diff --git a/src/scenex/util.py b/src/scenex/util.py
index 3ca7cfcb..8fc8ddfa 100644
--- a/src/scenex/util.py
+++ b/src/scenex/util.py
@@ -182,7 +182,7 @@ def show(
     reg.get_adaptor(canvas, create=True)
     app().create_app()
     for view in canvas.views:
-        projections.zoom_to_fit(view, zoom_factor=0.9, preserve_aspect_ratio=True)
+        projections.zoom_to_fit(view, zoom_factor=0.9, letterbox=True)
 
         # logger.debug("SHOW MODEL  %s", tree_repr(view.scene))
         # native_scene = view.scene._get_native()
diff --git a/src/scenex/utils/projections.py b/src/scenex/utils/projections.py
index 9cd19df1..416515c0 100644
--- a/src/scenex/utils/projections.py
+++ b/src/scenex/utils/projections.py
@@ -102,7 +102,7 @@ def zoom_to_fit(
     view: View,
     type: Literal["perspective", "orthographic"] = "orthographic",
     zoom_factor: float = 1.0,
-    preserve_aspect_ratio: bool = False,
+    letterbox: bool = False,
 ) -> None:
     """Adjusts the Camera to fit the entire scene.
 
@@ -119,12 +119,10 @@ def zoom_to_fit(
         approaches 0, the scene will linearly decrease in size. As the zoom factor
         increases beyond 1.0, the bounds of the scene will expand linearly beyond the
         view.
-    preserve_aspect_ratio: bool
-        Whether to apply aspect ratio correction to prevent distortion. When True,
+    letterbox: bool
+        Whether to letterbox/pillarbox to prevent anisotropic distortion. When True,
         squares will appear as squares regardless of view dimensions. When False,
         content may be stretched to fill the view. Default False.
-
-        FIXME: Is this the correct name for this behavior?
     """
     bb = view.scene.bounding_box
     center = np.mean(bb, axis=0) if bb else (0, 0, 0)
@@ -133,32 +131,60 @@ def zoom_to_fit(
     # projection matrix calculations
     w, h, d = np.maximum(np.ptp(bb, axis=0) if bb else (1, 1, 1), 1e-6)
 
-    # Apply aspect ratio correction only if requested
-    if preserve_aspect_ratio:
-        if not (canvas := view._canvas):
-            raise Exception("Cannot preserve aspect ratio without a canvas.")
-        _, _, pw, ph = canvas.rect_for(view)
-        aspect_ratio = pw / ph if ph else None
-        if aspect_ratio is not None and h > 0:
-            if w / h > aspect_ratio:
-                h = w / aspect_ratio
-            else:
-                w = h * aspect_ratio
     if type == "orthographic":
+        if letterbox:
+            # The scene has aspect w:h. If that doesn't match the viewport's aspect
+            # ratio ar, setting the orthographic frustum to those values will distort
+            # world-space squares. We can correct this distortion by expanding whichever
+            # dimension is too small to make w:h == ar. This is kinda like letterboxing,
+            # except you'll see extra scene background instead of black bars.
+            if (ar := _aspect_ratio(view)) is not None:
+                if w / h > ar:  # scene wider than view - expand height
+                    h = w / ar
+                else:  # scene taller than view - expand width
+                    w = h * ar
         view.camera.transform = Transform().translated(center)
         view.camera.projection = orthographic(w, h, d).scaled([zoom_factor] * 3)
     elif type == "perspective":
-        # Compute the distance a to the near plane of the frustum using a default fov
-        o = max(w, h) / 2
         fov = 70
+        # First, we need to figure out how far away to place the camera so that the
+        # entire scene fits within the frustum defined by the FOV. Calculation borrowed
+        # from
+        # https://www.scratchapixel.com/lessons/3d-basic-rendering/perspective-and-orthographic-projection-matrix/building-basic-perspective-projection-matrix.html
+        if letterbox and (ar := _aspect_ratio(view)) is not None:
+            # perspective() produces a square frustum (equal x/y FOV). If the viewport
+            # isn't square, world-space squares will appear distorted. We correct this
+            # by adjusting two different camera parameters. First, if the scene is wider
+            # than the viewport, our distance will actually have to increased based on
+            # that view aspect ratio.
+            o = max(h, w / ar) / 2
+        else:
+            o = max(w, h) / 2
         a = o / tan(fov * pi / 360) / zoom_factor
 
