Sakurai on Creating Games: Graphics (20 Topics)

Masahiro Sakurai, the creator of the Kirby and Super Smash Bros. series, shares a wealth of game-development know-how on his YouTube channel “Masahiro Sakurai on Creating Games”. This article summarizes and restructures the content of that channel by topic.

That said, a summary is only an entry point. So much of the value — Sakurai’s own words, his real-world examples, his pacing, and the footage itself — can only be gotten from the videos. So please don’t stop at reading the article; I strongly encourage you to also watch the original videos embedded under each topic.

This category covers picture-making — from the idea of “painting light, not texture” to supervision and staging. Here we bring together the key points of all 3 videos in the “Graphics” category (20 individual topics), structured in 3 parts following the 3 source videos. Each part opens with its explanatory video, so please watch along.

Part 1: Paint light not texture, emphasize hitboxes, retouch supervision (#01–#06)

This category is about picture-making, in both 2D and 3D. Centered on what Sakurai finds himself “snagging on” in visual supervision, it covers points artists should watch practically.

https://www.youtube.com/watch?v=tcTGUopJduE

1. Don’t paint the texture, paint the light

What Sakurai often tells background artists is “don’t paint the texture; paint the light reflecting off it.” We recognize everything as “light reflected off an object.” It’s not as simple as a material being red so it’s red, blue so it’s blue; how it looks is determined by many factors.

For example, making a tree: just pasting a brown-trunk, green-leaf texture doesn’t look very nice. Colored light streams in, bringing overall light and shadow. A degree of hue shift deepens the impression, and light diffuses too. Add atmosphere and you start to feel the depth of the world. Told to “make a tree,” you don’t just draw one tree. The final output is to express the light in the environment.

2. Emphasize things with hitboxes

A key point he says he discusses every time when making stages across the Smash series: emphasize things with collision detection, and don’t emphasize the rest. Smash is an action game, so it’s obvious you should know where the floor is — but very important.

With 3D polygons and perspective, hitbox positions get hard to read. Smash’s backgrounds are quite elaborate, but if you put too much into picture-making where there’s no hitbox and it stands out, the crucial terrain hitboxes risk being lost. Make parts with collision detection clear, and lower the prominence of backgrounds without it. Doing this while keeping the look beautiful is hard.

Simply applying atmospheric perspective to the background to reduce its presence makes hitboxes clearer, but harms the look. Graphical beauty matters too. It’s all balance, but leaning a bit toward the game side is just right. Artists shouldn’t just draw pictures either; understanding the game’s function and drawing the appropriate picture matters.

3. Distinguish the major from the minor

Game-machine performance is lower than ordinary people imagine, and you can’t realize things without trimming. At such times, what matters is distinguishing the “major” from the “minor” — discerning where the effect is high and where it’s low.

On Smash’s screen, even if the stage is like a celestial sphere, what’s mostly seen is near the screen center. The camera’s front should be weighted more, prioritized over near terrain that tends to fall off-screen. Between fighter and stage, you’d think the fighter is prioritized, but fighters are often shown small, so what matters is how much of the whole screen they occupy (though fighters are used in close-up too, so they get other priorities).

Sakurai sometimes sees odd fixations from the stage team. If the deadline and performance fit, playful touches are welcome, but the car tires actually rotating in the background of the Inkling stage was too much of an “invisible fixation,” so he had it reduced. Allocate resources to the most effective places. Not just graphics — this is essential to games made with trimming.

4. Get the sense of scale right

A tuning Sakurai raises fairly often in graphics supervision is matching the sense of scale. A huge object and its plastic model (scale model) have the same color and shape. In other words, even with the same shape, the sense of size looks quite different. Usually it’s the direction of “something that should be big looks small.”

A scale model has high resolution, strong specular (gloss), and is less affected by ambient light, so the same material area looks monotonous. Making it normally in CG approaches the model-side look, so ingenuity to compensate is needed. Keep texture resolution modest, reduce the area and strength of gloss, match the normal map to the scale, and add grime to large things. It matters to work with the feeling of making a scene, not a model.

5. Retouch supervision

For graphics supervision, nothing beats actually showing the adjusted result. Sakurai introduces several examples of supervision images for backgrounds. He shows a retouched image — “the colors are drab, so make it like this” — to convey the direction.

