The Emoji Google's 3D Redesign Changed the Most, Ranked

The green square 🟩. Compared pixel by pixel with Google's own flat artwork, 76.9% of it changed in the 3D redesign, more than any other of the 1,661 emoji we measured. Next come sunrise 🌅 at 73.9%, the yen chart 💹 at 73.1% and the purple square 🟪 at 71.0%. Faces, the obvious guess, sit lower. Symbols moved most, objects least, and not one of the 3,740 image files that exist in both sets is byte-for-byte identical.

This post is the ranking. If you want the background on the redesign itself, which phones have it and how to get the old look back, that is covered in our Noto 3D explainer. Everything below is a number we measured on 22 October 2026, and every description of what an image looks like comes from opening both versions side by side.

How we scored each emoji

Google keeps both designs in its public googlefonts/noto-emoji repository, the flat set in a 2D folder and the new one in a 3D folder. We pinned one commit (e20cbc2, dated 24 September 2026) so every figure here points at the exact files measured.

For each emoji we took both images, shrank them to 64 by 64 pixels on a white background, and compared them square by square. We add up how far the three color channels moved at each pixel, giving a number from 0 to 765, and call the pixel changed once that total passes 96. That is enough to ignore soft edge noise and still catch a recolored tile or a missing outline.

We recorded three things per emoji:

• Change: the share of pixels that visibly differ.
• Ink: how much of the square the drawing covers, meaning any pixel visibly darker than white.
• Brightness: the average lightness of the drawing's own pixels, from 0 to 255. A shift of more than 10 counts as darker or lighter.

Scope was every distinct emoji on Unicode's own list, without skin tone variants (they would just repeat their base). Country flags were left out because the old 2D folder had no flag artwork to compare against; the 3D folder draws 259 of them. That leaves 1,661 emoji, with a median change of 20.7%.

The 12 emoji that changed the most

1. 🟩 green square, 76.9%
2. 🌅 sunrise, 73.9%
3. 💹 chart increasing with yen, 73.1%
4. 🟪 purple square, 71.0%
5. ✳️ eight-spoked asterisk, 67.2%
6. ⚧️ transgender symbol, 66.4%
7. ♀️ female sign, 65.8%
8. ✅ check mark button, 65.4%
9. ✝️ latin cross, 65.3%
10. ❇️ sparkle, 64.7%
11. ℹ️ information, 64.7%
12. 🆔 ID button, 64.2%

Ten of those twelve are symbols, and the pattern holds further down: 23 of the top 25 and 87 of the top 100 are in Unicode's Symbols group. Only 31 emoji crossed 60%, and none reached 80%.

Sunrise is the odd one out at number two, and it earned the spot honestly. The flat version was a pale pastel sky with sunbeams over light blue water. The 3D one is a dark, saturated orange and red sky with the sun sitting low on the horizon. Same subject, nearly every pixel different.

Google changed two colors, and left the rest alone

The most useful finding in the whole measurement is hiding in the colored squares. Look at all nine:

• 🟩 green 76.9% and 🟪 purple 71.0%
• 🟥 red 14.8%, 🟦 blue 11.3%, 🟫 brown 11.1%, 🟧 orange 9.9%, ⬛ black 9.5%, ⬜ white 7.4%, 🟨 yellow 6.7%

Seven of them barely moved. Two were repainted. Opening the files shows why: the old green was an olive, slightly grey green, and the new one is a bright, saturated grass green. The old purple was a muted mauve; the new one is a vivid violet.

The circles say the same thing. 🟢 changed 59.6% and 🟣 52.1%, while the other seven circles stayed between 6.0% and 10.0%. Among the hearts, green heart 💚 changed 42.1% and purple 💜 23.7%, while the remaining ten colored hearts stayed between 2.7% and 8.0%.

