Roadmap

WoalzCraft today is a Blueprint generation agent. Over the next 12–18 months it expands toward becoming a Technical Artist-grade tool covering Materials, Niagara VFX, Animation Blueprints, Editor Utility Widgets, and Procedural Content Generation.

This page is honest about what's shipped, what's next, and the rationale. We update it as priorities shift; the dates are intent, not promises.

Where we are today (v0.7.x — Phase 2.17 shipped)

| Domain | Status | | ---------------------------------- | ------------------------------------------------------- | | Blueprint generation | ✅ Stable. Single-BP and multi-BP plans both supported. | | Brain knowledge base | ✅ 113+ curated entries, public search at /search. | | Plain-English error translator | ✅ Stable. | | MCP server | ✅ Claude Code, Cursor, Cline, Gemini CLI all work. | | User accounts + activity audit | ✅ Just shipped (Phase 2.16). | | Public knowledge browsing | ✅ Just shipped (Phase 2.17). |

What's stable for production-ish use right now: gameplay Blueprint generation in UE 5.3 → 5.7, with Brain-backed few-shot examples and a chat agent that can plan multi-BP systems.

The seven capability tracks

The phases below all fall into one of seven parallel work streams. Some tracks ship features faster than others depending on dependencies; this list helps frame where any single feature sits in the bigger picture.

| Track | What it covers | Phase footprint | |---|---|---| | A — Domain Mastery | Graph generation: Blueprints, EUW, Materials, Niagara, Behavior Trees, Anim BPs, PCG, Control Rig | Phases 3–7 | | B — Asset Pipeline Bridge | Quixel, Mixamo, Sketchfab, Suno integration. WoalzCraft orchestrating asset import + setup, not just BP | Phase 9 | | C — Runtime Intelligence | LLM-driven NPCs as a drop-in component. Memory, personality, voice. | Phase 8 | | D — Platform & Genre Expansion | Multiplayer replication, save/load, mobile, VR, console, live service patterns | Ongoing | | E — Studio & Collaboration | Multi-user sessions, team Brain, pattern marketplace, shared libraries | Phase 10 | | F — Education & Accessibility | Multilingual UI (Indonesian first), interactive tutorials, voice-driven authoring | Parallel | | G — Quality, Velocity, Refactoring | Project-wide rename, perf coach, AI test agent, version migration agent | Phases 10–11 |

The internal master vision document covers each track in depth. Public roadmap below focuses on Track A (where most user-visible features land), with Tracks B–G mentioned at their respective phases.

Phase 3 — Editor Utility Widgets (next, ~3 months)

Why first? EUW is literally a Blueprint for the editor. The plugin's existing graph mutator covers ~90% of EUW generation with minimal new code. High-value, low-friction win — perfect bridge from "BP tool" to "broader editor tool".

What unlocks

"Generate an EUW that batch-renames selected assets to PascalCase"
"EUW that finds and lists all unused textures"
"EUW for setting collision profile on a mesh batch"
"EUW that validates every Material has a proper LOD"
"EUW that organizes Content/ folder by asset type"

Why Technical Artists care

TAs author tools-of-tools constantly. Replacing 30 minutes of manual EUW clicking with a 30-second AI prompt compounds across every TA in the studio. This is also where the productivity ROI is highest — TAs are usually senior, expensive, and time-constrained.

Effort + scope

Knowledge: +30 EUW patterns + node references
Plugin: EUW generator (forks the existing BP graph mutator)
MCP tool: generate_editor_utility(prompt)

Phase 4 — Materials (after Phase 3, ~4–6 months)

Why second? This is the most-requested TA capability and the graph paradigm is familiar — Material Editor uses the same approach as Blueprint Editor with a different node namespace. Once Phase 4 ships, the architecture for Niagara / PCG / Anim BP becomes much easier because the "graph generation in another domain" path is proven.

What unlocks

"Master material for foliage with wind animation"
"PBR base material with subsurface scattering"
"Material function for triplanar projection"
"Material instance with 5 color variants"
"Glass material with Fresnel + refraction"
"Dissolve effect material with noise mask"

Sub-domains to cover

Master materials (Surface / Decal / Volume / PostProcess domains)
Material functions (reusable graph snippets)
Material instances (parameter overrides on a master)
HLSL custom nodes (the deep end — for engine programmers)

Effort + scope

Knowledge: +80 material patterns, node references, and HLSL snippets
Plugin: Material graph mutator (the new bit — Material nodes have different metadata than BP nodes)
MCP tools: generate_material(prompt, domain), generate_material_function(prompt), analyze_material_instructions(path)

Pivotal moment

This phase shifts WoalzCraft from "Blueprint tool" to "graph generation tool". After Materials, every other graph-based domain (Niagara, Anim BP, PCG, Control Rig) is incremental — same engine, different vocabulary.

