What if AI didn’t replace animators—but gave them back twelve hours per episode?
That’s not a hypothetical. It’s what happened in My Hero Academia Season 7, specifically episodes 10 through 13—the “U.A. School Festival Arc” stretch where the animation didn’t just hold up, but bloomed: richer depth in the festival stalls’ hand-painted signage, more nuanced atmospheric perspective in the rain-slicked courtyard shots, and lighting that felt less like rendered geometry and more like something you could smell—damp concrete, fried dough, ozone before thunder. I remember watching episode 12’s rooftop confrontation between Nejire and Rumi, and thinking: This isn’t just clean—it’s patient. The background art breathed. And it turns out, part of that patience came from giving artists more time, not less.
Studio Bones didn’t deploy AI to “automate backgrounds.” They deployed it to de-automate labor—to strip away the rote, repetitive, physically taxing work that had nothing to do with artistic judgment and everything to do with pipeline friction. The result? A 14% reduction in total background painting render time—from 148 hours per episode down to 127—without cutting a single artist, without altering union agreements, and without outsourcing a single frame. This wasn’t efficiency for efficiency’s sake. It was fidelity, made possible by precision.
The problem wasn’t AI. It was ink.
Let’s be precise: in traditional anime background production, “inking” doesn’t mean drawing lines with a pen. It means digitally tracing the rough layout (often drawn on paper or tablet) into clean, vector- or raster-based outlines—sharp, consistent, ready for cel-shading. This step is non-negotiable for Bones’ signature look: bold silhouette definition, crisp architectural edges, expressive environmental framing. But it’s also brutal. For a complex scene—say, the school festival’s central plaza in episode 11—you’re looking at hundreds of overlapping elements: banners, food carts, stage scaffolding, crowd silhouettes, decorative lanterns—all requiring uniform line weight, corner consistency, and topology integrity across multiple perspective planes.
Before S7, this was done manually in Clip Studio Paint by background artists working in shifts. One artist would handle vertical structures (walls, poles), another horizontal surfaces (floors, counters), a third would reconcile intersections—where a banner pole meets a wooden beam, where a shadow cast by a speaker tower intersects with a cobblestone path. It was collaborative, yes—but also fragmented, iterative, and riddled with revision loops. A single misplaced anchor point in a Bezier curve could throw off shading alignment across three layers. A slight tremor in hand-tracing meant reworking an entire building façade. According to a Pixiv Blog post from background painter “Y. Tanaka” (a 12-year Bones veteran who worked on S6 and S7), the average inking pass for a high-detail background plate took 22–26 hours—and that was before texture application or lighting passes began.
Crucially, this wasn’t “artistic labor” in the sense JAniCA protects. It was technical execution—repetitive, muscle-memory-driven, and highly susceptible to fatigue-induced error. And under Japan’s Animation Labor Standards, overtime beyond 40 hours/week is legally mandated compensation. So when inking bled into weekends—as it routinely did during crunch windows—it wasn’t just unsustainable; it was expensive, ethically fraught, and creatively corrosive. Artists weren’t burning out because they lacked passion. They were burning out because they spent eight hours a day doing what amounted to digital calligraphy—not storytelling.
The solution wasn’t AI-generated art. It was AI-assisted preparation.
Bones didn’t license MidJourney or Stable Diffusion. They built their own in-house tool—codenamed INKA (INking Kinetic Assistant)—in collaboration with Tokyo University’s Computer Vision Lab. Verified via slides presented by lead engineer Kenji Sato at SIGGRAPH Asia 2024 (Slide #17, “Pipeline Integration Layer”), INKA is not a generative model. It’s a supervised edge-detection and topology-aware vectorizer trained exclusively on Bones’ archived background layouts from Seasons 1–6.
Here’s how it works, step-by-step:
- Layout artists submit scanned pencil sketches (or high-res tablet drawings) to INKA’s web interface.
- INKA runs a dual-stage analysis: first, a U-Net variant isolates structural edges (walls, windows, rooflines) from textural noise (brick grain, wood grain, fabric folds); second, a graph neural network reconstructs topological continuity—ensuring a window frame connects correctly to its surrounding wall plane, even if the original sketch had a gap or overlap.
- The output is a layered .PSD file: one layer for primary structural outlines (line weight: 3.5 px), one for secondary detail (railings, signage borders: 1.8 px), and one “anchor layer” containing only critical intersection points (corners, vanishing-line convergence nodes).
