Ok, there's a lot in there.. :-)
To tweak the tonal balance of already-recorded audio tracks you'll need EQâ€”usually parametric EQ. You might want to adjust the low frequencies (the bass)â€”around 100Hz or soâ€”lowering them by a few dB (deciBels) on the deep-voiced speaker. If the software you're using provides audio processing plug-ins, most any EQ will do.
Regarding any level inconsistencies, a Compressor plug-in can automatically control levels, evening them out from start to finishâ€”IF it's set correctly. Unfortunately compressors can be one of the most difficult studio tools to really master. Again, if the software you're working in has audio plug-ins, try a compressorâ€”if it has one of those compressor plug-ins that emulate vintage hardware compressors, try a model of the Teletronix LA-2A. That model is easy to use, and is ideal for vocalsâ€”it has just two knobs, amount of compression and overall output level. If the software has only a generic virtual compressor, try setting the Ratio at around 4:1, and the Threshold so the (gain reduction) meter jumps mostly on the louder lines.
As to the leakageâ€”the presence of the other speaker's voice in the wrong micâ€”well, there may not be too much you can do about that.. If the unwanted voice is only noticeable when the desired voice in that mic is not speaking, then you can use a Gate to cut it out, by setting the gate's threshold so it kicks in when the desired speaker (in each channel) pauses. If the unwanted voice is audible while the desired voice is speakingâ€”and IF it sounds bad ("phasey", nasal) when it combines with the other trackâ€”well, the only possible way to get rid of it might a high-end Audio Repair software package like iZotope's RX7, which includes a module to (try to) eliminate leakage.
Here are a couple of links to some of our courses that discuss these topicsâ€”the first one is a basic introduction that covers, EQ, Compression, and Gating; the second is a more advanced course on RX7 (the relevant modules for what you describe would be (11) De-bleed and (13) Spectral Repair (Voice)):