Why Silksong Stayed on Unity: The Boring Engine Behind Hornet's Faster, Bigger Sequel

Hollow Knight: Silksong runs on Unity. Not Unreal Engine 5, not a custom C++ engine, not Godot. The infobox answer is one word — Unity — for a 2025 Metroidvania that launched on September 4, 2025 across nine targets: Linux, macOS, Windows, Nintendo Switch, Switch 2, PlayStation 4, PlayStation 5, Xbox One and Xbox Series X/S.

That is the short answer to "what engine was Silksong made in" and "what game engine does Silksong use." Team Cherry, the same tiny Adelaide team behind the original, kept the same 2D hand-crafted stack and spent seven years on content instead of engine migration. The result sold over seven million copies by December 2025 and drew acclaim for graphics, soundtrack, story and world design, with debate focused on difficulty rather than tech.

Most summaries stop at "indie team uses Unity because it is easy." That misses why staying was the harder, smarter bet in 2019-2025, when every mid-size studio was being told to switch to Unreal for Nanite and Lumen. My read after checking Steam, the official site, Wikipedia and Unity history: Silksong proves that for hand-drawn 2D, the engine was already solved in 2017. The sequel's risk was never shaders. It was animation budget, tool iteration and nine-platform shipping — exactly what Unity already gave Team Cherry.

The team and the constraint

Team Cherry developed and published Silksong itself. Directors and writers Ari Gibson and William Pellen returned, with Pellen and Jasmine Vine on programming and Gibson on art, and Christopher Larkin on music. Wikipedia's Silksong overview lists that four-name core, the Unity engine credit and the full platform list.

The project started as downloadable content for Hollow Knight — a stretch goal to make Hornet playable from the 2014 Kickstarter — then grew in scope until Team Cherry announced in February 2019 it would be a standalone sequel. After a delay past a 2023 window, it shipped September 4, 2025. During those seven years Team Cherry said very little, which fed heavy speculation and a meme culture around the wait.

Staffing matters here. The original Hollow Knight was three people with PlayMaker visual scripting carrying enemies. Silksong stayed tiny while the design grew: a new kingdom Pharloom, a faster protagonist, quests, tools, crests and silk skills. No engine team was hired. Every month spent evaluating Unreal or rewriting for a new renderer would have been a month Pharloom did not get bigger. Staying on Unity was scope protection, not nostalgia.

By December 2025 the game had passed seven million copies, with a free Sea of Sorrow expansion planned for 2026. The same overview carries the sales figure and the expansion plan alongside the mixed verdict on high difficulty — critics praised the world, players argued about the spike.

What Unity actually did for Silksong

Forget Nanite-versus-Unity discourse. Silksong uses Unity as a 2D delivery truck with cheap tricks, same as its predecessor:

• Hand-crafted 2D, not 3D showcase. The official site describes Pharloom as "brought vividly to life in a traditional, hand-crafted, 2D style" with gilded cities, lakes of fire and misted moors plus a vibrant orchestral score. The official Silksong site sells that look directly, and links its methods write-up to Unity. No complex lighting rig is needed when scanned PNGs and transparent shapes already sell depth.
• Nine-platform day one. Windows, macOS and Linux for PC, plus Switch, Switch 2, PS4, PS5, Xbox One and Series X/S. Unity is a cross-platform engine with C# scripting that became popular precisely for letting small teams ship everywhere. Unity's engine history traces that democratization pitch from its 2005 Mac debut through mobile and console dominance — context for why a micro-studio would double down on it.
• Content density over shader density. Steam lists the scale: explore mossy grottos and gilded cities, master acrobatic combat, craft weapons and traps, solve shocking quests, face over 200 ferocious foes and vanquish over 40 legendary bosses, plus Steel Soul mode. Silksong on Steam reads like a checklist of things that must feel precise, not things that must ray-trace.

That is Unity's real payoff: a four-person core shipping hand-drawn density to seven-plus targets without rewriting. The site even keeps the old Unity methods link at the bottom — learn about the methods used to make Hollow Knight, and about the engine under-pinning the whole game — a quiet admission the sequel did not reinvent the base.

