2:3 Pulldown Cadence Diagram

Why pulldown exists

Film shoots and finishes at 24 (or its NTSC-compatible cousin, 23.976) frames per second — progressive, one full image per frame, no fields. NTSC video runs at 59.94 interlaced fields per second, which is 29.97 full frames, each frame built from two half-resolution fields captured 1/59.94th of a second apart. Twenty-four does not divide evenly into thirty, or into 29.97, so getting a film frame rate onto an NTSC signal without changing the film's actual running speed means spreading each film frame across an uneven number of video fields. 2:3 pulldown — also written 3:2 pulldown, depending which end of the pattern a text starts counting from — is the standard telecine cadence that does this: every four film frames become ten interlaced fields, which pair up into five video frames. Run that cycle continuously through a whole telecined show and 24 film frames per second becomes exactly 60 fields per second — 29.97 video frames per second once NTSC's 1000/1001 slowdown is folded in.

The 2:3 cadence, field by field

Label four consecutive film frames A, B, C, D. Pulldown does not give each film frame the same number of fields — it alternates 2, then 3, then 2, then 3:

That is 2 + 3 + 2 + 3 = 10 fields from 4 film frames — the "2:3" the cadence is named for. Those ten fields then pair up, two at a time in the order they were produced, into five interlaced video frames:

Walk the arithmetic: 23.976 fps × 10 fields per 4 film frames = 59.94 fields per second = 29.97 frames per second. That is the whole trick in one line — real NTSC telecine runs at exactly these 1001-based fractional rates; "24-into-30" is a convenient rounding, not what actually happens on the wire.

Dirty frames and comb artifacts

F3 and F4 are dirty because each interleaves two half-resolution snapshots of different moments — a field from B's exposure sitting directly above a field from C's exposure. On a static shot that is invisible. The instant anything in frame moves, the two source images no longer line up scanline to scanline, and the result is the classic telecine "comb" — fine horizontal tearing that shows up worst when the material is paused, stepped frame by frame, or scaled without deinterlacing. A dirty frame holds no more image data than a clean one — every video frame in the group is built from exactly two fields; what sets F3 and F4 apart is that their two fields come from two different film exposures, two distinct moments in time, instead of one.

Reverse pulldown (detelecine) and why edits need to remove it first

Because the cadence is a fixed, repeating 2:3:2:3 pattern, it can be undone: reverse pulldown, or detelecine, inspects ten fields at a time, finds which pairs are clean and which are the two-source splices, and reconstructs the original four progressive film frames — recovering true 23.976 progressive from 29.97 interlaced video. This only works if the cadence stays intact. Cut a telecined master with a standard video editor that does not track the pulldown pattern, and an edit point can land in the middle of the repeating 2:3:2:3 cycle instead of on a film-frame boundary. That either strands a dirty frame at the cut, breaks detelecine software's ability to find the pattern across the splice, or — worst case — introduces a repeated or dropped frame that shows as a visible motion judder right at the cut. That is why professional telecine workflows either conform on the original film-rate media, or run detelecine before any cutting happens, so every downstream edit works on clean progressive frames instead of an interlaced cadence that has to be rediscovered at every cut.

2:3:3:2 "advanced" pulldown — the sibling cadence

A variant cadence, 2:3:3:2 (sometimes called "advanced" pulldown), rearranges which film frames get 2 versus 3 fields so the two dirty pairings land next to each other instead of one apart — B and C effectively swap roles — which produces a group of five video frames where only a single frame ends up interleaving two film sources, instead of two separate dirty frames spread across the group. That single dirty frame can then be dropped losslessly by codecs like DV and HDV that store fields as discrete, addressable units, recovering exact 23.976 without a full detelecine pass. This tool diagrams the standard 2:3 cadence only — 2:3:3:2 is a related, out-of-scope cadence worth knowing about if the footage in front of you does not match the pattern above.

Field order: TFF vs BFF

Every interlaced frame is drawn from two fields, and one of them has to be assigned to the upper set of scanlines and one to the lower — that assignment is field order. Top-field-first (TFF) draws the first field of each pair on the upper scanlines; bottom-field-first (BFF) draws it on the lower scanlines. The field sequence — which fields exist and in what order they were produced — never changes; only which physical scanlines each field lands on does. Get field order wrong on export or capture and every frame still displays, but each one draws its two fields in the wrong vertical position relative to when they were actually captured — motion appears to stutter or judder within each frame, and any downstream detelecine will mis-pair fields against the wrong neighbor.

1. Set the field order to match your source, top field first or bottom field first.

2. Drag the slider, or step one field at a time through the full ten-field cadence.

3. Watch which film frame feeds each field and where F3 and F4 comb, then move a timecode between rates in the Timecode Frame-Rate Converter.