Graphical Differences - Dragon Quest XI

Original (OG) and Definitive Edition (DE)

Models

Models, by and large, are the single worst hit area of Definitive Edition (DE) compared to the original (OG) version. Poly counts are decreased by around 33-50% overall, but this varies wildly depending on the specific assets being viewed.

Some asset types have relatively minimal impact, while others are quite large. We'll go through some specific examples below and break these models into 2 types: skeletal and static meshes.

Skeletal Meshes

Skeletal meshes include character models, enemies, and anything that is going to be animated throughout the world. They have a "skeleton" which defines how they are going to move and deform when animated.

Skeletal meshes are more noticeable and far worse hit than static meshes between these 2 versions.

Hero / Luminary's main outfit (OG)

Hero / Luminary's main outfit (DE)

In the top left of each image, you can see a poly count ("Tris" in this case) and a vert count (short for "Vertices").

It should be evident that the Definitive Edition (DE) model has about 1/7 the number of tris as the original (OG) model (an ~86% decrease!). These losses are most notable in certain places such as hair, fingers, sleeves, and other such small details.

This is a massive difference by the numbers, especially when you consider this is the main character of the game.

This model is seen (1) very frequently and often (2) very closely

--> these are 2 crucial factors to consider when looking at individual models and how (subjectively) "noticeable" they would be in game.

Let's look at another one...

Hero / Luminary's Horse (OG)

Hero / Luminary's Horse (DE)

Here we have another hard hit model. This time about 1/6 the number of tris (~83% decrease).

Slime (OG)

Slime (DE)

Now we have the slime. Essentially the series mascot. This one is far less egregious with about 1/3 the number of tris (a 66% decrease).

This may still seem like quite a lot, but at (1) the distance the player will typically see them and (2) the frequency of which they are seen. This is considerably minor compared to the above examples.

Static Meshes

The other model type we have to discuss are static meshes. These include basically anything in the environment or world that are not animated.

These tend to be impacted much less compared to skeletal meshes.

Arboria - Serenica Statue (OG)

Arboria - Serenica Statue (DE)

Kingsbarrow - Pillars (OG)

Kingsbarrow - Pillars (DE)

Save Statue (OG)

Save Statue (DE)

Common Heliodor Region Grass (OG)

Common Heliodor Region Grass (DE)

Common Heliodor Region Tree (OG)

Common Heliodor Region Tree (DE)

Here we have a variety of static meshes of different categories. In part, to demonstrate how each category is affected: from Architecture, to Statues, Foliage, Trees, and even small props.

  1. In the first example, we have the statue seen towering over the town of Arboria. This one is reduced by approximately 1/2 (50%). However, as noted above, this is seen at a very far distance, making this relatively insignificant.

It should also be noted that at a certain distance, the game is likely to load a lower LOD anyway (in which case, something like an INI tweak would have more of an impact to force higher LODs to stay loaded).

  1. In the second example, we have a typical pillar that you'd see in just about any interior. Once again we have about a 50% reduction right on the nose.

However, in this case we have something that you see all the time and very closely. Despite being the same level of reduction, this one is much more impactful.

  1. In the third case, we have the save statue seen all over the game at every campsite/town/etc. This is technically listed as a "Countryside Prop" in the game's files

This mesh is the exact same between OG and DE

  1. In the fourth case, we have a simple blade of grass.... Actually, no. This is the most common grass mesh seen throughout the game.

It is a major reduction, but it doesn't tell the full story. This is covered much more in-depth in the "Foliage" section below

  1. In the fifth case, we see a fairly typical tree you might find just about anywhere in the overworld (but particularly common in Heliodor). This one is about a 38% reduction, and compared to many other cases, is rather minor and won't be nearly as easily noticed by players.

UV Maps

Before moving into "Textures", it may be important to touch on UV Maps. UV maps effectively determine how a texture is going to be "wrapped" around a model.

Different models use different UV maps and that will play a role in how a texture is going to look on a given model. Even if it's the same texture.

Textures

Resolution

In most cases, DE (2020) uses an identical texture resolution to OG and 2-4x that of the 2019 Switch release.

Characters

Character textures are probably one of the more interesting areas. The UV maps are changed significantly here and the DE version uses merged layers

This is a mixed bag, as in DE's case a separate copy of the body/skin, face and hair are stored with every outfit texture

This generally results in a quality reduction for these features in order to fit on the same layer as everything else.

This is also faster, however, as the game only needs to reference a single texture for the entire character/outfit. Much like texture atlases would do for world elements.

One could argue, however, that it's less efficient than having a single hair texture and a single face texture on a separate layer, as in the vast majority of cases these don't change between outfits.

From a modder's perspective, this could go either way-

  • if you want to create an outfit that changes the hair and face for just that outfit, use the DE mesh

  • if you want to create a mod that changes the hair or face for any outfit the player has equipped, use the PR mesh

Mirroring

DE has one major advantage with its reworked textures- it takes advantage of texture mirroring.

Basically, DE is able to save on space with the texture by mirroring certain parts that are otherwise identical (i.e. left side and right side of a body -> only store the left side and use mirroring to render the right side, rather than storing the whole thing).

This results in much better preservation of detail in some cases (more detail is able to be placed on the same texture), even considering the layering reduction.

Monster textures benefit the most from this change, as there is no layering reduction as in the case of character textures.

