Lofting: can I copy a Shape-Keyed mesh for each frame of its animation?

• I have made a simple profile mesh (an edge loop) which I've animated along a path using 'Follow Path'.
• It has a number of Shape Keys, and as it moves along the path, it changes shape.
• I would like to capture a copy of the full state of the mesh, in place, at each frame of its animation.

This would be, if you like, a sort of 'physical onion skinning'. Dupliframes capture the transformations of the mesh at each frame, including tilts and scales derived from the path, varying those along the path, so come very close. But they don't capture the Shape Keys. If I alter the shape, it changes in all the duplicates,irrespective of keyed frames.

I'm attempting this in the hope of modeling with the resulting copies: creating a mechanism for lofting between varying profiles along a curve with a reasonable degree of accuracy, and ease of workflow.. so if there's an alternative route to that, please comment.

• Think I have something similar re this question A follow path constraint gives an easy way to put shape at either end, and fill in the "ribs". Be simple enough to convert from offsets to frames .Could you provide a small sample file? Nov 8 '18 at 11:03

Bmesh script

Select the profile mesh, run script. Creates a copy for each frame from scene frame start to scene frame end.

A simple edge loop animated with follow path constraint, scale, and a mix of two shape keys. BMesh.from_object gives a mesh snapshot at that frame (Apologies for low quality gif)

import bpy
import bmesh
from mathutils import Vector, Matrix
context = bpy.context

dg = context.evaluated_depsgraph_get()
scene = context.scene
ob = context.object
me = ob.data
bm = bmesh.new()
f = scene.frame_start
step = 1
while f <= scene.frame_end:
scene.frame_set(f)
bm.from_object(ob, dg)
rme = bpy.data.meshes.new("Rib")
bm.to_mesh(rme)

copy = bpy.data.objects.new("Rib", rme)
copy.matrix_world = ob.matrix_world
bm.clear()  # interesting without.
f += step


Skinning the thing.

Rather than having an object per rib, could also make one object. Taking advantage of the order each rib ring is added, can bridge these edgeloops with bmesh.ops.bridge_loops Could be done on a lower level by creating faces.

import bpy
import bmesh
from mathutils import Vector, Matrix
context = bpy.context

dg = context.evaluated_depsgraph_get()
scene = context.scene
coll = context.collection
ob = context.object
mw = ob.matrix_world.copy()
me = ob.data
nverts = len(me.vertices)
nedges = len(me.edges)
bm = bmesh.new()
f = scene.frame_start
step = 1
rings = []
while f <= scene.frame_end:
scene.frame_set(f)
bm.from_object(ob, dg)
bmesh.ops.transform(bm, verts=bm.verts[-nverts:], matrix=ob.matrix_world)

rings.append(bm.edges[-nedges:])
f += step

# build from rings
next = rings.pop()
while rings:
ring = rings.pop()
bmesh.ops.bridge_loops(bm, edges=ring + next)
next = ring

rme = bpy.data.meshes.new("Rib")
bm.to_mesh(rme)
copy = bpy.data.objects.new("Rib", rme)
#copy.matrix_world = mw