DEV Community

dev game
dev game

Posted on

Why Vampire Survivors Started on Phaser and JavaScript: The Web Stack Behind Bullet Heaven — And Why It Had to Leave

Vampire Survivors should not have been a Steam hit. A browser-first 2D framework, JavaScript garbage collection, and thousands of bats on screen is a recipe for lag — yet a tiny prototype built by unemployed developer Luca Galante became the defining bullet-heaven of 2022, won Best Game at the BAFTAs, and spawned a whole genre of Survivors-likes.

The short answer for developers is this: Galante built Vampire Survivors in 2020-2021 with Phaser, the 2D HTML5 game framework, shipped it as a free browser game and then as a cheap Steam Early Access title, and only later ported it to Unity at version 1.6 to fix performance and reach consoles. That two-act engine story is the lesson — but my take is different from most postmortems: the port was not proof that Phaser failed. It was proof that the right starter stack depends on where you intend to finish.

This updated analysis re-verifies the history in November 2026, adds Unity WebGL browser performance evidence, and shows what to do if your destination is the browser itself rather than consoles.

Who this is for and what is covered

You are an indie or solo developer choosing between a lightweight web stack and an industry-standard engine, or a technically curious player wondering why early Vampire Survivors felt janky late-game and then suddenly got smoother.

Scope as re-verified on November 19, 2026: original code used HTML5 and was released on itch.io as a free browser game, then on Steam for Windows under Early Access in December 2021, with 1.0 on October 20, 2022. Engine is listed as Phaser pre-version 1.6 and Unity from version 1.6 onwards. Full release platforms were macOS and Windows on October 20, 2022, Xbox One and Series X/S on November 10, 2022, Android and iOS on December 8, 2022, Switch on August 17, 2023, and PS4 and PS5 on August 29, 2024, with a Meta Quest VR port in November 2025. Search demand remains strong: Google suggestions still cluster around vampire survivors evolutions, wiki, evolution chart and characters, while phaser HTML5 game tutorial and unity webgl performance optimization return stable tutorial clusters. That is the niche-plus-popular signal — a browser library powering a BAFTA winner, plus evergreen web-performance questions.

What Phaser actually is — and how niche it is

Phaser is a fast, free, and fun open source HTML5 game framework that offers WebGL and Canvas rendering across desktop and mobile web browsers. It was first released in April 2013, is written in JavaScript, is available under the MIT License, and is developed by Photon Storm lead Richard Davey. It uses both Canvas and WebGL internally and can swap between them based on browser support, using the Pixi.js library for rendering.

Its own documentation is unusually honest about the niche: Phaser is an HTML5 framework designed specifically for web browsers, it is 2D only, you code in JavaScript or TypeScript, and it is not the right choice if you want to make a game fully in 3D, publish on PS5, Xbox or Switch, or avoid learning JavaScript. Native desktop and mobile builds are possible only via third-party wrappers such as Apache Cordova or Electron. Physics is built-in as Arcade Physics for high-speed AABB collision and Matter.js for full-body simulation.

How niche? It is commercially maintained by Phaser Studio Inc with an open-source community and is one of the most starred JavaScript game frameworks on GitHub, with thousands of shipped web games, YouTube Playables and jam prototypes. The minified full build is only around 345 KB gzipped, smaller than most texture images. Compare that to Unity or Unreal editors measured in gigabytes, with console SDKs and lightmappers. Phaser is a JavaScript library you include with npm install phaser, not an engine you install.

That is the contrast most coverage misses: a browser library for web toys powering a BAFTA Best Game winner.

Why JavaScript plus Phaser fit a one-person prototype

Verified history: Italian developer Luca poncle Galante began developing Vampire Survivors in 2020 using the Phaser framework while he was unemployed. Development of the initial Early Access version took about a year, during which Galante spent around £1,100 on assets, art and music. He was inspired by the mobile game Magic Survival from 2019 with auto-attacking combat, by Castlevania — one asset pack included sprites in a similar style to early Castlevania games — and by his prior work in the gambling industry, using flashy slot-machine graphics knowledge for chest-opening animations.

Galante released the first version on itch.io but saw little engagement, then released on Steam before the end of 2021 keeping cost low since the game had minimal art. In early 2022 YouTube creators picked it up and Steam sales spiked. Success exceeded expectations and let him quit his job in February 2022. In March 2022 he hired freelancers and said a major mid-2022 milestone would be porting to an industry-standard engine to improve performance. All of that arc is documented in the development history.

My reading of why Phaser was rational, not accidental:

  1. Zero fear of shipping. No license fee, no console NDA, no build farm. A browser page plus itch.io free upload means you can publish in an afternoon. For an unemployed dev, that low shame threshold matters more than frame time.

