I've been like a mad scientist the past few days playing with Synplant 2. It's a worthy upgrade from v1. Genopatch is a lovely endless inspiration source, and using the new Bulb Modes is great fun. But I keep feeling like Synplant's strictly seed-based mutation scheme is not sufficient for creative expression. I would love to be able to tweak the genes for individual branches as I see fit, which doesn't seem to be possible. (Perhaps I missed a passage about this exact topic in the manual, though I have read it through twice now.)

Discrete branch editing would be ideal in scenarios using range Bulb Mode, allowing for (for example) pitched drum / percussion kits of bass drums, timpani, toms, congas, bongos, celesta, and glockenspiel. It's hard to get all those sonic distinctions right just by mutating a single gene set.

Discrete branches would also be ideal in velocity Bulb Mode. One might imitate a string section, where low velocities could be pizzicato and spiccato, middle velocities could be various subtle iterations of arco, and high velocities could be marcato / martelé sounds with varying amounts of inharmonicity. Again, difficult to get such a specific result without fully autonomous branch genes.

All that said, I understand if this isn't in line with Magnus and Fredrik's vision for Synplant. Feature creep can ruin a good product, and Synplant 2 is already good! — But I'd like to submit my vote for implementing per-branch editing, if you have been considering it.

I think that somebody could create scripts for solving such challenges

If the branch seed values in the preset file referenced an entire set of genes, then yes, that would be a fine solution. Glancing at the preset format, I'm not sure the branch seed values contain enough data to directly correspond to a set of genes. In an unmodified patch, branch seed values seem to occur in order (for example, 0x20a1a20c, 0x20a1a20d … to 0x20a1a217) and yet each branch produces wildly different results when mutated.

Perhaps seed values point to "macro" tables in which each value corresponds to pseudo-randomized offsets for specific combinations of genes, so that all sounds mutate in "useful" ways. (Probably not. I know nothing. lol) At any rate, I'd be very interested to know how it works.