When I first started using MLN or QFN packages, I didn't have solder paste, so I used the second method.
Firstly, I basically tinned every pad so that each pad wasn't completely flat but very so slightly raised (it really doesn't take much solder at all). Then I applied flux to the pads and the underside of the chip, including the centre pad, placed the chip on top, and reflowed with hot air using one of those cheap '898D' soldering stations.
Surface tension really is your friend and with flux, once up to temperature the chip will move into place. You can very carefully nudge the chip with the end of your tweezers and if it moves back into place (its now sitting on molten solder) you know its done. Check each connection with a continuity tester, one probe on the trace away from the chip, and the other on the side of the chip (QFN''s and NLF often have an exposed by of copper for that pin on the side).
I remember being rather nervous about it, but it took maybe 30 seconds total, 10 seconds to heat the chip up. In some ways, quicker than soldering say TQFP32, but there is always a chance that one pad didn't solder. In this case, you can add a tiny but more solder to the pad from the side (and I mean tiny), dab with flux, and reflow. As the chips pad and the PCB pad heats up, it will draw in the tiny solder ball you made on the side to under the pad. I have only had to do this once and that was just because I was being sloppy.
I have since used solder paste and had similar results, although it was a bit faster - just quite difficult to control the solder paste flow by hand. As the solder paste already contains flux you can usually just put thin lines across the pads, and four small blobs on the centre pad, plonk the chip on and heat with slow circular motion.