Pardall Games is a solo project by Alfred Baudisch: an interactive digital museum and archive of assets from PlayStation, Nintendo 64 and Sega Saturn games. You can examine 3D models, maps, levels, effects, sprites and textures, compare models in a 3D viewer or side by side in a 3D arena, and watch animations of characters, monsters, NPCs and environments.
It is free, non-commercial, and view-only — nothing can be downloaded.
The view-only decision is what makes it possible
That restriction is the first thing anyone notices and the reason the archive exists at all.
These assets are owned, and they are owned by companies that are still trading. A site offering downloadable Final Fantasy VII models is a takedown waiting to happen; one offering a viewer for studying them is a research and reference resource that has a plausible position. The framing is explicit — viewing, exploration, research and reference browsing only — and it is what lets a solo project include Final Fantasy VII, GoldenEye 007, Gran Turismo, the first three King’s Field games, Resident Evil 2 and Spyro the Dragon without immediately disappearing.
It also happens to be the right design for the actual use. You do not want these models in your project. A 1997 character mesh is built for a 320×240 framebuffer, no z-buffer, affine texture mapping and a few hundred triangles — dropping it into a modern engine gives you something that looks wrong rather than retro. What you want is to look at how it was solved, and a viewer is strictly better for that than a download.
Why these three consoles are the interesting ones to study
The PS1, N64 and Saturn generation is the single richest period in real-time graphics for anyone learning the craft, because the hardware was powerful enough to attempt 3D and far too weak to do it properly. Every asset is a negotiated settlement with a specific constraint.
The PS1 had no z-buffer and sorted polygons per-object, which is why things visibly pop through each other, and why artists built geometry to avoid the cases where it would. It also used affine texture mapping with no perspective correction, which is the source of the characteristic warping on large flat surfaces — and why level geometry is subdivided far more than the shape requires. Vertex positions were integers, hence the famous jitter.
The N64 had perspective-correct texturing and a z-buffer but almost no texture memory — commonly 4KB of texture cache — so its look is low-resolution textures stretched large, with heavy use of vertex colours and fog in place of detail. The blurriness is not nostalgia-filter; it is the hardware’s bilinear filter applied to a tiny texture.
The Saturn drew quadrilaterals rather than triangles and had two different video processors with incompatible ideas about geometry, which is why its 3D looks unlike either of the others and why so few developers got good results from it.
Being able to open the actual asset and see how many triangles, how big the texture, where the seams are, how the UVs are laid out turns all of that from trivia into something you can read off the object. That is a different kind of learning from reading a hardware spec.
What is in it
16 titles currently, with assets taken directly from the original games and converted into formats a web browser can display — which is its own quiet piece of work, since each console’s asset formats are undocumented and different.
The browsing is built for finding things rather than listing them: search across characters, environments and visual effects, with filters by scenarios, textures, styles and animations.
The side-by-side 3D arena is the feature that suggests the project understands its own audience. Comparison is how you actually learn from this material — the same kind of character in two games, the same effect solved two ways, one studio’s budget against another’s. A single-model viewer shows you an artefact; two models at the same scale show you a decision.
Contributing
Baudisch does not accept asset contributions, which is consistent with keeping the provenance and the conversion pipeline under one pair of hands. He does take error reports: incorrect names or tags, broken models, missing tags.
That is a sensible division for a one-person archive. The work that benefits from many hands here is cataloguing, not extraction.