It’s unrealistic for the director to prepare all retouch images, but each such retouch takes about 5 minutes of work, under 10. This is more effective than conveying it verbally. For a cave picture, say, copy a layer in Photoshop, raise the contrast, shift the hue a little, and erase where light hits — it becomes like a reflected glow and starts to look nice. Incidentally, Sakurai doesn’t use a pen tablet; he retouches with a mouse only. Supervision is only the direction. Polishing it to product level requires the artist to work hard from there.

6. Let’s look at the stages closely

Smash Ultimate has 115 base stages alone. Trying to keep 8 fighters at 60 frames per second on the Switch — a far-from-rich environment — yet the range is so wide it doesn’t seem like one game. Because it expresses the worlds of various games.

This installment is a showcase of footage letting you look closely at some of those stages. Having this much breadth contained in one game is rarely seen. It’s quite hard on the makers, but being able to survey many game contents is part of the fun. He says he proceeded without cutting corners, and enjoyed it.

Part 2: Pixel art, matching styles, the miniature effect (#07–#13)

From the basics of pixel art to ingenuity for unifying characters of various styles onto one screen, and the relation of picture and game, the practice of picture-making is dug into further.

https://www.youtube.com/watch?v=Xt7ELzq7-4w

7. A beginner’s talk on pixel art

Sakurai himself laid the pixels for the original Game Boy Kirby. He drew many enemy characters and backgrounds too, which were reused for the Famicom version. He lightly explains techniques for low-color pixel art.

On the Famicom and Game Boy, 16×16-dot sprites (the unit for moving art) are common. Kirby is 16 dots wide too. But when drawing a circle, 15 dots looks nicer than 16 — even though it’s shown small. The reason is you can take a center line (the exact middle line). Even in coarse-dot crafting/building games, an odd width is handy in various ways.

Even with only two tones, black and white, you can vary the texture. Just an outline with the highlight removed looks mesh-like; blacked out with only highlights drawn becomes an iron ball. With one more color, you can use shadow for more careful expression. Low-color pixel art is also charming and fun to make if you include the animation (aside: Kirby’s Adventure’s Kirby gets a bit fatter when he has a Copy Ability).

8. Matching styles

If you do game-making as a job, you must handle all kinds of work regardless of your strengths. Artists are the same: people who join the same team match the same art style. Someone good at realistic expression may keep drawing a fancy style. Drawing itself has many rivals too, so you get through on technique.

To match styles, an image board is often made. This renders the game’s background, worldview, characters, items, and so on into pictures. A skilled person leads in drawing it, and by deepening models and lighting based on it, you reduce gaps in skill and taste and steer toward the intended direction.

In Sakurai’s works, Smash’s Mario has denim stitching drawn on. This is ingenuity to match styles as much as possible in Smash, where realistic and comical mix. Without it, you risk an expression like “anime and live-action fighting.” The higher the machine’s expression, the higher the cost of matching styles, but since it’s necessary without exception, you just have to push through.

9. Like anime and live-action fighting

Smash has characters of all kinds of world settings, comical and realistic. Avoiding dissonance when they stand on the same field requires considering many things. Just matching the official image gives a picture that won’t come together at all. Sakurai lists five adjustments done for this.

① Make vivid reds and yellows pale so they take ambient light more easily. ② Add detail to comical types and lighten the weight of realistic ones to bridge the gap. ③ Tweak proportions from the original to reduce dissonance (Bayonetta and others differ quite a bit in proportions). ④ Properly refine Olimar’s suit, R.O.B., etc., to add detail. ⑤ For Donkey Kong and Yoshi’s “manga expression of closing the eyes,” Smash specially makes it close the eyelids. From the viewpoint of “if these characters really existed, this is how it’d be,” the detail is deepened.

10. Do graphics relate to a game’s fun?

You occasionally hear “graphical beauty has nothing to do with a game’s fun.” Is that really so? Gameplay and visuals seem separable, but Sakurai says gameplay clearly does set off the visuals.

A game’s background, bluntly put, is just polygons. Without hitboxes, it’s a prop you pass through. But its presence changes with the game’s content. For example, in a die-a-lot game like Dark Souls, on a first playthrough you’re extremely wary of your surroundings. You proceed carefully, watching for casual background motion and omens, so you keep an eye everywhere and the presence of a single background piece is set off. With the same graphics, an expression of Sonic dashing through at high speed would give a completely different impression.

