cycles per second.

Television engineers adopted this “universal” pulse as the basic timing circuit for video. Since video in those days was interlaced, where a single frame (complete image) was composed of two fields (a portion of the image consisting of all the odd or even scan lines), each field pulsed at 1/60th of a second.

Ta-dah! 30 fps video.

Except, over time it was discovered that high-voltage electricity “evaporated” from transmission lines when the cycle rate was too high. 50 cycles preserved more power over distance than 60 cycles (now called Hz). So, when much of the world was rebuilt after World War II, the utility companies, to save money and power, dropped the cycle rate to 50 Hz.

From there, the video industry derived 25 fps video, because interlacing was still in vogue.

So, at the dawn of the HD era in the early 1990’s, we had three principle frame rates: 24, 25, and 30 (which, with the advent of color was slightly modified to 29.97 fps, because why should this story be particularly simple?)

And, as we all know, with the rise of HD, our industry came together as a group and standardized on a single frame size and single frame rate.

– – –

Sigh… No such luck.

At last count, we now have nine different frame rates: 23.98, 24, 25, 29.97, 30, 48, 50, 59.94 and 60. (And, yes, 100 and 120 fps are knocking on the door. Please keep that door shut…!)

No WONDER we’re all confused. We’ve been handed a veritable Gordian Knot of frame rates!

CONVERSION OPTIONS

Frame rate conversion is the process of duplicating or removing frames such that, when the clip is played in a sequence that matches the frame rate of the clip, all action appears at “normal” speed. (I put “normal” in quotes because I couldn’t figure out an easy or good way to define normal.)

Most of the time, video editing software will automatically handle frame rate conversions. And, most of the time, I suggest you not worry about it, because, most of the time, it will look fine.

REMEMBER: “Camera-native frame rates always look better than converted frame rates.”

When it comes to frame rate conversions, there are easy options and hard options. Following the wooden block analogy I introduced earlier, we can’t just stretch frames to different rates because each frame is made of wood, not Silly Putty. Instead, we change frame rates by inserting or removing entire blocks.

EASY: 50 fps to 25 fps – or 60 fps to 30 fps

Assuming the video is progressive, conversion simply deletes every other frame.

If the video is interlaced, one field is deleted while the scan lines in the other field are duplicated. (Yes, this option reduces image quality. That’s one reason I hate interlaced video.)

EASY: 29.97 fps to 59.94 fps – or 25 fps to 50 fps.

Here every frame is duplicated. This does not create slo-mo because the video plays back at 50 fps, yielding the same movement as playing 25 fps video in a 25 fps project.

This does not significantly degrade movement quality, but movement will look more fluid if you shot 50 fps (or 60) originally.

MOSTLY EASY: 24 fps to 25 fps

The traditional way of converting 24 fps to 25 fps is increasing the speed of the 24 fps material 4%. This allows all frames to be displayed and, while the action is a bit faster, it isn’t so much faster that the audience will perceive it.

NOTE: Yes, this speed change means we need to speed the audio as well. There’s no free lunch.

HARD: 24 fps to 29.97 fps

This was done traditionally when converting films for television broadcast using a telecine.

Here, we need to create, essentially, six “new” frames every second. (The difference between 24 fps and 30 fps.) But we are dealing with children’s blocks here, we can’t create new images, we can only create new frames that contain existing images.

There are several ways to do this, depending upon whether you are working with interlaced or progressive images. The interlaced methods are quite complex and involve duplicating specific fields, not just frames.

But, here’s a simple method to illustrate the process. Take a group of four frames, then duplicate the last frame in the group. Over 24 frames this creates 6 new frames.

When played back at 30 frames a second, most viewers won’t notice the duplicated frame. However, for the discerning, your action will slightly stutter every five frames. This illustrates why you want to avoid converting frame rates.

HARD: 60 fps to 24 fps

Three quick reminders:

• “Where possible, always shoot, edit and output the same frame rate. Camera-native frame rates always look better than converted frame rates.” There’s no free lunch, converting frame rates will not look as good as shooting 24 fps natively.
• The “cinematic look” is not based on frame rate; converting to 24 fps won’t make your movie look “filmic.”
• Frames are indivisible. We can’t invent new frames.

NOTE: Optical flow seeks to do just that, invent new frames. However, while good in theory, the results are often worse than not using optical flow.

Here’s an example of how this could be handled: We need to remove 36 frames from every second of video. Since both 60 and 24 are divisible by 3, we can divide each second into three “blocks,” or sections.  This means that a  20 frame block in the source clip needs to be converted into an 8 frame block in the destination clip. To do this:

• Play frame 1
• Delete frame 2 in the source clip
• Play frame 3, this becomes the new frame 2 in the destination block
• Delete frames 4 and 5
• Play frame 6, this becomes the new frame 3
• Delete frame 7
• Play frame 8, this becomes the new frame 4
• Delete frames 9 and 10
• Play frame 11, this becomes the new frame 5
• Delete frame 12
• Play frame 13, this becomes the new frame 6
• Delete frames 14 and 15
• Play frame 16, this becomes the new frame 7
• Delete frame 17
• Play frame 18, this becomes the new frame 8
• Delete frames 19 and 20
• Repeat this process until you reach the end of the clip

As you can see, asymmetrical trimming (remove 1 frame, then remove 2, then remove 1…) gets us to the frame rate we need, but at the expense of potentially adding jitter to movement; say during an actor’s walk or as a car drives through a scene.  Whenever we convert frame rates asymmetrically, we run the risk of damaging the movement in the clip.

A SIDEBAR ON VIDEO COMPRESSION

Video can be compressed in one of two ways:

• As individual images (I-frame)
• As a group of pictures (GOP)

I-frame formats include: ProRes, GoPro Cineform, AVC-Intra and the DNx family of codecs.

GOP formats include: AVCHD, H.264, HDV and most formats that generate very small file sizes.

As I was writing this article, it occurred to me that camera-native GOP-format video will probably suffer more from image degradation as you change frame rates than video that was shot as I-frame media.

I haven’t tested this, and would like to hear other opinions, but if you are seeing blurry images when changing frame rates by small increments, I would suspect your video format is too blame.

SUMMARY

Frame rates are complex. However, a little planning ahead can simplify headaches. The best option – always – is to shoot the frame rate you need to output.

And, keep in mind, that a “film-like look” does not necessarily require a “film-like frame rate.”

---

Bookmark the permalink.