I prefer to hire musicians for productions. Real, live, creative guitarists, bassists, drummers.
Hire someone brilliant, and then get out of their way. They'll be brilliant.
But - when I was creating New Metal From Old Boxes (and This Is Not A Love Letter), the musicians I wanted to hire were busy. They're professionals! And professionals work.
And so, what to do? I needed to get this music out of my system, and out into the world, as quickly as possible. It'd been too long since my last production, and it was much more important to me to get a great album of music created and shipped than it was to wait until all of the stars aligned and I could get the perfect group of people on the thing.
And so, just me. I had to rely mostly on my technical and programming chops to create tracks that sounded cool. I didn't have a guitarist? Great - that created an opportunity for me to get creative with creating "guitar-like" sounds for the arrangement. And so, after creating a stack in the Native Instruments amp simulator (Guitar Rig 5 Pro), I dialed-up a few sounds: a sampled guitar, a sampled English Lute, a tone generated by a Korg Mono/Poly. The solos I played are unlike anything a guitarist would play (which was the point, right?) - they're evocative of a monophonic, distorted solo but very rooted in keyboard technique.
No bass player? Keyboard bass. Sounded good to me, and still does, and the performances are rooted in keyboards.
The drums required a lot of technical work. The software I used creates procedurally-generated drum sections: patterns and fills based on tempo, time signature, and so on. I hand-edited some sections and fills, depending on what I needed, but most of the parts are generated and then played through a sampled, huge drum kit in a beautiful room (Native Instruments again, Abbey Road Modern Drummer).
Although I didn't have access to the source code of the drummer software, I was able to take some educated guesses as to the input parameters I could tweak to get the performance I wanted. I've been around procedural generation of music before (in C and in Python programming languages), and had an idea of how tempo, time signature and key signature would effect the output of the program.
Here's a good example: if I handed a chart to a drummer that had alternating 5/4 and 6/4 measures, the drummer would create a part that fit the feel of 11/4, or 11/8 depending on the overall style of the song. But the software would produce wildly different outcomes based on whether I set the DAW to (5/4, 6/4), or (5/8. 6/8), or 11/8, or 11/4, or even (4/4, 4/4, 3/4).
In the end, I wanted to great music that I loved, and I did that. Would the album be better with live musicians? No. It would be different, but not better.
And what now? Well, for my next productions, I've hired some incredibly talented musicians. I can't wait until it's ready for the world, and I can't wait for everyone to hear these people. I'm sure I'll create more music that 100% "me", but it's much more interesting to get a bunch of musicians together and see what happens.