Last updated: July 2026.
This is the sibling page to our aspect ratio cheat sheet: every frame rate you’ll meet in post, what each one is for, and the exact numbers behind the weird ones, all on one page. Same drill as last time… bookmark it. Cmd+D on a Mac, Ctrl+D on a PC.
The quick table comes first. Below it: why 29.97 exists at all, the drop-frame arithmetic, a conform table with speed and pitch figures for every common rate pair, how the three big NLEs handle mixed-rate timelines, and delivery norms by region.
The frame rates you’ll actually use
Eight rates cover essentially everything. The first trap is the naming: camera and NLE menus love to say “24,” “30,” and “60” when they mean 23.976, 29.97, and 59.94. If you’re in the US and you didn’t deliberately set an exact integer rate, you’re on the fractional one.
| Rate | Exact value | What it is | When to use it |
|---|---|---|---|
| 23.976 (23.98) | 24 ÷ 1.001 | The NTSC-compatible film rate | The default “24p.” Streaming originals, US film-look work, most cameras’ 24p mode |
| 24 | 24.000 | True cinema rate (DCI) | Theatrical DCP and film-out. Don’t pick it for “film look” on a streaming job unless the spec says so |
| 25 | 25.000 | PAL broadcast rate | Anything delivering into 50 Hz countries: UK, Europe, Australia, most of Asia and Africa |
| 29.97 | 30 ÷ 1.001 | NTSC broadcast video | US and Japanese broadcast delivery: news, sports, unscripted, corporate with a broadcast leg |
| 30 | 30.000 | Round web rate | Web and social work with zero broadcast legacy |
| 50 | 50.000 | PAL high frame rate | European HFR sports and broadcast; slow-motion source for 25p timelines |
| 59.94 | 60 ÷ 1.001 | NTSC high frame rate | US sports and HFR broadcast; what most cameras record when the menu says “60” |
| 60 | 60.000 | Web and gaming HFR | Screen recordings, gaming, web HFR with no broadcast target |
Frame rate picked? The other half of the delivery puzzle is resolution and bitrate, and the per-platform numbers for YouTube, Instagram, and TikTok live in our export settings reference. And if you landed here mid-panic because a delivery sheet says “PAL,” the conform table below has your exact speed and pitch numbers.
Why 29.97 exists (blame color TV)
Every fractional frame rate in your NLE traces back to one decision in December 1953.
Black-and-white NTSC ran at a clean 30 frames per second, 60 interlaced fields, locked to the 60 Hz US power grid. When the FCC approved color broadcasting, the color information had to ride along in the same channel without breaking the millions of black-and-white sets already in living rooms. The fix was a color subcarrier at 3.579545 MHz, chosen so its energy would interleave neatly with the luminance signal. To keep that subcarrier from beating against the audio carrier and putting visible interference on screen, the horizontal line rate shifted from 15,750 to 15,734.264 lines per second. Divide by 525 lines per frame and you get 29.97003 fps: 30 divided by 1.001, or 0.1% slow. The change was small enough that the old sets kept working, and we’re still typing semicolons into timecode fields 70 years later.
23.976 follows directly. Film runs at 24, and to broadcast film on NTSC, telecine machines used 3:2 pulldown (next paragraph) to map 24 frames onto 60 fields. Once video slowed to 29.97, film had to slow by the same 0.1% so the pulldown landed exactly: 24 ÷ 1.001 = 23.976. Doubling gives 59.94.
Europe never had this problem. PAL and SECAM ran 50 fields on 50 Hz mains, 25 frames per second, and their color encoding avoided the subcarrier conflict, so 25 and 50 stayed exact integers. That’s the whole family tree: everything descended from NTSC is off by 1000/1001, everything descended from PAL is clean.
3:2 pulldown in one paragraph
Four film frames, call them A B C D, become ten video fields: A is scanned for three fields, B for two, C for three, D for two. Ten fields is five interlaced video frames, so 24 film frames per second come out as 60 fields, or 30 frames, running at the corrected 29.97. To edit that footage as clean progressive frames you reverse the process (“reverse telecine”), which discards the duplicated fields and reconstructs the original frames at 23.976.