Also, on Steam and the like, countless games line up as banners, and you won’t get played unless the visuals draw interest. In that sense too, it’s a multiplier with fun. The view that “visuals don’t matter” isn’t wrong, but balancing overall and raising both to a high level is best.

11. Various supervision

This installment generously shows real examples of the graphics supervision Sakurai actually does. In telework it’s basically text, but he sometimes directs with retouches or diagrams.

Introduced are many concrete supervisions — Pyra, Kazuya, the Mishima Dojo, Sephiroth. For example, Pyra’s walk: “put weight forward, improve the silhouette”; Pyra’s facial: “raise the corners of the mouth crisply”; Kazuya: “raise the eyes and brows so he looks angrier”; the moonlight reflection on the Mishima Dojo; the problem of fighters in shadow being hard to see… each a fine adjustment piled on (all fixed in the retail version).

“A picture is worth a thousand words” — he wanted to show, by example, what supervision even is. And Sakurai conveys “thanks and apologies” to the staff who fix things properly.

12. The miniature effect

The “something that should be big looks small” problem in graphics. Related to this, Sakurai explains how to make a miniature photo: ① prepare an overhead landscape photo, ② blur the top and bottom (depth of field), ③ raise the saturation, and if possible lower the blue a little (remove atmospheric perspective). This easily gives a miniature effect.

Conversely, putting these two in can make big things look small. Depth of field is prized for its atmosphere, but where to focus differs by person, so it’s better not to blur much beyond the intended part. Saturation should be held down, since the bigger the object, the more atmospheric perspective applies. It’s necessary to view objectively and be conscious of the final output.

13. Analog color unification

Something that has long troubled Sakurai in graphics checks is monitor variance. Each staffer’s monitor color differs a little. Even the same model from the same maker shifts in relative color by the surrounding lighting.

This may not concern many roles, but for artists and Sakurai supervising them it’s a life-or-death matter. Especially fatal if the checking monitor is off. There are devices that measure and adjust color, but since the game’s output destination is sometimes an ordinary TV, they’re unreliable. In the end, he makes some monitor visible to everyone and adjusts by eye.

Even so, the gap can be too big to fully close, and the product and the game actually being made can really be made in different expressions. It can normally happen on your TV too. At least within the same team, it’s good to build a mechanism that minimizes color differences. Note that at one company the moody yellow lighting cranked up the picture-making overall, so it’s recommended to use lighting with as little personality as possible.

Part 3: Final output, manga-like expression, design that gives identity (#14–#20)

This part centers on the pursuit of picture-making that “makes you feel something is there when it isn’t” — blur expression, final output, and manga-like expression.

https://www.youtube.com/watch?v=xH1InUOcYSU

14. Motion blur

Blur is the afterimage expression when something moves. An expression filling the gap between one frame and the next, it reduces the frame-by-frame feel and helps dynamic expression. Kirby Air Ride’s Wheel ability and Smash’s rails have blur added so they look natural while advancing.

Wind-streak expression is a kind of blur too, but a little different in nature, applied not only to characters but to the background when you move the camera. To express blur in a game, there’s a hand-drawn way and a hardware-computed way, but beware overdoing either. Artificial blur, for now, strengthens dissonance, and risks making the game hazy and hard to read. Opinions tend to split among staff too, so it’s good to provide adjustment options to tune to taste. The more natural the blur, the less dissonance, so we look forward to the development of technology.

15. Make you feel something is there when it isn’t

There are works with graphics that seem prettier than reality, but from the maker’s side the game world is a prop, a set. With no mass; without hitboxes the terrain slips right through; without programming it doesn’t even fall. Just a combination of CG polygons, a dry thing with no mass.

That you don’t think so during the game is because it takes on reality through many elements beyond visuals — sound, programming, parameters. Making you feel something is there when it really isn’t — this is comprehensive “expression.” A graphic designer, even when making a table or props, shouldn’t think they’re just making the same thing, but control even the placement, accounting for texture, how it looks under the light source, sense of volume, and sense of life. Survey from a pulled-back view, and make you feel something is there when it isn’t. Note that easily overlooked is airflow. In Smash there’s ingenuity where things sway in environmental wind, or a slight wind arises near an explosion.