Phase 5 — Niagara VFX (after Phase 4, ~3–4 months)

Niagara reuses Phase 4's graph mutator. New work is mostly content (curating Niagara module patterns) and editor integration (emitter/system structure).

What unlocks

"Fire particle system with smoke trail and ember sparks"
"Water spray on impact"
"Magic missile trail with light flicker"
"Damage number popup with float-and-fade"
"Foliage interaction wind effect"
"Death dissolve with rising particles"

Why this matters for TAs

Niagara is high-skill, high-impact, and where artists spend disproportionate time. A solid pattern library + AI generation moves TAs from "I'll spend a day building a fire VFX" to "I'll prompt one in 5 minutes, then iterate visually for 30 minutes". 80% of routine VFX needs become trivial; bespoke artistic VFX still needs human authorship.

Phase 5.5 — Behavior Trees (parallel to Phase 5, ~2 months)

Behavior Trees power every "smart NPC" in UE5 — patrol, chase, attack, retreat, coordinate. The graph-mutator architecture from Phase 4 transfers cleanly; new work is curating BT-specific patterns and integrating with the AI Controller / Blackboard model.

What unlocks

"AI Controller + BT for ranged enemy: take cover, peek-shoot, retreat at low HP"
"Squad coordination BT: leader assigns flank/center/support roles"
"Boss BT with three phases (aggressive → defensive → enraged)"
"Stealth NPC BT: idle patrol, suspicious-investigate, alert-search, hostile-attack"
"Companion AI BT that follows player and assists in combat"

Why pair with Phase 5 (Niagara)

Smart NPCs without VFX feedback feel sterile. Smart NPCs with on-trigger VFX (alert effect, attack telegraphs, death dissolve) feel alive. Shipping Phase 5.5 alongside Phase 5 lets users build the full enemy experience in one workflow.

Visual Feedback Loop (cross-cutting, lands with Phase 5)

This is the single most important capability that transforms WoalzCraft from "code generation tool" into "creative collaborator" — and the reason Phases 4–6 deliver iteration workflows worth using.

The problem it solves

AI is text-based. Most UE5 outputs are visual (a fire VFX, a material's appearance, an animation pose). Without seeing the result, AI can't iterate intelligently — it can only re-roll based on the user's text description of "what's wrong".

The solution

A screenshot capture pipeline built into the plugin:

User clicks "Show AI my result" in the panel
Plugin captures the relevant viewport (Material preview, Niagara preview, animation pose, level viewport, etc.)
Image sent to a vision-capable LLM (Claude 4.5 Sonnet, GPT-4V, Gemini 2.5 Pro multi-modal)
AI analyzes: "I see a fire VFX with too much smoke and not enough warm color variation in the flame core."
AI proposes specific changes: "Reduce smoke emitter spawn rate by 50%, add color ramp gradient from yellow → orange → deep red in the flame's Color over Life curve."
User approves → AI applies → user reviews again

Use cases per domain

| Domain | What visual feedback unlocks | |---|---| | Materials | "This metal looks too plastic" → AI sees, suggests roughness/normal map adjustments | | Niagara | "Make it more dramatic" → AI sees current state, knows what dramatic means contextually | | Anim BPs | "Walk cycle looks robotic" → AI analyzes pose, suggests blend space tweaks | | UMG | "This menu feels cluttered" → AI sees, suggests layout reflow | | Lighting | "Scene looks flat" → AI sees, suggests light/shadow adjustments |

Constraints we set up-front

Cost: vision API calls are 5–10× pricier than text. Reserve for explicit "show AI" actions, not auto-watching every edit.
Privacy: editor screenshots may contain unfinished IP. Each capture is opt-in per call. Never auto-uploaded.
Latency: vision LLMs are slower (~5–10s). Position as "iterate carefully" not "real-time".

Phase

Lands during Phase 5 (~12–14 months out). Immediately benefits Phases 4 (Materials), 5 (Niagara), 6 (Anim BPs).

Phase 6 — Animation Blueprints (after Phase 5, ~4–6 months)

Anim BPs add complexity: state machines, blend spaces, retargeting, and the dual-graph (event graph + anim graph) structure. We hold this until Phase 4–5 prove the multi-domain workflow.

What unlocks

"Locomotion anim BP with idle/walk/run/jump"
"Combat state machine with attack/dodge/block/stunned"
"Aim offset blend space generator"
"Crouch/stand transitions"
"Climbing system with state machine"

Tricky bits

State machine logic — transitions and blend rules are graph-shaped but with different semantics than BP exec flow
Retargeting — links to Skeleton + IK Rig assets, has cross-asset dependencies
Animation Montages — slotted notifies, branching points

We expect Phase 6 to take longer than 4–6 months in practice. Calibrating optimism against reality.

