[douglas-carmichael]

The Contents... dialog is just as useful for a folder of files which each hold one preset as it is for a bank: it lists the preset names stored inside the files - which are not necessarily the names of the files - with their zones, key ranges and categories, everything can be auditioned before converting, and the conversion afterwards only reads the files of the ticked presets. The hard-coded list of multi-preset formats is therefore gone and the button is shown for every source format.

The tree grew a level for this: the folders below the source folder come first, then the file - but only when it holds more than the one preset, otherwise a folder of presets would be doubled by a one-child folder per file - then the containers inside of the file. The folders were part of the container path before and rendered below the file of a preset instead of above it. A preset which is named differently than its file additionally shows the file name in its row, so the two can still be matched.

This surfaced a real defect: the Sample Files format reports a folder as the source of a multi-sample, so a selection made in the Contents dialog converted nothing - the conversion ran the folder through the file detection, which filters by file ending, and finished with an empty output. AbstractDetector now detects a folder-shaped source as a folder, which also makes the audition work for it.

Measured with the harness reproducing the dialog's tree plus a real conversion of the selection:

Source	Contents run	Selection
Elektron Tonverk factory library (175 .tvpst)	175 presets, 6.5 s, grouped under their folder	4 ticked → 4 written, 177 ms, unticked files not read
Synthstrom Deluge pack (nested folders)	46 presets, 0.2 s	-
1206 loose WAVs (Sample Files)	81 multisamples, 1.0 s	1 ticked → 1 written (0 before the fix)
SF2 banks in sub-folders	242 presets, folder → file → preset	3 ticked across 2 banks → 3 written
E-mu EIII CD image (regression)	1255 presets, 1.1 s, image → bank → preset unchanged	-

Re-reading every entry through the audition path and comparing with the contents run: 175/175 Tonverk, 46/46 Deluge, 2/2 Sample Files - 0 mismatches, 0 unreadable.