So I’ve found a number of posts here and there saying you can’t do a proper Vangelis lead on the Deepmind 12 because you can’t precisely detune two saw waves. Most of them go on to approximate it by using the unison detune with the voice mode set to unison-2.
Typical settings are something in the ballpark of ~2Hz triangle LFO doing a bit (16ish on DM12) of pitch mod, and the two saws detuned by a few cents. Then slow down the attack a bit and add some long reverb. Here most people just turn off everything but the saw and set it to unison-2 as mentioned. But the DM12 has a very nice selection of mod sources, including “uni voice.” Nobody seems to know what this does, but I can tell you: with unison on, each voice after the first is numbered: 1,2,3,4… (the first is zero) and the mod amount is multiplied by this number. That means if you want to detune two saws by, say, 4 cents, set the source to uni voice, amount to 4 and destination to osc1 fine. The main voice is unaffected (mod amount multiplied by zero, essentially) but the second voice gets detuned by 1x4 cents.
This also works for octave stacks: +12, osc1+2 pitch. You can also use this kind of thing to emulate parallel filters, by offsetting the settings for the second filter bank in the mod matrix. I saw someone trying to use “voice num” for that and it was a big mess.
I know some of you have a DM12, so I hope this info proves useful.