The Unreal Engine Feature Radar

Nanite has been production-grade since UE5.0 and has only widened its remit since. For static, high-density meshes it is the default we reach for — it removes the LOD-authoring grind and lets artists work at source fidelity. 5.7's move to bring first-person rendering with Nanite to production closes one of the last awkward gaps.

The judgement to make is no longer *whether* to use Nanite but *where*: skeletal meshes, world-position-offset foliage and heavy translucency remain the places to measure rather than assume. Profile, don't guess — but start from Nanite-on.

Lumen is the other half of the modern UE5 look and is firmly production-ready. Software Lumen scales down to current-gen consoles at 60Hz with care; hardware ray tracing buys sharper reflections and cleaner interiors where you have the GPU budget, and 5.7 continued to improve both.

It is not free — Lumen rewards teams who treat lighting as a performance discipline, budget for it early and validate on target hardware. But as a starting point for dynamic lighting on a new project, it's the right one.

Virtual Shadow Maps are the shadowing counterpart to Nanite and Lumen, and like them they're production-ready and the expected default in a modern UE5 renderer. They deliver consistent, high-resolution shadows across large scenes without the manual cascade tuning of the past.

Cost lives in overlapping casters and heavy foliage; the mitigations are well understood and documented. Adopt, and keep an eye on the VSM stats when scenes get dense.

Substrate replaces the fixed shading models with a layered, principled framework, and Epic moved it to production-ready in 5.7. The capability is real and the future of UE materials clearly runs through it.

This is the sharpest Epic-says-Production, we-say-Trial call on the radar. The framework is ready; the *ecosystem* around it — third-party materials, team muscle memory, and the cost of converting an existing project — is still catching up. For a brand-new project, trial it with intent. For a mid-flight one, plan the migration deliberately rather than flipping it on.

MegaLights lets you place far more shadow-casting dynamic lights than was previously affordable, and it has moved quickly: experimental, then beta in 5.7 with directional lights, Niagara particle lights, translucency and hair support. The 5.8 preview pushes it to production-ready.

We place it in Trial: beta in the current *stable* engine is exactly the 'commit where you can absorb risk' case, and we don't promote a feature on the strength of a preview build. When 5.8 ships, expect this to move to Adopt — it's one of the most promising lighting changes in years.

Hardware Lumen and ray-traced reflections/shadows are production-ready and deliver a visible quality step on capable GPUs. 5.7's reflection and interior-lighting improvements make the case stronger again.

We hold it at Trial rather than Adopt only because the decision is so target-dependent: it's an easy yes for high-end PC and current-gen console showcases, and an easy no for the lower end of your hardware spread. Decide by your minimum spec, not your hero shot.

World Partition, with Data Layers and Level Instances, is the production-ready, expected approach to large worlds in UE5 — it replaced the old sublevel workflow and is what Epic's own large-world tooling is built around. 5.6 and 5.7 both invested in making big worlds run smoothly at 60Hz.

Adopt for anything open-world or simply large. The learning curve is real (One File Per Actor changes how teams collaborate in source control), but it's the right default and the ecosystem assumes it.

PCG graphs let you scatter, place and procedurally build world content non-destructively, and Epic designated the framework production-ready in 5.7, alongside continued work on biome and vegetation tooling. It has gone from an interesting experiment to a dependable part of the worldbuilding pipeline in a short time.

We agree with the Production label and Adopt it. Treat the newer biome/procedural-vegetation editors as the less-settled edge (see Nanite Foliage), but the core PCG framework is ready for real work.

Nanite Foliage and the accompanying Procedural Vegetation Editor arrived as experimental features in 5.7, aiming to bring Nanite's density advantages to vegetation — historically Nanite's hardest case. The early results are exciting and the direction is clearly right.

Experimental means exactly that: Assess it, prototype with it, file feedback — but don't build a shipping vegetation pipeline on it this release. This is a strong candidate to climb the radar quickly over the next version or two.

Mesh Terrain is a new experimental system in the 5.8 preview for building large-scale landscapes from a 3D mesh rather than a heightfield — a meaningful rethink of the terrain pipeline if it matures.

It is Assess, and barely on the board: it exists only in a preview build today. We're flagging it so you know it's coming, not recommending it for any current work. Revisit when 5.8 ships.