Phase 7 — Procedural Content Generation (after Phase 6, ~3 months)

PCG is graph-based and reuses the architecture from Phase 4. Niagara is for runtime particles; PCG is for build-time content placement.

What unlocks

"Generate a PCG graph that scatters rocks across a landscape with density falloff near roads"
"Foliage scatter graph: trees in clusters, grass between them, mushrooms only in shaded areas"
"Urban kit assembler: place buildings along streets with random rotation and snap-to-grid"
"Cave system PCG: stalactites from ceiling, pebbles on floor, moss on walls"

Why this matters

PCG is one of the highest-leverage features in modern UE5 — it lets a small team build worlds at AAA scale. The graph paradigm is now familiar territory; this phase is mostly content (PCG node knowledge) + integration polish.

Phase 8 — LLM-Driven NPCs (year 2, ~3–4 months after Phase 7)

Track C delivers. Drop-in component for runtime LLM-driven characters.

What unlocks

WoalzAINPC component drops onto any Pawn → conversational behavior
Dialogue UI templates (text input, history, choice menus)
Memory persistence per NPC per save slot
Personality system (per-NPC system prompt, traits, mood states)
TTS integration (ElevenLabs, OpenAI TTS, local Bark)
Voice input for player dialogue (Whisper, Google STT)
Cost-aware routing — cheap models for ambient banter, expensive for crucial story beats

Example flow

Player approaches village merchant.
Player types/speaks: "Have you seen any dragons lately?"
WoalzAINPC component → Brain (lookup village context, NPC memory)
                    → AI provider (with personality + memory + scene state)
                    → Dialogue: "Dragons? That's the third time someone's
                       asked this week. Something stirs in the mountains.
                       I won't sell to dragon-hunters at full price."
TTS plays dialogue with merchant's configured voice.
NPC remembers: player asked about dragons, treats them as adventurer henceforth.

Why this is a marketing moment

Modern smart-NPC is the most-talked-about gaming AI category. WoalzCraft's existing AI infrastructure is 70% of what's needed; Phase 8 packages it into a drop-in component. First UE5 plugin with first-class LLM-NPC support has positioning advantage.

Why we wait until Phase 8

Skipping straight to LLM-NPCs without strong Blueprint + Animation + VFX support produces shallow demos. Smart NPCs need: idle anim BP, perception system, dialogue trigger logic, VFX feedback (alert, hostile, defeated). Phase 8 stacks on top of Phases 1–6 to deliver the full experience, not just the chat layer.

Phase 9 — Asset Pipeline Bridge (year 2, ~6 months)

Track B delivers. WoalzCraft becomes the orchestrator across the asset pipeline, not just BP / graph generation.

What unlocks

| Integration | What it does | |---|---| | Quixel Bridge | "Find me a desert rock cluster, import 4 LODs with material instances" — direct Megascans search | | Mixamo workflow | Animation download + auto-retarget to project skeleton + AnimBP scaffolding | | Sketchfab import | Free model search with auto-LOD generation | | Suno / ElevenLabs | Music + voice generation with drop-in audio components | | Stable Diffusion bridge | Texture generation hooks (advanced; careful UX) | | Marketplace + Fab search | "Find me a free explosion VFX" — search, preview, install in-editor |

Why this matters

WoalzCraft handling Blueprints in isolation captures 30% of the value. Orchestrating the whole pipeline captures 80%. After Phase 9, a developer can describe "I want a desert village with a few NPCs", and WoalzCraft can fetch + place + wire enough assets to make a runnable scene.

Phase 10 — Quality, Velocity, Refactoring (year 2–3, ongoing)

Track G delivers. The work that matters AFTER initial generation.

What unlocks

| Capability | What it does | |---|---| | Project-wide rename | "Rename PlayerStats → CharacterStats across 50 BPs and their callers" | | Extract function | Select nodes → AI extracts to reusable function with proper inputs/outputs | | Convert to component | "This BP has 200 nodes; split out movement into a UCharacterMovement component" | | Detect duplication | "Find BPs doing similar things; suggest consolidation" | | AI code review | Reviews BP changes before commit, catches anti-patterns (Tick on every actor, etc.) | | Compile error fixer | "Why does this fail to compile?" → AI explains + applies fix | | Performance coach | Reads UE5 Insights captures, suggests optimizations, applies safe ones | | UE5 version migration | Auto-detect UE version, compare to target, apply safe upgrades, flag risky ones | | Test generation | Generate test cases for a BP based on its event graph | | Doc generator | Auto-README from project structure + BP comments |

Why this matters

Studios + experienced solo devs spend at least as much time refactoring + fixing as they do building new things. Phase 10 features compound — each saves hours per week, every week, for as long as the project lives.