16. Final output

On powerful machines, it’s become an era where final output especially talks. The picture a console generates doesn’t appear on screen right away; it’s first stored in a frame buffer, where background, characters, and UI are drawn, and through shaders and rendering, once everything is drawn, it’s sent to the screen. How you wield the post-processing at that point matters.

Smash Melee has quite distinct object colors, but Smash Ultimate adds color tones from ground reflection and the like, improving the final output. The era of polygon characters merely being pretty is over; the final output matters. Eyes tend to go to fine details, but pulling the viewpoint back and having a coordinator of the image’s final output control the post-processing greatly raises quality. “I’d love to hire such a specialist if there were one,” he says.

17. Down to a single wall break

The wall break on the Mishima Dojo stage was supervision Sakurai especially struggled with. Small in scale, yet it just wouldn’t feel good. From the remaining supervision images, that trial and error shows.

He retouched the first submitted image, then piled on requests one by one: “the board’s cross-section looks cut by a coping saw,” “make it jagged front-to-back, not perpendicular to the board,” “the jags are regular, so make them irregular,” “scatter the three regularly-remaining pillars”… stacking detailed requests one by one. Even so, problems remained — jags rounded or regular — and only in the retail version did it finally reach the best expression.

A key attitude Sakurai shares here: make it OK even if it strays from your image, as long as quality holds. And no matter how much you supervise, the image gap or quality assurance sometimes falls short, and then you have to cut your losses and ship it as-is. “The goal isn’t to make everything ideal, so I cut it off surprisingly readily. But I do regret it.” Having them swing big and then reining it back via supervision tends to come together better, he also says.

18. Manga-like expression

Japanese manga and anime have many interesting expressions — eyes visible through the side frame of glasses, a mouth poking out to the side even in profile… These are applicable to games too. The profile mouth poking sideways in a Mega Man Legends event, or Sonic’s mouth moving to the side. Strange when you think about it, but good as character expression.

For manga expression, you sometimes deliberately bring on a tough CG structure. The Tom and Jerry-like expression of squeezing through a narrow space into a wide one is quite hard to do in CG (using shape animation and such). But Sakurai says he doesn’t think manga-like expression should be avoided just because it isn’t realistic; rather he’d recommend it. More such fun expressions is welcome. He also introduces directing the mouth models of Kirby and Jigglypuff, which had old expressions (whisker-like cross-sections), to “be like this,” and refining the Mii Fighter Nakoruru to feel true to the original.

19. Design that gives identity

What’s the secret to making an imagined picture look cool? There are various elements, but what Sakurai weighs most is the silhouette.

Make a silhouette whose outline works even when blacked out. If it’s muddled and unreadable, or indistinguishable from other characters, it’s lacking. The same goes for poses: even the same character looks different by the tendency of poses it favors. You should always be conscious of the character in everything it does. With the person in charge understanding the backbone, sharpen it further. It’s not something one person can finish; it’s an element the whole staff sets off together.

20. Various supervision: Sora edition

This installment generously shows supervision images Sakurai actually did, centered on the Smash Ultimate fighter Sora (Kingdom Hearts).

“Silhouette tuning to make the back jump look livelier,” “fixing the misalignment of the landing stance,” “making the attack follow-through plausible from the side,” “the hit-effect stars were small and clustered, so fling them back,” “shift the sea color in Hollow Bastion toward green to look more sea-like,” “the same material should be strong in front, weak in back in contrast (it was reversed, so he pointed it out)”… countless fine adjustments are piled on (all fixed in the retail version).

Game creation needs this repetition of steady adjustment. Polishing without giving up raises quality. “I ask the staff for a lot of things, sorry — but I’m probably not saying anything wrong, so I’d be glad if you stick with me persistently,” he closes, with gratitude for the relationship of trust.

Summary

We’ve now gone through all 3 videos and 20 topics of the “Graphics” category. Finally, let’s recap the key points of each part (= each source summary video) by individual topic.

Part 1: Paint light not texture, emphasize hitboxes, retouch supervision (#01–#06)

Part 2: Pixel art, matching styles, the miniature effect (#07–#13)

Part 3: Final output, manga-like expression, design that gives identity (#14–#20)

There’s a lot here, but I hope you’ll revisit it starting from whichever part caught your interest. Please also check out the related categories.