-        # So that the bounding cube's front plane is mapped to the canvas,
-        # the camera must be a units away from the front plane (at z=(center[2] + d/2))
+        # Place the camera so the bounding box's front face (z = center[2] + d/2)
+        # maps to the near plane of the frustum.
         z_bound = center[2] + (d / 2) + a
         view.camera.transform = Transform().translated((center[0], center[1], z_bound))
-        # Note that the near and far planes are set to reasonable defaults.
-        # TODO: Consider making these parameters
-        view.camera.projection = perspective(fov, near=1, far=1_000_000)
+        proj = perspective(fov, near=1, far=1_000_000)
+        if letterbox and (ar := _aspect_ratio(view)) is not None:
+            # Second, if the viewport is non-square, we'll have to adjust our (square)
+            # projection matrix to reflect the non-square viewport. The result is kinda
+            # like letterboxing, except you'll see extra scene background instead of
+            # black bars.
+            proj = proj.scaled((1.0 / ar, 1.0, 1.0))
+        view.camera.projection = proj
     else:
         raise TypeError(f"Unrecognized projection type: {type}")
+
+
+def _aspect_ratio(view: View) -> float | None:
+    if not (canvas := view._canvas):
+        # If the view isn't attached to a canvas, we can't get viewport dimensions, and
+        # can't compute an aspect ratio.
+        return None
+    _, _, pw, ph = canvas.rect_for(view)
+    if pw <= 0 or ph <= 0:
+        # If the view has non-positive dimensions, we can't compute an aspect ratio.
+        return None
+    return pw / ph
diff --git a/tests/utils/test_projections.py b/tests/utils/test_projections.py
index 9d5197a1..cc46276e 100644
--- a/tests/utils/test_projections.py
+++ b/tests/utils/test_projections.py
@@ -141,7 +141,7 @@ def test_zoom_to_fit_orthographic() -> None:
     )
 
 
-def test_zoom_to_fit_orthographic_preserving_aspect_ratio() -> None:
+def test_zoom_to_fit_orthographic_letterbox() -> None:
     # Put a view on a (square) canvas
     canvas = snx.Canvas(width=400, height=400)
     view = snx.View(
@@ -161,9 +161,9 @@ def test_zoom_to_fit_orthographic_preserving_aspect_ratio() -> None:
     # ...and [100, 200, 0] to NDC coordinates [1, 1]
     assert np.array_equal((1, 1), tform.map((100, 200, 0))[:2])
 
-    # Now with aspect preservation the longer axis (y) should fit exactly, and the
+    # Now with letterbox=True, the longer axis (y) should fit exactly, and the
     # shorter axis (x) should have some padding to maintain aspect ratio
-    zoom_to_fit(view, type="orthographic", preserve_aspect_ratio=True)
+    zoom_to_fit(view, type="orthographic", letterbox=True)
     # Assert the camera is moved to the center of the scene
     assert view.camera.transform == snx.Transform().translated((50, 100, 0.5))
     # Projection that maps world space to canvas coordinates
@@ -220,6 +220,46 @@ def test_zoom_to_fit_perspective() -> None:
     assert np.all(_project(tform, (100, 100, 0)) < _project(tform, (100, 100, 1)))
 
 
+def test_zoom_to_fit_perspective_letterbox() -> None:
+    view = snx.View(
+        scene=snx.Scene(
+            children=[snx.Points(vertices=np.asarray([[0, 100, 1], [100, 0, 0]]))]
+        )
+    )
+    # Put the view on a non-square canvas
+    canvas = snx.Canvas(width=800, height=400, views=[view])
+    zoom_to_fit(view, type="perspective", letterbox=True)
+
+    # Assert the camera is moved to the center of the scene
+    # Depth isn't particularly important to test here.
+    assert np.array_equal(
+        view.camera.transform,
+        Transform().translated((50, 50, view.camera.transform.root[3, 2])),
+    )
+    # Projection that maps world space to canvas coordinates
+    tform = view.camera.projection @ view.camera.transform.inv().T
+    # Assert the camera projects [0, 0, 0] to NDC coordinates [-1, -1]
+    assert np.allclose((-1, -1), _project(tform, (-50, 0, 1))[:2])
+    # ...and [100, 100, 0] to NDC coordinates [1, 1]
+    assert np.allclose((1, 1), _project(tform, (150, 100, 1))[:2])
+
+    # Now, do the same thing but with height > width
+    canvas.size = (400, 800)
+    zoom_to_fit(view, type="perspective", letterbox=True)
+    # Assert the camera is moved to the center of the scene
+    # Depth isn't particularly important to test here.
+    assert np.array_equal(
+        view.camera.transform,
+        Transform().translated((50, 50, view.camera.transform.root[3, 2])),
+    )
+    # Projection that maps world space to canvas coordinates
+    tform = view.camera.projection @ view.camera.transform.inv().T
+    # Assert the camera projects [0, 0, 0] to NDC coordinates [-1, -1]
+    assert np.allclose((-1, -1), _project(tform, (0, -50, 1))[:2])
+    # ...and [100, 100, 0] to NDC coordinates [1, 1]
+    assert np.allclose((1, 1), _project(tform, (100, 150, 1))[:2])
+
+
 def _project(mat: snx.Transform, world_pos: tuple[float, float, float]) -> np.ndarray:
     # Inverting the behavior of vec_unproject
     proj = np.dot(mat.root, np.asarray((*world_pos, 1)))