Timecode Frame Rate Converter

Knowledge Base

Two conversions, and why the answer differs

Converting a timecode between frame rates has two correct answers, and they are not interchangeable. Most of the trouble people have with conform and sync comes from taking one when they needed the other, so this tool shows both rather than choosing for you.

Same frame count

Every frame keeps its position in the sequence. The five hundredth frame is still the five hundredth frame, and only the label on it changes. Nothing is added and nothing is thrown away, so the picture is untouched.

What moves is the duration. A thousand frames at 23.976 last a little over forty-one and a half seconds. The same thousand frames interpreted at 25 last exactly forty. This is the conversion behind conforming a 23.976 master to 24 for a cinema deliverable, and behind the old habit of speeding a 24 fps film up by four per cent for a 25 fps television market. The audio has to be pitched or stretched to match, because it is now playing against a different clock.

Same moment in real time

The instant is preserved and the frame number is recalculated to whatever lands nearest. One hour into a 23.976 recording is still one hour into the 25 fps version, and the frame count either side of that instant is different because the two are counting at different speeds.

This is the conversion you want when matching a camera to a sound recorder, lining up two cameras that were running at different rates, or checking a delivery against a clock. It is also the one that introduces rounding. A moment that falls between two frames at the target rate has to pick one, so a conversion and its reverse will not always return the exact frame you started on.

Where the fractional rates come from

The rates that look wrong are the ones ending in .976 and .97. They exist because of a decision taken when colour was added to a monochrome television system. The colour subcarrier had to sit where it would not interfere with the existing sound carrier, and the arithmetic that solved it slowed the frame rate by a factor of exactly 1000 divided by 1001. Thirty became 29.97, and twenty four became 23.976 when the same relationship was carried into digital cinema workflows.

The consequence is that a timecode counter running at 29.97 and labelled in whole frames drifts against a real clock. After an hour it is roughly three and a half seconds ahead of where the wall clock says it should be. That error is small enough to ignore on a short clip and large enough to ruin an hour-long programme delivered to a broadcaster.

What drop frame does, and what it does not do

Drop frame is the fix for that drift, and its name misleads almost everyone who meets it for the first time. No pictures are discarded. What is dropped is two labels per minute, skipped at the start of every minute except every tenth, so the counter catches up with real time. A recording numbered in drop frame contains exactly as many frames as one numbered without it.

You can see the skip directly. In drop frame the count runs to 00:00:59;29 and the next frame is labelled 00:01:00;02. The two labels in between were never used. Over an hour that removes one hundred and eight labels, which is why one hour of drop frame timecode counts 107,892 frames rather than 108,000.

Drop frame applies only to the 29.97 and 59.94 families. There is no such thing as drop frame at 25, because 25 is exact and has no drift to correct. A file marked as drop frame at 24 or 25 has been mislabelled somewhere upstream.

The convention for writing them apart is a semicolon or a full stop before the frames in drop frame, and a colon in non-drop. It is worth honouring, because the same digits mean different instants under the two schemes.

FAQ

Q: Which conversion should I use?
A: If the media file is unchanged and you are relabelling it, use the same frame count. If you are matching something to a clock or to another source, use the same moment in real time.

Q: Why does converting back not always give me the frame I started on?
A: Only the real-time conversion has that problem, and it is rounding. A moment that falls between two frames at the target rate has to choose one, and the choice is not always reversible.

Q: Is 23.976 the same as 24?
A: No. They are related by 1000 divided by 1001, which is a difference of about one frame in a thousand. Over a two hour feature that is around seven seconds, so treating them as the same is not safe on anything long.

Q: My camera says 30 but the file says 29.97.
A: That is normal. Cameras and menus routinely round the label. The file is authoritative, and if the two disagree the fractional rate is almost always the real one.