  2. Auto-attack removes what Phaser is bad at. Vampire Survivors is described on Steam as a time-survival game with minimalistic gameplay where hell is empty and all you can do is survive. Weapons attack automatically; you only move, collect XP gems, choose 1 of 3-4 weapons or passives per level up to 6 weapons plus 6 passives, open chests for evolutions, and try to survive 15, 20 or 30 minutes until The Reaper spawns. No precise aiming, no rollback netcode, no 3D culling — exactly the systems a Canvas/WebGL sprite batcher can do.

  3. Asset-flip aesthetics as feature. £1,100 of store sprites plus screen-filling numbers looks intentional when the fantasy is mow thousands of night creatures. Slot-machine chest pops, damage ticks and level-up pauses create juice without shaders or VFX graphs.

The ceiling: when thousands of enemies meet garbage collection

Early Phaser builds were notorious in player reports for late-run slowdown: hundreds of enemies, projectiles, XP gems and damage numbers alive at once, all as JavaScript objects with per-frame updates. JavaScript is single-threaded, garbage-collected, and Phaser 3's renderer was not built for 5,000 independently colliding bats. The official docs admit Phaser itself is not for consoles — yet poncle needed Xbox, Switch, PlayStation and mobile to grow.

The verified response was a rewrite: Phaser before version 1.6, Unity from version 1.6 onwards, with the March 2022 roadmap explicitly framing the port as an industry-standard move to improve overall performance. The 1.0 release on October 20, 2022 was already on the new foundation, enabling Xbox in November 2022, mobile in December 2022, Switch in August 2023 and PlayStation in August 2024.

In other words, Phaser bought speed to market; Unity bought headroom and certification.

Bullet heaven as an emergent property of constraints

Verified genre fact: Vampire Survivors received accolades at the Golden Joystick Awards, New York Game Awards and D.I.C.E. Awards, was nominated in five categories at the British Academy Games Awards winning Game Design and Best Game, is considered one of the first bullet-heaven games and led to numerous auto-combat clones.

My analysis — labeled as opinion: I think bullet heaven was not designed despite Phaser limits, but partly because of them.

If you can only cheaply render sprites, collide AABBs, and run logic in one JS thread, what is the most fun you can still deliver? Remove aiming, remove precision platforming, remove enemy AI pathfinding beyond seek-player, and pour all budget into quantity. Being the bullet hell — the supernatural indie phenomenon that lets you be the bullet hell — inverts the performance problem into fantasy. Lag becomes power fantasy up to a point, then it becomes a bug report. That knife-edge is the genre.

Provocation 1: Would Galante have invented auto-combat if he started in Unity with free NavMesh, Mecanim and particle systems? I doubt it. Abundance invites complexity. Constraint forced him to delete the stick-twirling and keep movement plus build choices. What would your prototype look like if you were forbidden from adding input verbs and only allowed to add enemies?

Provocation 2: Performance pain as a genre filter. The worst Phaser stutter happened exactly when the build was most fun — 20 minutes in, evolved garlic plus lightning rings melting 2,000 skeletons. Players forgave it because the fantasy justified it, then demanded the Unity port because they wanted more of it. A smoother engine at day one might have hidden that signal. Do we sometimes optimize too early and never learn where fun breaks?

My contrarian take: if your destination is the browser, do not follow the port

Most retrospectives treat Unity as the happy ending. I disagree for web developers. Galante left the web because he needed consoles, achievements, first-party saves and retail certification. If you do not need those, copying his exit is the wrong lesson.

Here is why, with current Unity Web evidence:

Unity provides support for developing games on the Web platform, where WebGL lets you publish real-time interactive 3D graphics into a browser. That sounds like a free upgrade from Phaser, but the same manual warns that on the CPU, Emscripten translates your code into WebAssembly, the performance of which depends on the browser you are using, that when using WebGL API for rendering, the CPU side dispatch of WebGL operations is slower than in native OpenGL, so avoid large numbers of draw calls per frame and use batching and instancing, and that standard C# jobs are largely limited while only native and Burst-compiled code multithreads well on Web. Setting Maximum Memory Size above 4 GB slightly reduces runtime performance, and exception stack traces have a cost unless you target WebAssembly 2023.

Translation: Unity WebGL reintroduces a heavy runtime, larger downloads, longer load times and stricter memory discipline to solve a problem — console reach — that a browser-only game does not have. Phaser, by contrast, stays a few hundred kilobytes, starts instantly, and its WebGL plus Canvas fallback across desktop and mobile browsers is already tuned for instant-play portals where players bounce in under 10 seconds.

My decision framework, offered as opinion:

  • Ship Phaser/HTML5 if your win condition is instant link sharing, itch.io virality, school Chromebooks, or ad-free portals like EasyHub.games, which runs hundreds of HTML5 and Unity WebGL games instantly with no downloads.
  • Plan a Unity rewrite only when you can name the trigger in advance — console request, 60fps collapse past X live enemies, or DLC entitlement pain — as Galante did in March 2022. If you cannot name the trigger, you do not have a plan, you have hope.
  • Do not assume Unity WebGL is automatically smoother than well-pooled Phaser. Thousands of pooled sprites with texture atlases and AABB-only collision will often beat a naive Unity WebGL build with hundreds of draw calls and unbatched UI canvases.