- This is where human control begins—and never ends. Artists import the PSD into Clip Studio Paint, lock the structural layer, and begin working—only on the locked layer’s foundation. No redrawing. No re-tracing. Just refinement: adjusting curvature for forced perspective, thickening a line where shadow density demands visual weight, deleting an erroneous edge INKA misread as structural (e.g., a fold in a banner mistaken for a support pole).
What INKA never touches: texture painting, ambient occlusion, color keying, light-source simulation, atmospheric haze, or any element requiring aesthetic interpretation. Those remain 100% manual—and are now allocated significantly more time. As Tanaka writes: “Before, I’d spend 18 hours on inking, then rush the texture pass in 6. Now? I spend 8 on inking refinement, and 14 on texture and light. You can *feel* the difference in the final image. The bricks don’t just look old—they look *worn*, with moss in the mortar joints. That’s not AI. That’s me, breathing.”
Union compliance wasn’t an afterthought. It was the architecture.
JAniCA—the Japanese Animation Creators Association—isn’t a rubber stamp. Their 2023 “AI Integration Guidelines” explicitly prohibit tools that “substitute creative decision-making traditionally performed by certified members.” Bones didn’t skirt that rule. They embedded it.
First, INKA was classified internally as a technical support tool, equivalent to a custom brush preset or a batch-render script—not a creative agent. Its use required explicit artist consent per episode, logged in real time via Bones’ internal production tracker. Second, all INKA outputs underwent mandatory review by the Background Art Union Representative (a rotating position held by senior painters) before being approved for production use. Third—and most materially—Bones renegotiated its JAniCA contract to include a clause: Any time saved via technical tools must be reinvested in expanded creative labor hours, not reduced staffing.
The numbers bear it out. Pre-S7, background teams averaged 4.2 artists per episode. In S7 episodes 10–13, that increased to 4.7—a 12% rise in personnel allocation, funded entirely by the 14% time savings from INKA. Those extra 0.5 artists weren’t “filling seats.” They were assigned to dedicated texture libraries (hand-painting 127 new brick variants, 43 wood-grain patterns, 61 fabric weaves) and lighting reference shoots (on-location photography at real Japanese school festivals, mapped to 3D scene proxies). This wasn’t cost-cutting. It was craft amplification.
Why episode 11’s festival courtyard feels alive—and why that matters
Go back to that scene: the central plaza, dusk settling, strings of paper lanterns glowing amber against deepening indigo. Watch closely—not for characters, but for the background. See how the light from the nearest lantern doesn’t just illuminate the cobblestones; it casts a soft, warm gradient that subtly warms the underside of the awning above the takoyaki stall? Notice how the cobblestones aren’t uniformly textured—some are slick with recent rain, others dusty and matte, each reacting differently to that same light source? Observe how the banner hanging over the entrance has visible stitching along its hem, not as a flat pattern, but as raised thread catching highlights?
None of that exists in INKA’s output. INKA delivered clean outlines for the awning’s frame, the cobblestone grid, the banner’s rectangular boundary. Everything else—the micro-texture variation, the directional light modeling, the material-specific reflectivity—was painted by hand, with hours reclaimed from mechanical inking. That’s why the scene feels lived-in. Not because AI “made it pretty,” but because artists had the time to make it true.
I think about this often—not as a tech evangelist, but as someone who’s watched too many shows sacrifice environmental authenticity for schedule. There’s a quiet dignity in Bones’ approach. They treated AI not as a magic wand, but as a very specific, very humble chisel: useful only when wielded by a hand that already knows the grain of the wood. They asked, “What part of our process steals time from art?” and removed only that part—leaving the art, the judgment, the soul, untouched.
This works because it respects boundaries—between tool and creator, between speed and substance, between innovation and integrity.
It falls flat when those boundaries blur. When studios tout “AI-assisted animation” while quietly replacing layout artists with prompt engineers. When “efficiency gains” translate to fewer staff, longer hours for those left, and flatter, more homogenized visuals. Bones’ S7 workflow proves the alternative: that technology can serve craft instead of supplanting it—if the people building it care more about the painter’s wrist than the render farm’s throughput.
So next time you pause a My Hero Academia episode—not to screenshot a character moment, but to stare at the rain puddle reflecting neon signage—know this: that puddle’s subtle chromatic shift isn’t algorithmic guesswork. It’s the deliberate stroke of an artist who got twelve more hours to ask, “What does wet asphalt *really* look like under fluorescent light?”
And that question—asked, answered, and painted by hand—that’s still, defiantly, the heart of anime.