Enhanced Input is the production-ready, expected way to handle input in UE5, with input actions, mapping contexts and modifiers replacing the legacy action/axis mappings. It's mature, well-trodden and assumed by most current learning material.

There's no reason to start a new project on the old system. Adopt without reservation.

The Gameplay Ability System is shipped-in-AAA, production-grade code for abilities, attributes and effects, with networking handled. For anything with skills, cooldowns, buffs or replicated stats, it saves you reimplementing a hard problem badly.

Our one honest caveat keeps it from being an unqualified Adopt for every team: official documentation remains thin, so the real cost is the ramp-up. With a team willing to climb that curve (or lean on the strong community material), Adopt.

State Tree gives you a hierarchical, data-oriented state machine usable well beyond AI — UI flow, gameplay states, sequencing. It reached production-ready maturity over recent releases and is increasingly Epic's recommended structure for this kind of logic.

We place it at Trial: it's solid and we'd commit to it, but it's newer to most teams' toolkits than the systems above it, so spend a project getting the patterns right. Pairs naturally with Mass and GAS.

Mass is UE5's data-oriented (ECS-style) framework for simulating very large numbers of entities — crowds, traffic, swarms — and it's genuinely capable at scale. Epic uses it in its own large-world demos.

It remains experimental, the API still shifts between versions, and the documentation is sparse — so it's an Assess. If you have a real high-agent-count problem and engineers comfortable on the bleeding edge, prototype with it; otherwise watch it and don't build core gameplay on it yet.

CommonUI is production-ready and the sensible foundation for UI that must work across mouse, gamepad and touch with proper focus navigation and input routing — the things plain UMG leaves you to solve yourself. Epic's own titles ship on it.

We say Trial rather than Adopt because the up-front complexity is significant and overkill for a small single-platform UI. For anything console or cross-platform, it's worth the climb; size the decision to your UI's real scope.

Verse is Epic's programming language and a clear strategic direction for the engine's future. Inside UEFN / Fortnite it's a shipping reality. For a standard, standalone Unreal Engine project, however, its availability is still rolling out and experimental.

So the verdict splits by context. If you're building for UEFN, Verse is simply part of the platform. If you're building a standalone UE5 game, treat Verse as Assess — learn it, watch the roadmap, but keep C++ and Blueprint as your production languages for now. This is the entry most likely to be re-judged as the rollout proceeds.

Niagara is the production-ready VFX system in UE5 and has been for the whole 5.x line. It's deep, GPU-capable, and the assumed target for modern Unreal effects work; the old Cascade system is legacy (see our Migration radar).

Adopt without hesitation for any new effects work. It rewards investment in reusable module and emitter libraries.

MetaSounds is the production-ready, node-based audio system that replaced the old Sound Cue workflow, giving you sample-accurate, fully procedural control over sound generation. It's mature and the expected way to author audio in UE5.

Adopt for new audio work. It's one of the cleaner 'new system fully replaced the old one' stories in the engine.

Control Rig is the production-ready system for building rigs and procedural animation directly in-engine, underpinning a lot of UE5's modern animation workflow (including much of the MetaHuman and motion tooling).

Adopt for in-engine rigging and runtime procedural animation. It's well-established and the surrounding animation tools assume it.

Pose Search / Motion Matching gives you high-quality, data-driven locomotion without hand-built blend trees, and Epic's Game Animation Sample turned it from a curiosity into something teams genuinely ship. The tooling has matured well across recent releases.

We place it at Trial: it's ready to commit to, but it changes how your animation team works and needs a good motion dataset to shine, so adopt it deliberately on a project that can invest in the pipeline.

MetaHuman matured significantly when its creation tools moved in-engine (from 5.6 onward) and the licensing opened up, making high-quality digital humans and their animation pipeline far more accessible. The character and animation tooling around it keeps improving each release.

We say Trial: the core is production-grade and we'd happily build on it, but the in-engine toolset is still settling release-to-release, so pin your engine version and validate your pipeline rather than assuming stability across upgrades.

Chaos is UE5's physics system and fully replaced PhysX; the core rigid-body simulation is production-ready and the default. For standard physics it's a non-decision — it's simply what the engine uses.

Adopt for core physics. The specialised sub-systems (destruction, cloth, flesh) vary in maturity from solid to experimental — judge those case by case rather than assuming the whole Chaos umbrella is equally mature.