Phase 11+ — Studio, Marketplace, Test Agent (year 3+)

Tracks E, B, G converge into the most ambitious capabilities.

Studio tier (Track E)

Multi-user editor sessions (built on UE5's Concert + WoalzCraft AI layer)
Private team Brain (project-scoped patterns, never shared)
Permission tiers within a team (Admin / Author / Reviewer / Read-only)
Cross-team audit log
Shared pattern libraries (your studio's conventions, codified)

Pattern marketplace (Track E)

Creators publish patterns to a public marketplace
Free + paid patterns
Verified contributor program
Quality reviews
Revenue share with creators

AI test agent (Track G — ambitious)

AI plays the game with a defined goal ("complete the tutorial")
Captures gameplay (screen + game state)
Reports issues ("crashed when picking up item X", "stuck on this jump")
Compares behavior across builds (regression detection)

This is multi-year ambition. Realistic to expect functional v1 in year 3, robust v2 in year 4–5.

What we're not planning

To stay focused, here are things we're explicitly not building:

❌ Mesh / texture / audio generation (use third-party tools — Quixel, Mixamo, Stable Diffusion, Suno, etc. — and import results)
❌ Level design / world placement automation (use UE5 World Partition + manual)
❌ A separate IDE / editor (UE5 IS the editor — we're a plugin in it)
❌ Cloud rendering or remote build farms (out of scope; orthogonal product space)
❌ Replacing Unreal Marketplace or Fab (we integrate, not replace)

If something on this list becomes critical for users, we'll revisit. But the principle is: stay in the graph generation lane, do it excellently.

How priorities change

This roadmap shifts based on:

Real user feedback. If Phase 5 (Niagara) gets requested 5× more than Phase 6 (Anim), we'll re-order.
Technical learnings. Material graph generation might surface deeper issues that delay Phase 4 — we'd rather extend the timeline than ship a broken Phase 4.
Contributor recruitment. Some domains (Materials, Niagara) need TA-domain contributors to seed the knowledge base. We hire those signals into the schedule.

How you can help

If you live in a specific UE5 domain — Materials wizard, Niagara guru, Anim BP veteran, EUW poweruser, BT designer, multiplayer specialist, mobile/VR developer — we'd love your knowledge:

Contribute patterns — submit your reusable patterns via the Contributing flow. We credit contributors and you get free perpetual user-tier (and eventually Pro-tier) access.
Review beta domains — when Phase 4–11 hit beta, real domain experts trying real workflows will catch gaps faster than we can.
Tell us what you'd want — support@woalz.com. The more concrete ("I'd kill for an EUW that does X"), the more we can prioritize.
Translate the plugin UI — Track F needs Indonesian + 4 other major languages. Native speakers welcome.
Beta-test LLM NPCs — when Phase 8 hits beta, we need games + creators willing to try LLM-driven characters in real projects.

Honest risks

The above is ambition. Things that could derail or delay:

Solo dev bandwidth. WoalzCraft is currently maintained by one person. Each phase is a sequential time investment, not parallel.
UE5 release cycle churn. Epic ships UE updates every ~6 months that change node namespaces. Multi-domain support means multi-version tracking — expect occasional regressions on new UE versions.
TA contributor recruitment. Without knowledge from real TAs, Phase 4–7 risk being shallow. Recruitment is on the critical path.
Scope discipline. The temptation to add Domain X "just because we can" is real. We'll push back on it; the goal is depth before breadth.

We'd rather under-promise + over-deliver than the reverse. If a phase ships late, you'll see it on this page first.

Current version

WoalzCraft is in public alpha (v0.9.x plugin, Brain v0.7.x). All Phase 1–2 capabilities are stable; Phase 3+ is the work ahead. File issues generously — early feedback shapes the next year of work.

For deeper detail

This roadmap is the public view of the plan. The internal master vision document covers:

The 7 capability tracks (this page covers Track A — Domain Mastery)
Tracks B–G: asset pipeline, runtime intelligence, platform expansion, studio collaboration, education, quality+refactoring
The visual feedback loop (cross-cutting innovation that unlocks Material/VFX/Anim iteration)
Pricing tier strategy
Success metrics + risk register
Resource plan + decision framework

Living in the source repo: WoalzCraftBrain/docs/MASTER_VISION.md. Quarterly updates as reality diverges from plan.

If you're considering contributing or want to understand the long-term direction, read it.