That one palette decision explains a large chunk of the top of the ranking. The green buttons (✅ ✳️ ❇️ 💹 ❎) all changed between 57.4% and 73.1%. The religious symbols, which Google draws on purple tiles (☮️ ✝️ ☪️ 🕉️ ☸️ ✡️ ☦️ 🛐), all changed between 61.1% and 65.3%. Virgo ♍ changed 43.1% while the median zodiac sign changed only 11.0%, and the reason is plain once you see it: Virgo sits on a green circle.

So most of those "most redesigned" symbols were not redrawn at all. Their shapes are close to what they were. The tile behind them changed color, and a solid tile fills most of the square, so the score goes through the roof.

Symbols first, objects last: the category ranking

Here is the median change for each Unicode group, with the number of emoji measured in brackets:

• Symbols (224): 27.8%
• Smileys and Emotion (172): 24.5%
• People and Body (390): 22.5%
• Food and Drink (132): 20.8%
• Activities (85): 20.2%
• Animals and Nature (161): 20.1%
• Flags, non-country only (8): 19.2%
• Travel and Places (221): 17.2%
• Objects (268): 15.0%

Faces did change more than most things. A sample of 14 everyday faces we picked, including 😂 😭 🥹 😍 😅 and 😎, has a median of 30.4%, well above the overall 20.7%. But by pixel count the symbols moved further, and that is mostly the green and purple story above.

The biggest mover in every category

• Symbols: 🟩 green square, 76.9%
• Travel and Places: 🌅 sunrise, 73.9%
• Objects: 💹 chart increasing with yen, 73.1% (another green tile)
• People and Body: 🧗‍♀️ woman climbing, 53.8%, with the man and person versions at 53.2% and 53.0%
• Animals and Nature: 🐸 frog, 53.3%, from a flat lime face to a greener, glossier one
• Food and Drink: 🍍 pineapple, 52.5%, now tilted on its side with darker golden skin
• Activities: 🎿 skis, 50.9%, a blue boot on a single ski redrawn as a yellow boot in front of a crossed pair
• Smileys and Emotion: 💟 heart decoration, 45.1%, its magenta tile now violet
• Flags: 🏳️‍🌈 rainbow flag, 27.1%, drawn smaller and lighter

The climbers are the interesting case. Put the two versions next to each other and it is still a small figure on a tree trunk. The trunk got wider and moved in from the edge of the frame, which pushes the climber across, and almost nothing lines up any more. The skis are the opposite: a genuinely different picture, new color and new composition.

The gender signs grew the most in the frame

The three gender signs were bare symbols in the flat set. In 3D each sits on a rounded colored tile, purple for female sign ♀️, teal for male ♂️ and orange for transgender ⚧️, and each now fills about three quarters of the square:

• ♀️ went from 31.5% of the square to 77.8%
• ⚧️ went from 35.3% to 77.8%
• ♂️ went from 42.3% to 74.1%

Next on the list are a set of results you should not believe. 🪧 placard went from 17.1% coverage to 51.1%, 📅 calendar from 43.4% to 69.2%, and 🍽️ fork and knife with plate from 21.3% to 46.5%. Open the files and they are almost the same size. What changed is that their white surfaces now carry faint grey shading, and faint grey counts as "drawing" in an ink measure. The placard's change score, 3.1%, gives it away: if a drawing truly grew by a third of the square, far more than 3% of pixels would differ.

Across the whole set, 415 emoji covered more of the square by more than 2 points and 368 covered less. The median barely moved, from 52.2% to 52.5%, so this was not a set-wide zoom.

What got smaller

• 💨 dashing away, 58.1% to 33.3%
• 🫆 fingerprint, 65.0% to 41.8%
• 📛 name badge, 71.5% to 54.7%
• 💟 heart decoration, 77.4% to 62.0%

The fingerprint is the clearest real case. The flat version had a pale blue filled shape behind the ridges. The 3D version dropped it, leaving only the dark blue ridges, so it reads lighter and thinner. The fog emoji 🌫️ technically tops this list, but it is a pale grey cloud on a pale tile, exactly the kind of low-contrast drawing the ink measure handles worst, so we leave it out.