Drop-frame vs. non-drop-frame timecode
Timecode has no way to count fractional frames, so 29.97 footage gets labeled as if it ran at 30. That lie costs 3.6 seconds per hour, and drop-frame timecode is the accounting trick that pays it back.
The arithmetic: one wall-clock hour at 29.97 fps is 107,892 frames. A counter labeling 30 frames per second needs 108,000 labels to reach 01:00:00:00. So when non-drop timecode shows exactly one hour, an hour and 3.6 seconds have passed in the real world. That’s 108 frames of drift per hour, close to a minute and a half per day. For editorial nobody cares. For broadcast, where a “one hour” program has to fill exactly an hour of air time, it’s a scheduling problem.
Drop-frame fixes it by skipping labels, not frames. “Wait, it deletes frames from my footage?” No. Not one frame of video is touched. The counter skips frame numbers 00 and 01 at the start of every minute, except minutes ending in 0, so a DF counter rolling past a minute mark jumps from 00:00:59;29 straight to 00:01:00;02.
29.97 drop-frame
| Labels skipped | 2 per minute (;00 and ;01) |
| Exempt minutes | :00, :10, :20, :30, :40, :50 |
| Skipped per 10 minutes | 9 x 2 = 18 |
| Skipped per hour | 108 |
| Residual error vs. clock | about 0.1 frame per hour |
59.94 drop-frame
| Labels skipped | 4 per minute (;00 through ;03) |
| Exempt minutes | :00, :10, :20, :30, :40, :50 |
| Skipped per 10 minutes | 9 x 4 = 36 |
| Skipped per hour | 216 |
| Residual error vs. clock | about 0.2 frames per hour |
The notation is the tell: colons throughout for non-drop (01:00:03:18), a semicolon before the frames field for drop-frame (01:00:03;18). Only the two NTSC-fraction video rates ever use drop-frame. Timecode at 23.976, 24, 25, 30, 50, and 60 counts straight, no skipped labels. (23.976 drifts from the clock the same way 29.97 does, but no drop-frame variant exists for it; film-rate projects run NDF and nobody schedules air time on them.)
Conform math: speed changes and pitch
Every conform between these rates is a speed change, and every speed change moves audio pitch unless you correct it. The pitch shift is 12 x log2(target rate ÷ source rate) semitones. Run the common pairs through that and you get this table (positive = faster and sharper):
| Conform | Speed change | Pitch shift | Notes |
|---|---|---|---|
| 23.976 → 24 | +0.1% | None needed | Imperceptible; the standard conform when a streaming master becomes a DCP |
| 24 → 23.976 | -0.1% | None needed | Reverse of the above |
| 24 → 25 (“PAL speedup”) | +4.167% | +0.71 semitone (71 cents) | The classic theatrical-to-PAL conversion |
| 23.976 → 25 | +4.27% | +0.72 semitone (72 cents) | The more common real-world case, since most masters are 23.976 |
| 25 → 24 | -4.0% | -0.71 semitone | PAL material onto a true-24 timeline |
| 25 → 23.976 | -4.1% | -0.72 semitone | A UK-shot 25p project delivering to a US streaming spec |
| 29.97 ↔ 30 | ±0.1% | None needed | Same negligible shift as 23.976/24 |
Two notes on that table. Forums quote the PAL speedup pitch shift as anything from 0.68 to 0.72 semitones, and the spread comes down to which ratio people feed the formula: a rounded flat 4% gives 0.68, the exact 25/24 gives 0.71, and 25/23.976 gives 0.72. The computed values above are the ones to use. And yes, plenty of European DVDs and broadcasts shipped with the 4% speedup uncorrected, which is why films run a few minutes shorter on PAL discs with every voice sitting most of a semitone sharp. Pitch-correcting the conform is standard practice today.
Slow motion is the same relationship pointed the other way: capture rate ÷ timeline rate = slow-motion factor. 60 fps on a 24 timeline plays at 2.5x slow, 120 at 5x, 240 at 10x. The NTSC fractions keep the same factors, since the 1.001 cancels: 59.94 ÷ 23.976 is exactly 2.5.