What changed: Hornet is not the Knight

Silksong's gameplay is similar to Hollow Knight — 2D platforming and combat — but Hornet moves more acrobatically and heals faster, with three Masks restored in a single heal. Side-quests called Wishes are tracked in menus. Equippable Tools split into blue passive, yellow support and red offensive weapons with limited charges refilled with Shell Shards at Benches. Crests change tool capacity plus attack speed and range, each with its own mechanic. Silk builds as Hornet hits, then pays for heals or Silk Skills. Benches still heal and respawn.

Movement is where Unity's 2D physics gets tested: Clawline grappling-hook swings, Drifter's Cloak wind-catch glides and slow falls, Cling Grip wall climbs and Silk Soar vertical needle leaps. These are momentum toys, not shader toys. They live in C# update loops and hand-tuned colliders, the same layer the original tuned for Mega Man X-style air control.

In other words, the sequel tax was animation states and level interlock, not engine features. Pharloom is taller and more vertical than Hallownest by design — ascend on pilgrimage to the citadel peak — so every new move had to work in 200 enemy encounters and 40 boss arenas without breaking readability.

What staying cost

Staying was not free:

1. No procedural escape hatch. Hand-placed Pharloom had to be hand-built. Seven years for a DLC-turned-sequel is the receipt. Unity does not generate a kingdom for you.
2. Difficulty as engine debt. When every boss is a hand-tuned state machine, balancing 40+ bosses plus Tools and Crests is combinatorial. The mixed reception on difficulty is partly a content problem Unity cannot solve.
3. Engine-business risk. Unity's 2023 runtime-fee proposal rattled every Unity indie, Team Cherry included. Unity canceled it in September 2024, but the scare is a reminder: betting on proprietary Unity plus Asset Store dependencies means your roadmap includes other companies' pricing.
4. Cheat-friendly by construction. Search suggestions show a second meaning of "engine" — "silksong cheat engine," "does cheat engine work on silksong." Unity Mono/C# games are famously easy to inspect and patch, which fuels trainers and cheat tables. That is the flip side of an accessible stack: easy to ship, easy to poke.

None of those killed the game. All of them shaped the seven-year wait.

My take: boring engine as editorial filter, not shortcut

Here is where I diverge from "Unity is for indies who can't code."

1. Staying was an anti-portfolio decision. From 2019 to 2025 the fashionable move was to announce an Unreal Engine 5 migration for Nanite, Lumen and World Partition. Team Cherry refused the demo. I think that was correct because Silksong's beauty is animation count, not polygon count. A Hornet sprint cycle read at 60 fps beats a photogrammetry rock at 30 fps. Unity already did 2D sprites, sprite packing, 2D physics and multi-platform builds well enough. Switching would have bought prettier screenshots and delayed Pharloom by years. Boring tech funded interesting traversal.

2. Silk is a UI design, not a graphics design. Tools, Crests, Silk spool and Wishes are loadout and quest UI problems. Unity's Canvas UI plus ScriptableObjects make that iteration cheap: designers can rewire a Crest's tool slots without touching the renderer. A custom engine team would have spent months rebuilding editors Team Cherry already had. If you want the lesson, copy the split: C# for continuous feel (acceleration, coyote time, Clawline release windows), visual graphs or scriptable data for discrete choices (Tool loadouts, Crest perks). Tutorials that pit engines against each other miss that the hard part was content tooling.

3. Nine platforms is the feature. Solo and micro teams obsess over frame-rate graphs. Team Cherry obsessed over port count. Linux/macOS/Windows plus last-gen and current-gen consoles plus Switch and Switch 2 on day one is something no hand-rolled engine delivers with four people. Unity's boring build pipeline is the moat. When players ask why Silksong is only a few gigabytes while UE5 games are 100GB, the answer is PNGs plus simple shaders plus no lightmaps — a distribution advantage that matters more on Switch storage than any Lumen bounce.

