Frames, time, and why the numbers are strange

A video is a stack of still pictures. That is the whole of it. Everything difficult about time in editing comes from decisions made between 1927 and 1953 that nobody has been able to undo since.

The frame is the unit, not the second

You cannot cut halfway through a frame. It is the smallest thing that exists, so it is the smallest thing you can edit on. At 30 frames per second a frame is 33 milliseconds, which sounds like nothing until you are trying to land a cut on a consonant and you are one frame late every time.

This is why timecode counts frames in its last field rather than hundredths of a second. 00:01:23:14 means one minute, twenty three seconds, and fourteen frames. At 30fps the frame after :29 is :00 of the next second. At 24fps it is the frame after :23. The number you see depends on the rate, which is the first thing that catches people out.

24, 30, 60, and what each one means

24 is the film rate. It was chosen in the late 1920s because it was the slowest speed that gave acceptable sound quality on an optical track, and sound was the reason the rate had to be fixed at all. It has the stutter people call cinematic. Pans at 24 judder. Everyone has decided to like this.

30 is the North American television rate, derived from the 60Hz mains frequency. 25 is the European one, derived from 50Hz. Both exist because early receivers used the power line as a timing reference and any mismatch showed as a rolling bar on screen.

60 is for things where fast motion has to stay readable: sport, gameplay, anything where a viewer is tracking a small fast object. It costs you roughly double the bits of 30 for the same apparent quality, because there are twice as many frames to encode.

Higher is not better. It is a different feel. Shooting a talking head at 60 does not make it look more expensive, it makes it look like local news.

23.976, and the tenth of a percent that ruins everything

When colour arrived in 1953, the colour information had to be squeezed into a signal already carrying black and white, and the subcarrier that carried it beat against the audio carrier in a way viewers could see. The fix was to slow the whole frame rate down by a factor of 1000/1001. 30 became 29.97. 24, when transferred to video, became 23.976.

Seventy years later, with no black and white receivers left anywhere, cameras still record 23.976 and 29.97 because everything downstream expects them.

Here is what it costs you. Put 23.976 footage on a 24fps timeline and the picture plays a tenth of a percent fast. That is 0.0864 seconds every minute. Over a five minute piece the sound ends about four tenths of a second away from the picture, which is enough to see. Over an hour it is three and a half seconds, which is a disaster you find at the very end.

Drop frame timecode drops no frames

At 29.97fps, labelling frames 0 through 29 for each second means your timecode runs fast against a real clock by 3.6 seconds every hour. Broadcast could not live with that, because a half hour programme has to be a half hour.

Drop frame skips two frame numbers at the start of every minute except every tenth minute. No pictures are removed. The labels just skip from 00:01:59:29 to 00:02:00:02. Over an hour that recovers exactly the 3.6 seconds of error.

You will see it written with semicolons, 01;00;00;00, to distinguish it from non drop. If nobody has asked you for broadcast delivery you can ignore all of this, so long as you do not mix the two in one project.

Speed, and why slow motion usually looks bad

Playing a clip at half speed means the software needs two frames for every one it has. It can hold each frame twice, which stutters. It can blend neighbouring frames, which looks like a double exposure. Or it can run optical flow and invent the in between frames by guessing how the pixels moved, which works well until something passes in front of something else and then hands melt.

The real answer is to shoot at a high frame rate and slow it down to a normal one. 120fps played at 30 is a true quarter speed with a real frame for every moment, and it looks right because it is right.

Speeding up has the opposite problem, which is audio. Running a clip at 150 percent raises its pitch by about a fifth unless the tool corrects for it. Most do by default now, and the correction is audible on music long before it is audible on speech.

Motion blur is doing more work than the frame rate

What makes 24fps read as movement rather than a flipbook is that each frame is slightly smeared. The sensor was open for half the frame time, so anything that moved during that window left a streak. That is the 180 degree shutter convention, and at 24fps it means a shutter speed of 1/48th.

Footage shot at 1/500th at the same frame rate has no blur at all, and it stutters horribly on any pan. Saving Private Ryan used that deliberately for the landing sequence. Your interview footage shot at 1/500th because it was bright outside did not.

No software puts convincing motion blur back. Check the shutter at the shoot.

What to actually do

• Find out the frame rate of your source before you build a timeline. Every editor shows it on import.
• Build the timeline at the source rate. Convert once, deliberately, if you must.
• If you have 23.976 and 30fps material in one piece, decide which one the timeline is and let the software conform the other. Do not eyeball it.
• Shoot at 1/48th or 1/50th at 24 and 25, and 1/60th at 30, unless you want the stutter.
• Shoot high and slow down. Do not shoot normal and ask software to invent frames.