Mixed frame rates on the timeline
Sooner or later a 59.94 drone clip lands in your 23.976 timeline, and each NLE handles that moment differently. The differences cause real conform mismatches on round-trips, so it pays to know all three even if you live in one.
DaVinci Resolve
Resolve handles it at the project level: Project Settings, General Options, Conform Options, “Mixed frame rate format.” The dropdown greys out once media is in the project, so set it before you import. Leave it on None and Resolve conforms every clip to the timeline rate, so a bin full of 23.98, 29.97, 50, and 59.94 material all plays at your timeline’s 24, running slower or faster to get there. The other settings name an application instead, and they conform mixed rates using that application’s retiming math, which is what keeps an imported timeline matching the one it came from. When you’re round-tripping XML or AAF from Premiere or Media Composer, set that conform option to “Resolve” rather than one of the legacy modes; it gives the closest match to the source timeline.
Premiere Pro
Premiere’s default is the opposite of Resolve’s: nothing. Drop a mismatched clip on a sequence and it keeps playing at its own native rate, redistributed across the sequence timing, with real-time speed unchanged. To force a clip to the sequence rate and accept the speed change, right-click it: Modify, Interpret Footage. That dialog is Premiere’s version of Resolve’s conform. The hand-off is where this bites: mixed-rate sequences sent to Resolve by XML come out mismatched, so conform everything to one rate before the export.
Final Cut Pro
FCP sets the project frame rate when you create the project. You can change it afterward, but Apple’s own warning is that changing the frame rate can shift every edit point in the project in time, so pick correctly before the first clip lands. Mismatched clips get one of five selectable rate-conform methods, set per clip, which Apple ranks by quality: Floor, the default, duplicates frames or rounds down to the nearest whole frame and renders fastest; Nearest Frame rounds up or down to the nearest whole frame instead; Frame Blending mixes adjacent frames for smoother motion at the cost of some softness; Optical Flow interpolates new frames from motion vectors, looks better, renders slower, and can fall apart on complex motion; Machine Learning, rated best and available only on Apple silicon Macs, builds its in-between frames with a bidirectional optical flow model. Apple’s advice is also the practical one: shoot matched or cleanly multiplied rates (24 with 48, 30 with 60) and the conform never has to work hard.
If your source mess goes past frame rates into codecs and playback, that’s a different page: our proxy workflow guide covers what each of these NLEs does with heavy media, mixed or not.
Delivery norms by region
Broadcast families still follow the old analog color systems, decades after the transmitters went digital: 1000/1001 countries stay fractional, 50 Hz countries stay 25/50.
| Region | Family | Standard rates | Notes |
|---|---|---|---|
| US, Canada, Mexico, Japan, South Korea, Taiwan, Philippines | NTSC-derived | 29.97 (59.94i interlaced legacy); 59.94p for HFR | The whole family is 1000/1001 fractional |
| UK, Europe, Australia, NZ, most of Asia and Africa | PAL/SECAM | 25 (50i legacy); 50p for HFR | Clean integers tied to 50 Hz mains |
| Streaming (Netflix and friends) | Native rate | Accepts 23.976, 24, 25, 29.97, 59.94 | Deliver the project’s native rate, not a broadcast conversion |
| Digital cinema (DCP) | DCI | 24.000 standard; HFR variants exist | True 24, not 23.976 |
The streaming row deserves its own sentence: Netflix’s delivery spec requires native frame rate. Whatever rate the project was shot and finished at is the rate you deliver, so a 23.976 doc goes out at 23.976 and a 25 fps UK series goes out at 25, with none of the tape era’s pulldown or speedup conversions. Native-rate masters are why the conform table above matters less for streaming than it did for broadcast… and why it still matters at the last mile, because a festival DCP is still true 24. When that last mile arrives, here’s how to export a feature properly.
In conclusion
And that’s it! Eight rates, two families, and one 1953 committee decision to blame for the semicolons. If a number here doesn’t match a delivery sheet on your desk, leave a comment and I’ll dig into it. These reference pages get updated as the standards move. This one, mercifully, moves slower than most.
