Aspect ratios reduced exactly by the greatest common divisor, with the reduction multiplied back as a check, resizes that say whether they are exact, and named ratios that carry a note where the name is not the ratio.
Initializing in your browser…
Exact arithmetic on rationals, with both evaluation conventions named and the precedence table the parser runs on printed on the page.
Add and subtract durations with the sign kept, so one hour minus two hours is minus one hour. Also moves a time of day forward or back and says how many midnights it crossed.
Add, subtract, multiply, divide, compare and simplify fractions of any size, exactly. Reads mixed numbers and repeating decimals, and gives the decimal expansion with its repeating cycle.
A 1920 by 1080 frame. What is the reduced ratio, and how do you know it is right?
Input
1920 x 1080
Result
16:9 checked: 16 x 120 = 1920 and 9 x 120 = 1080
A reduced ratio has a check that needs nothing external: the reduced ratio times the divisor has to give back the original, and that multiplication is printed. The previous version of this tool could not pass it, because it scaled both sides by a thousand to handle decimal inputs and then divided by the scaled divisor AND by a thousand again; measured in a browser, 1920 by 1080 printed as 0.02:0.01 and 800 by 600 printed as 0:0. The reduction is now a greatest common divisor over whole numbers, exact at any size, with decimals scaled before the division rather than after. 2560 by 1080 comes out as 64:27, which is what it is, rather than as the 21:9 no panel actually is.
The greatest common divisor over whole numbers, so 1920 by 1080 is 16:9 and 2560 by 1080 is 64:27. The reduced ratio times the divisor has to give back what you typed, and the page does that multiplication and shows it.
A reduced aspect ratio has an unarguable right answer and a check that needs nothing external at all: **the reduced ratio times the divisor must give back the original**. That multiplication is on the page.
The reduction is a greatest common divisor over whole numbers, so it is exact at any size. A decimal input like 2.35 or 1.85 is scaled to whole numbers BEFORE the division rather than after, which is where the previous version of this tool went wrong: it multiplied both sides by a thousand to handle decimals, ran Euclid on the scaled values, and then divided by that scaled divisor and by a thousand again. Every reduced ratio came out a thousandth of the right number, and rounding it to two places for display turned most of them into zero. Measured in a browser on 2026-09-06, 1920 by 1080 printed as 0.02:0.01 and 800 by 600 printed as 0:0.
**A resize is usually not exact, and the page says which it is.** A pixel is a whole thing: 1000 pixels wide at 16:9 is 562.5 high, and both 562 and 563 are wrong. It rounds to the nearer, then reports the ratio that really came out and how many pixels off the mark it landed, rather than printing a whole number as though it were exact. When the result IS exact, it says that too.
**Several named ratios are not the ratios they are named after.** No panel sold as 21:9 is 21:9: 2560x1080 and 3440x1440 are 64:27 (2.370) and 3840x1600 is 12:5 (2.400). CinemaScope has been 2.39:1 since 1970, and a film described as 2.35 today is usually 2.39. Facebook quotes 1.91:1 for a link card but recommends 1200x630, which is 40:21 (1.9048). And the ISO 216 A series is one to the square root of two, which is IRRATIONAL, so no whole-number ratio is it exactly. Every preset here carries that note, and every preset sets a size that really is the ratio it names, rather than forcing a fixed width and rounding: the previous version set 1920 by 823 for its 21:9 button, which is 2.3329.
Alongside that: the exact fraction, the decimal to four places, the pixel count in megapixels, the closest named ratio with how far away it is, a shape drawn to the real proportions, and a list of the sizes at which each named ratio lands exactly.
Reduces to 16:9, with a divisor of 120. The check shown is 16 x 120 = 1920 and 9 x 120 = 1080.
Reduces to 64:27, not 21:9. 64:27 is 16:9 cubed, which is why it tiles with 16:9 content; no ultrawide panel made is actually 21:9.
The exact height is 562.5. Rounded to 563, the result is 1.776199 against 1.777778, which is a quarter of a pixel out, and the page says so.
Lock one side and read the other, with the rounding stated.
Every named ratio lists the sizes at which it lands exactly.
Type its dimensions and read the reduced ratio and the nearest named one.
Because that is what it is. 2560 and 1080 have a greatest common divisor of 40, giving 64:27, which is 2.370. 21:9 is 2.333. The name is a marketing round number that no panel actually is, and 64:27 happens to be 16:9 cubed, which is why 16:9 content tiles into it neatly.
No. CinemaScope prints released between 1958 and 1970 are 2.35:1; the SMPTE standard has been 2.39:1 since 1970, and a modern film described as 2.35 is usually 2.39. Both are in the preset list with that note.
Because a pixel is a whole thing and most target widths do not divide evenly. At 16:9, a width of 1000 wants a height of 562.5, and neither 562 nor 563 is on the ratio. The tool rounds to the nearer and reports the ratio it really produced along with how far off it is, instead of printing a whole number as though it landed on the mark.
Yes. 2.35 and 1 works, and so does 1.85 and 1. Both sides are brought to the same number of decimal places and scaled to whole numbers before anything is divided, so the reduction stays exact.
The reduced ratio multiplied back by the divisor. If that does not give what you typed, the reduction is wrong. It needs nothing external, so it runs on every calculation and the result is printed rather than assumed.
Because ISO 216 defines the A series as one to the square root of two, so that halving a sheet gives the same shape. The square root of two is irrational, so no ratio of whole numbers is it exactly. 1000:1414 is the nearest four-figure approximation and is about 0.0015 percent off.
Every calculation runs locally in your browser. Your numbers and expressions are not transmitted or stored.