In this diagram, the servo at A only moves the orange piece ("foot") left or right, and the distance part of distance-times-weight is almost zero. Thus, very little torque is actually needed.
The servo at C is mounted to move the legs horizontally (hither or yon, into or out of the drawing) and there is no real load in that direction, thus that servo also needs very little torque.
Remains servo at B, the "knee" servo. When the rest of the body is still, it lifts the orange and green pieces, as well as the servo A itself. The center of gravity is probably somewhere towards the end of the green piece and a bit down. The horizontal distance from COG to the servo is thus shorter than the green piece. The torque needed for this lifting is weight of green plus orange plus servo A, times approximately two thirds the length of the green piece.
Now, the other servos in position B, on the other legs, keeping the body up, are each lifting 1/3 of the weight of blue leg pieces plus purple body plus servo C. This is assuming each of the legs (green + orange) are fixed, in contact with the ground. Each of them see a center of gravity approximately at servo C (depending on weight distribution between purple and blue, could be closer.) This may be a bit bigger than a servo B sees when it's lifting the outer limb part.
And this is simplified -- I'm not a mechanical engineer, so the statics and dynamics math for the actual solution isn't something I know how to solve, but this "rough estimate" works for several others that have done walkers, so I believe it :-)
So, torque for servo B is the maximum of either:
weight(green + A + orange) * length(2/3 of green)
weight(4/3 * blue + 4/3 * C + 1/2 * purple) * length(blue)
If you lift two legs at a time, crosswise, then the second term becomes:
weight(2 * blue + 2 * C + 1/2 * purple) * length(blue)
And, again, add a 60-100% safety margin.
All distances are measured horizontally only (projected to the plane perpendicular to gravity, if you're good with 3D geometry.)