In short: Phaser was scaffolding for poncle because his building moved to consoles. If your building stays on the web, the scaffolding can be the house.

Feel the loop in 30 seconds: three browser trainers

Reading about auto-combat is abstract. These instant-play examples isolate one Vampire Survivors skill without an install:

  • For the timed-survival pressure, try 10 Minute Mage, an action game where you play a mage with the goal of defeating all enemies within 10 minutes, leveling up to gain new spells with unique abilities as harder waves arrive. It compresses the 15/20/30-minute Reaper clock into a lunch-break run and teaches the core Vampire Survivors discipline: keep moving, pick scaling spells, do not greed for gems.

  • For the seek-and-melt movement fantasy, try Apocalypse Reborn, a top-down survival game where you gather resources, build shelters and fight zombies, and where simply walking near objects lets you interact, build or attack zombies automatically. That walk-to-attack rule is the closest browser cousin to auto-weapons — positioning replaces aiming.

  • For the build-crafting half, try AutoRPG Arena, a turn-based auto-battler RPG where every choice influences the fight as you collect powerful items, upgrade heroes and adapt your squad. It has no twitch dodging, but it trains the same muscle as choosing 6 weapons plus 6 passives and chasing evolutions: read synergies, not moment-to-moment aim.

None of these replaces Vampire Survivors, and none proves engine superiority. They are useful because they run where Phaser shines — instant, low-friction browser sessions.

What to steal (and what not to copy)

Steal this if you are solo:

  • Choose web tech to kill fear. Free itch browser build, cheap Early Access, instant patch. Validate fun before paying for consoles.
  • Design around one-thread logic. One movement vector, auto-fire, discrete level-up pauses that mask GC hitches.
  • Make numbers the art. XP gems, gold, DPS ticks and chest flashes are cheaper than shaders and read better on streams.
  • Pool everything. Reuse enemy, projectile and gem objects, batch sprites into atlases, keep collision to AABBs and circles, and cap live damage numbers — the exact habits that delay the Phaser ceiling and also fix Unity WebGL draw-call bloat.
  • Plan the rewrite trigger in advance. Galante announced the industry-standard port when YouTube traction hit, not after burnout. Define your metric: concurrent sprites, crash rate, or console request?

Do not romanticize:

  • Phaser does not give you console certification, first-party saves, achievements or age ratings. poncle needed a team and Unity to ship Switch and PlayStation almost a year apart.
  • JavaScript dynamic typing and Electron packaging bite at scale — anti-cheat, DLC entitlements and cross-save across Steam, Xbox, iOS, Android and Switch are not web problems.
  • Auto-combat clones now flood Steam. Copying Vampire Survivors mechanics in 2026 without a new constraint just adds to Survivors-likes. Add a browser-native constraint instead: one-button input, 10-minute cap, or walk-to-attack.

Verdict

Vampire Survivors did not succeed because JavaScript is secretly better for games. It succeeded because a niche web stack let one unemployed developer spend a year and £1,100 testing a single heretical idea — what if you remove aiming? — in public, then abandon that stack when success demanded headroom. Phaser was scaffolding, Unity was foundation — for a console future.

If you are starting a 2D systems game in 2026 for the browser, ask the sharper Vampire Survivors question: what stack lets me ship this month, and what signal will tell me to rewrite? If your players will always arrive via link, the honest answer may be to stay on HTML5, pool ruthlessly, batch draw calls, and never pay the Unity WebGL tax at all.

What boring web constraint would make your game stranger, not just cheaper? I do not have a rigorous answer. I think living with that question for a prototype weekend is more useful than any engine benchmark.

Sources and verification date

Core facts re-verified November 19, 2026 via direct English HTTP retrieval. Analysis and provocations are my own and labeled as such.

  • Engine split, release dates, development history, awards and bullet-heaven status
  • Phaser vs Unity engine split, HTML5 origins and pricing history
  • Phaser fast free fun open-source HTML5 definition, web-only scope and console exclusion
  • Phaser MIT, 2013 origins, Canvas plus WebGL via Pixi.js and physics systems
  • Unity WebGL overview and Web development index
  • Unity Web performance: GPU near-native, Emscripten CPU variance, draw-call batching, threading and memory limits
  • Steam minimalistic time-survival description and bullet-hell tagline
  • EasyHub portal instant HTML5 and Unity WebGL positioning
  • Browser trainers: 10-minute mage survival loop, auto-attack top-down survival, and auto-battler build optimization

Top comments (0)