Lighter: the speech bubbles lost their outlines

Only four emoji got brighter by more than 100 points on the 0 to 255 scale, and all four are bubbles:

• 🗨️ left speech bubble, 67 to 207
• 💭 thought balloon, 69 to 201
• 💬 speech balloon, 78 to 198
• 🗯️ right anger bubble, 93 to 196

In the flat set each had a thick black outline around a white interior, so most of the "drawing" was black. In 3D the outline is gone and the bubble is a soft shaded white shape. 🗨️ is the single most lightened emoji in the set. On a dark mode chat background that is a bigger visual change than the score of 28% suggests.

Other notable brightenings: 🗄️ file cabinet (91 to 139) went from dark slate to light grey metal, and 😱 face screaming in fear (122 to 168) lost the deep blue band across the top of its head in favor of a lighter one.

Darker: the mouse, the buffalo and a lot of shadow

Darkening was far more common. 487 emoji got darker by more than 10 points, against 110 that got lighter. Real lighting adds shadow, and shadow mostly takes brightness away.

• 🖱️ computer mouse, 219 to 89: white mouse became a black one
• 🐃 water buffalo, 169 to 109: light grey animal became dark charcoal
• 🌅 sunrise, 198 to 142: pastel dawn became a deep orange sky
• 🌱 seedling, 186 to 147: pale yellow green became glossy dark green
• 🌮 taco, 191 to 153, and 🥟 dumpling, 220 to 182: both picked up toasted, browner shells

The mouse is the extreme. A 130 point drop is bigger than any other change in either direction except the bubbles. Another darkening worth knowing: 💀 skull went from a pale blue-grey to an ivory bone color with deeper eye sockets (204 to 174).

The emoji that barely changed

Every file was re-rendered, but some came through nearly untouched:

1. 🔸 small orange diamond, 0.7%
2. 👚 woman's clothes, 0.9%
3. 🔹 small blue diamond, 1.5%
4. 🚪 door, 1.7%
5. 🕯️ candle, 2.1%
6. 🚏 bus stop, 2.5%
7. 💛 yellow heart and 🤍 white heart, 2.7% each

Side by side, these really do look the same apart from a slightly softer highlight. The diamonds are tiny, which helps their score, but the door and the blouse are full-sized drawings that Google simply kept. In total 37 emoji changed less than 5%, and 225 changed less than 10%.

The quietest emoji in each other group: 🎟️ admission tickets (3.6%) in Activities, 🦩 flamingo (4.4%) in Animals, 🥖 baguette (3.6%) in Food, and 🛌 person in bed (3.9%) in People.

Where a pixel score misleads

A pixel comparison is a blunt tool, and three of its blind spots show up in this data.

First, position beats meaning. Move a drawing a few pixels sideways and the score climbs, as the climbers show. Repaint a background tile and it climbs even more, as the green and purple symbols show. Neither tells you whether the emoji now means something different.

Second, pale grey fools the ink measure. White objects that gained gentle shading look like they grew. We flag those cases rather than rank them.

Third, 64 by 64 is small. Fine detail inside a face, like a change in eyebrow angle, can move only a handful of pixels. That is the size where the emoji is actually seen in a message, which is why we chose it, but the subtle redraws are better judged by eye. For a design change that genuinely shifted a meaning, the cases in emoji redesigns that changed what they meant are a better guide than any percentage.

Checking the numbers yourself

The full measured section, with before and after images for the most and least changed emoji, lives on our Noto 3D update page. Our Google emoji page and Android emoji page show the current designs in context.

Color changes like the new green also matter when two emoji were already easy to confuse, since a recolor can push a pair closer together or further apart on one platform only. The emoji lookalikes page measures those pairs across vendors, and why emoji look different on iPhone and Android explains why the person you are texting may be seeing the old green, or Apple's, rather than Google's new one.