4. Silence was scope confidence, not tech trouble. The popular theory was engine trouble or Switch 2 contract silence. I read it differently: because the engine was stable, Team Cherry could afford silence. There was no "we migrated to Unity 6, please wait" post to write. The delay was design sprawl — DLC to full sequel, Hornet speed breaking old level metrics, 200 foes needing tells — not shader compilation. If they had switched engines mid-way, we would still be waiting.

One heresy for your next jam: visual fidelity did not make Silksong anticipated for seven years. Readability at speed did. Hornet is faster than the Knight, heals faster, and carries more verbs. Making a faster character still readable on Switch handheld is harder than making a slower character prettier on PS5. Unity did not make that easy. It made it possible to attempt without hiring a graphics team.

What indie devs can steal this month

• Lock Hornet-feel in code week one. No-accel horizontal, generous air control, dash-cancelable jumps, Clawline release just past the bottom on the upswing. Get that in C# before designers add Tools.
• Prototype bosses as named states on paper. Idle / Telegraph / Active / Recover. If you cannot sketch it in four boxes, your graph will be spaghetti at boss 30.
• Budget graph review like code review. Shared templates for stagger, stun and death, one owner per boss graph. Forty bosses will find every unnamed transition.
• Use cheap art tech to buy level time. Scanned PNGs plus sprite shaders plus transparent-shape lighting shipped "most beautiful hand-drawn game" quotes twice. You do not need a custom lit pipeline for mood.
• Plan ports on day one. Pick the engine that already ships where you will be in five years. Day-one Switch 2 plus PS4 plus Linux is Unity's free goodwill.

Do not copy the stack if you need procedural worlds, rollback netcode or a hiring pipeline of pure C++ engineers. Then the FSM and Mono tax outweighs the iteration dividend.

Train Hornet's verbs in a browser

Silksong demands three separate reflexes — needle direction reads, Clawline momentum, and feathered wind-glides. You cannot practice Pharloom without owning it, but you can drill each verb in 60-second loops:

• For directional slash-and-dodge discipline, play a few rounds of Iron Snout. Wolves come from both sides, weapons must be caught or deflected, and spamming gets punished — the same centered, pattern-reading patience Hornet's needle duels demand against fast Lace-style fencers. Its Classic endless waves and Sudden Death one-hit pressure map neatly to Steel Soul runs where one mistimed heal costs everything.

• For Clawline momentum, grind Stickman Hook. It is one-button grappling: hold to attach, release just past the bottom on the rising arc, chain without hesitation. That release window is the same intuition as Clawline chains and Silk Soar launches — look one anchor ahead, fix the mistake two swings earlier, never hold a good swing too long.

• For Drifter's Cloak feathering, load Space Waves. Hold to rise, release to fall through 33 handcrafted neon tunnels with green-to-nightmare difficulty faces. The winning method — micro-taps, recenter to mid-line, read two obstacles ahead, isolate the hinge section — is exactly how you survive Pharloom's wind shafts and spike corridors without panic-spiraling into ceiling-floor bounces.

None of these are Silksong. That is the point. They isolate one verb so your hands learn it before Pharloom tests all three at once.

Bottom line

Silksong's niche-plus-hot combo is why this question still ranks: a niche engine question (Unity versus Unreal, "what engine was Silksong made in," "what game engine does Silksong use") behind a massive Metroidvania (September 4, 2025, nine platforms, 7M+ by December, 200+ foes, 40+ bosses, Sea of Sorrow in 2026). Search data backs the curiosity — Google suggests "silksong engine," "silksong engine used," "silksong engine change" and "silksong engine game," while Bing adds "what engine was Silksong made in" and the cheat-engine polysemy — while Bing organic on full questions degrades to engine.com dictionary noise, which is why this piece leans on primary pages instead of SERP counting.

If you remember one thing: Team Cherry did not choose between art and tech. They bought boring tech so artists could author more animation, then spent seven years on a taller kingdom worth climbing. For a tiny team aiming at faster-than-Knight combat, that was not a compromise. It was the design.