World Elements

World textures, for the most part are the exact same between versions. With one important exception:

Compression

DE features major improvements in texture compression, switching from DXT1/DXT5 to BC7 in a number of cases.

This has huge advantages, both in terms of optimization and quality

Bush Leaves (OG)

Bush Leaves (DE)

Ground (OG)

Ground (DE)

Slime (OG)

Slime (DE)

UI Character Bust (OG)

UI Character Bust (DE)

Armor Icon (OG)

Armor Icon (DE)

Saving/Loading Icon (OG)

Saving/Loading Icon (DE)

1/2. The first two examples are relatively self-explanatory: the same UV map with the only difference being the compression quality improved in DE.

3. In the third example, we once again have our old friend the slime from earlier -- this time it's a different UV map

--> you can see the devs are taking advantage of texture mirroring by only including half the eye and making it larger (also gave them a proper smile :)

4. Here we have the character portraits seen in the menus -- DE benefits immensely from improved texture compression here

5. Icons seen in the menus. Essentially the same situation as #4 albeit slightly less impactful as most icons are scaled down to a smaller size


While we could go on with textures (just like models). For the most part, it's a slight win for DE in terms of world textures or UI textures, while character textures are more impacted by changes in UV maps and different layering, making those a slight win for OG (highly dependent on the specific texture - monsters may be a slight win for DE due to compression & mirroring improvements).

Lighting

Enlighten

The original (OG) version licensed Enlighten , a middleware (third party software which "plugs into" UE4) that provides real time global illumination and acts as an alternative to UE4's default global illumination system: Lightmass.

Lightmass

Definitive Edition (DE) stuck with the UE4 default (Lightmass). This change is thought to have been done because

  1. Switch support in Enlighten wasn't very mature at the time of DE's development (possible bugs, etc)

  2. The cost to license Enlighten was considerably expensive, especially for Switch.

  3. Performance (Enlighten likely would have reduced performance on Switch's hardware)

  4. Improvements made to Lightmass between 4.13 and 4.18


Take your pick. Bottom line: the lighting looks different -- let's check out a couple examples.

Enlighten (OG) on left, Lightmass (DE) in middle, and Tuned Lightmass (PR) on right (where available)

-- > Note that some screenshots are not up-to-date. Screenshots are taken by multiple community members on different hardware/configurations

Kingsbarrow Entrance (Enlighten)

Kingsbarrow Entrance (Lightmass)

Kingsbarrow Entrance (Tuned Lightmass)

Heliodor Castle Entrance (Enlighten)

Heliodor Castle Entrance (Lightmass)

Heliodor Castle Entrance (Tuned Lightmass)

Costa Valor (Enlighten)

Costa Valor (Tuned Lightmass)

Gondolia Entrance (Lightmass)

Gondolia Entrance (Tuned Lightmass)

Emerald Coast (Lightmass)

Emerald Coast (Tuned Lightmass)

Heliodor Church (Enlighten)

Heliodor Church (Lightmass)

Heliodor Church (Tuned Lightmass)

Tuned Lightmass

One small advantage to Lightmass is that it is more tunable in-engine. Tuning will not "make it more like Enlighten" as it is a different beast, but lighting effects can be improved in this way as well.

Foliage

Foliage density (more specifically, grass density), are one of the most hard hit areas by the changes in Definitive Edition (DE). It is arguably a toss up between this and character models for the single most egregious shortcomings of the DE version.

As noted in the static mesh comparison above, it isn't just that the grass meshes themselves have seen a reduction but the placement of grass in the umaps (map files) have a vastly reduced number of instances.

Unlike other titles, DQXI-S has density scaling disabled on its grass instances (set by bEnableDensityScaling in UE4).

Instead, grass instances are manually placed ("painted") on the landscape.

Foothills Grass (OG)

Foothills Grass (DE)

Foothills Grass (PR)

Performance

As noted on the general version differences page ("Under the Hood"), performance is improved in DE - one area that it thoroughly outshines OG.

This is debatable as a "graphical difference", but many consider framerates to be an important part of visual fidelity. Thus worth including here.

Miscellaneous

This section will cover any additional known details not mentioned above

  • Animations are identical between versions (as far as existing knowledge and documentation is concerned), including bugs present in the original (OG) release

  • CGI Cutscenes are also identical

  • A small handful of in-game cutscenes (previously rendered by the engine in real time) have been replaced with prerendered versions (converted to .usm and play back similarly to the CGI cutscenes) in DE

  • Certain surfaces have additional Screen Space Reflections (SSR) applied in DE (seen in the Heliodor Castle images under the "Lighting" section)

  • The default field of view is higher in DE than OG (in many cases the camera is moved further behind the player)

  • Certain particle effects (Magic Burst in particular) are believed to be slightly altered in DE (this may require more testing to confirm for other cases)

Magic Burst (OG)

Magic Burst (DE)

That's all for the (mostly) bad news. The good news is that Project Rebuild addresses most of these by combining the best of both worlds

--> This also allows custom models/retextures to choose between using the optimized models and simple layering of DE, or the higher quality, multi-layered OG/PR models. These can even be swapped at will to take advantage of each approach!

--> In addition to adding the higher quality individual grass models, this module also increases the density present in the models rather than simply adding instances on the map. In many cases, this can result in fuller, thicker looking grass with more realistic distribution and better draw distance than seen in the OG version.

  • Module 6 tackles the lighting of Definitive Edition.

--> Rather than try to emulate Enlighten, this module tunes Lightmass to accentuate its strengths.