What are good practices when retopologizing text?

Question

I am trying to add proper topology to text so I can do a beveled edge and a boolean extrude from a shape later on.

I am following this video, however I find it pretty confusing when it comes to applying the steps to my file.

In the video, the shapes always have the perfect amount of vertices on each side for quads to be generated, but in my project that is not the case:

What I am doing doesn't look right, some faces are triangles, some are quads, the distribution of which is irregular, and I can't really understand what I could be doing differently except adding a bunch of extra vertices randomly, for which there doesn't seem to be a keyboard shortcut in my version. Simply put, it doesn't look good.

From one letter to the next, I have either a lot of faces or very few, which might create inconsistencies later on.

Do I want to keep the amount of vertices low, or should I add a lot of them? How can the shapes have the same amount of vertices on each side of the shape to create quads without deformation?

Given the examples in the screenshots, what am I doing wrong and what is a fast workflow to create correct topology inside these letters?

I have been looking it up for a few hours, and the major changes from Blender 2 to Blender 4 make it difficult to find relevant answers. I do not want to use a Remesh modifier.

Answer 1

There's nothing special about it being text. You're showing us solidified planes. They happen to be in the shape of letters, but that doesn't matter.

The actively selected vertex is a "non-planar 5-pole": it connects five different faces, and those faces do not all lie in the same plane. This will affect C-C subdivision, giving us the divot we see, and, depending on how it gets triangulated, it will affect smooth shading. This is the error you are making in your pictures. (Sometimes, you're making non-planar 6-poles instead, which are even worse.)

Instead, we can do this:

All of the non-planar bits are now 4-poles. I've added some 3-poles: naturally, places where only 3 faces join at a single vertex. 3-poles are not great either, when they're non-planar. But these 3-poles are planar, so it's fine.

Another way of saying this: I've created a control loop. A third way of saying this: my edge flow here follows the edges of the mesh.

Generally speaking, I'll do this by selecting all the connected faces in a single plane and using an 'i' inset faces operation. This will often require a bit of manual cleanup to get as much inset as I desire, without creating any overlap; a loop tools/relax operation is usually a good, quick fix for overlap fixes (loop tools is an addon bundled with Blender but not enabled by default.)

This single loop is fine if I am smooth shading or subdividing, but not necessarily if I'm doing both. If I'm doing both, I want two control loops for each plane:

Why? Because the subdivision will take the vertices of the first loop out of plane, so I need the second internal loop to keep my 3-poles planar.

Once I have these control loops, the rest of the plane no longer matters, except that its faces remain convex (concave faces can screw up subdivision.) I can pepper it with convex triangles and ngons and 42-poles and the render will never, ever change:

So: follow the edge flow. Use 1 or 2 control loops to hold mesh edges, depending on your needs for the mesh. Do as little work as necessary to make anything interior convex, but otherwise ignore the topology of the interior of planar surfaces.

We can look at this, in practice, on a solidified K if we want:

I am subdividing and smooth shading, so I have two loops that follow the shape of the K. We can see that doing things this way, I create no non-planar poles, and so the subdivision and shading never do anything I don't want them to.