DIY metal casting definitely can be done. Not everything is machinable shape wise, casting allows you to make things that are either impossible or very hard to make out of a single billet.
But like any other skill at this level of complexity it will take substantial dedication to be able to do it at a level where you can do it with a useable degree of (repeat) accuracy. Even real foundries test new molds and more often than not improve them after they learn how a particular mold behaves in practice.
Sand casting (the preferred method for low quantity bronze and cast iron runs) is error prone even in the best of cases but in the end it is very much worth it. Lost foam is another process that can work well for for instance aluminum (and is comparatively easy).
One reason why you really might want to stay from all of this is simply because it is far more dangerous than you might think when looking at the gear. But if you do go for it I would suggest starting with very simple castings and then to cut them open afterwards to check them for integrity and to learn what causes issues and how to avoid them. All the magic and the hard bits are in the prep work, the pouring is the easy part.
Forgive my ignorance, but what is there that's not machinable shape-wise, but is still castable? The more difficult the shape, certainly, the more difficult the casting?
You're not necessarily missing anything if you haven't done a lot of machining. But if you have then you probably know what I'm getting at. Tooling requires space to work, far more space than is sometimes available in the workpiece. So complex shapes, especially shapes with a lot of interior work and lots of detailing on all sides of the workpiece are much easier to make as a casting than as a milled or turned piece.
Every time you take a piece off the machine you need to re-register it, perfectly. This is doable but requires a ton of work and can severely constrain your design, after all, you'll have to re-clamp it after changing orientation.
You can easily see this by looking at for instance the output of 3D printers and then to try to think about how you would go about making the exact same pieces using substractive processes. Plenty of times you'll find that this is super hard or even that it can not be done at all.
Of course, the reverse is also true: there are pieces that you can mill and turn but that you can't make using an additive process. For every workpiece there is an optimum set of tools and processes to make it, some workpieces have a lot of options and some have only one.
When the number of solutions is 'zero' or when you have only some tools available and no access to others you have to get creative, maybe split the piece up, use a different material or come up with your own tools or processes.
> what is there that's not machinable shape-wise, but is still castable
Yes. For example imagine a 3d jewellery made of a web of thin filaments.
It can be non-machinable for two reasons: The wires can be too thin individually and machining them out of a solid block would break them, while they can support themselves once cast. Or you might have shapes where one of the wires is surrounded by the other wires and you can't get the tool in.
Traditionally, hand carve the ring out of wax and use a lost wax casting method. These days you can 3D print in a wax filament or machine the wax in multiple pieces, hand finish, then cast.
But like any other skill at this level of complexity it will take substantial dedication to be able to do it at a level where you can do it with a useable degree of (repeat) accuracy. Even real foundries test new molds and more often than not improve them after they learn how a particular mold behaves in practice.
Sand casting (the preferred method for low quantity bronze and cast iron runs) is error prone even in the best of cases but in the end it is very much worth it. Lost foam is another process that can work well for for instance aluminum (and is comparatively easy).
One reason why you really might want to stay from all of this is simply because it is far more dangerous than you might think when looking at the gear. But if you do go for it I would suggest starting with very simple castings and then to cut them open afterwards to check them for integrity and to learn what causes issues and how to avoid them. All the magic and the hard bits are in the prep work, the pouring is the easy part.