Is there a way to output a MIDI stream not to the speakers, but to some sort of a waveform buffer that could be used for DSP manipulation (like reverb or spatial filtering) before output to the speakers?

No, as a MIDI has no intrinsic sounds, as it is just a list of instructions for an instrument to play.

However, you can 'render' a MIDI to a wave form, by running it through software that has instruments already set up. I hear that Timidity can do this.

EDIT: Although another route, if you are already happy with whatever is producing your MIDI output, is to use a feature of your sound card to record whatever it is playing. Creative cards often offer the 'What You Hear' option in the mixer for this purpose.

Let me rephrase and expand my question a bit:

I send the instrument/note messages using the standard windows "midiOutShortMsg" and I chose the default windows synthesizer to produce the sounds (whatever is under "0" when initializing "midiOutOpen"). I'm not even sure if that synthesizer is part of the sound card software/hardware, or a software synthesizer built into windows.

Is there any way to capture the output of that synthesizer before it goes to the speaker? I know I can record the "what you hear" from the soundcard, but is there a way to do that without actually letting that audio reach the speakers?

The effect I need to obtain is synthesis of musical instruments, but processed through a number of custom filters (for spatial rendering).

Should I just abandon the MIDI concept alltogether and get into STK (synthesis toolkit) or just generate my own wavetable work with that?

Why do you have to cut it off at the pass, so to speak? Why can't you apply filters in post?

Look into an online tool kit called "Virtual Audio Cables" if you're on a PC.

This should let you route the audio output across virtual audio cables and you should be able to move the sound output of the MIDI renderer to other programs.

If you're on a mac, then your program choice is... I think... "Sound Jack" or "Sound FLower" ???

One of those, the mac folks can correct me, I don't own one myself.

Seems like the VAC might be what I need. I'll try the demo this week.

The problem with VAC is that it uses WDM, so I can't speak to the latency, but it's worth trying out.

You can use Cubase to do this and with the VST Instruments(inside Cubase) you can select your sounds and effects for each track.