As this is for workflow only, I might not put the same precision in it as I would — if it is for final use. I care here about the steps I do to get those things done. After all, I have to make certain that no one expects me to do someone else’s work ;o) I hope we can agree in that point. Good. The file available via a link, see PM.
…this is just one method. ;o)
Inspection of the model
The provided model is a wild mix out of Triangles, Quads and N-Gons. All Normals are equally directed, but most polygons are non-planar.
For a clean mesh, with Subdivision Surface use later on, one should focus on quads, or if triangles are used, that those weight the center point equally. N-Gons are always only a intermediate step, not a final option, if one likes control over the model.
The size of each polygon, except for some frame polygons are nicely balanced,but the frames are long.
To use Bevel Deformer is possible, but since it introduce virtual points, I’m not really happy with those. But possible, even they limit more during that stage. Since they affect the area where the Batter-Lit has to be, not my way of modeling.
The provide low res model is a good start to work with, the subdivided version is useless for the next target (I guess it was more to check the current progress), but since we have the low res, all is fine, even if it needs the translation of the “Bevel” into polygons.
I do those examinations to get familiar with the mesh, after checking what the target is, and this work allows me to “think” about the next steps. I write all of that down, to support everyones work, not to critique anything.
First, I disabled all the Bevels, and checked one by one which one is really influencing the area we care about here. It is the first and third one, so I will shut all off, and recreate the Bevels 1 and 3 manually. This will allow me to solve some triangles into quads as well. The bevel-objects react to the given mesh, and most changes there might result somewhere to something I wouldn’t want. I can’t even see the results of those, while in Polygon mode. It makes it difficult to work at that stage. (I like Bevel Obj, don’t get me wrong).
Bevel Deformer #1 resulted on one end to create out of a quad an N-Gon and on the other end “exploded” in an “Complex point” resulting into a triangle cluster. However, as this is not the real point in question, I have only worked on the parts I needed. Perhaps the model is fine for a certain use, even with all the N-Gons in it, but I prefer to model with quads, and allow triangles only if there is no way around (happens, if re-start modeling isn’t possible).
I created then the Edge Splines as a guide, as there was none in the file. I set the given Sketch Texture to a higher resolution in the Material>Editor>Texture Preview Size. Given the bold stroke, 4MB was plenty here.
Based on this three (Back, Front, Side [camera based]) Splines, I was able to see what kind of mesh was needed. I copied the left to the right side.
The main decision would be now, a) to re-topologize the mesh and work with the sub-divided “previous results” and get a low res mesh that comes already close to the needs of the edges (body vs lit), or b) to work with what is given. A re-topologizing is discussed in a Cineversity tutorial (Pen Tool) so I will not go that route (even that would be best!), and instead discuss another route. Via subdividing the existent mesh.
There are two options: One to subdivide the mesh completely (but step by step!), with Mesh>Commands>Subdivide [Smooth Subdivision] or follow the spline and cut the crossing edges, to get the needed information into the mesh.
The Smooth option has the advantage that the mesh will be very close to the Subdivision Surface Object results, and with a lower level needed in the Subdivision Surface [faster] .
The Edge option will create an more flat area, if the Subdivision Surface Object would create a round surface there. Which can result in the need for supportive splines (guide/snapping) as well for this to counter the flat version. As this is a small object, I go with the first option. It produces more polygons, but as the total amount stays the same, the more efficient edge-cutting will be not used, and lots of time is saved. It’s a judgement call; Make your pick. I create copies always, call them BU [back up] and move further. Autosave is on as well, but I love to have it right inside the scene.
I decided on this level (first subdiv), which points can be moved, and perhaps what points will be available during the next iteration of Smooth Subdivision.
Moving points (Slide) along an edge is a one time action, never ever (except neighbor edges are on the same flat plane) use the slide tool again to move points again or back, cmd Z and then do it again, never twice. The results might be unusable.
On the second subdivision level, again only sub-div “Level 1” each time, I have to adjust previous and new points. Especially the neighbor points are crucial, as they will determine the next sub-division level and with that the position of the new points. To be sloppy here, will cause more work later.
For long round lower edge I decided to go with two triangles, and not subdivide further. This is an ugly short cut, and needs to be handled carefully, on the last and 3rd subdivision. I introduce this here more for variance reason. A new topology would be my wish, less resolution in one and more in the other direction of that roundish object, but here you go. I left the top edge (which is similar) alone, so you might explore it … :o)
On the back, the two half round, I roughly laid out the edge and then used an extrude inner to support the roundish edge (image). It needs more care.
These steps, once done one time through, hopefully allows to see how much each step has influence into the next. This will save time, to not go too crazy in each, but supports the work in the next one.
My best wishes