There's no software alternative that would be directly compatible with any Radiants, but I've been trying the same thing and here's my method, although it's for a completely different project. This will only work for simple MTRs that contain a header and an editor image.
I take all the editor JPGs and rename every file to prefix their folders into the filename, and put them all together into 1 folder, then I start to categorise them into what they actually are, based on (real-world) materials and features, such as 'crates_wood' and 'crates_metal', etc. Importantly, I need to keep the original subfolders in these new categories, but luckily Radiant can handle fairly long headers with a deep folder structure, at least as much as I need it to.
Once that's all done, I remove the prefixes in the filenames with a script so to return their original filenames. The only reason for doing it that way is to keep track of which subfolders they belong to when they are all lumped together into 1 folder and I'm categorising them out.
Now I can make an index of all these files to turn the results into a text file, which can then be MTR files, just a list of headers. This is where Grep will come in handy, to take a simple list of headers and place the correct structure underneath to give it an editor image, or in my case just the "diffusemap" line, with the curly brackets on the lines above and below, so a list of headers becomes one big MTR file. I need to manually split this file into separate MTRs, but that's not so difficult since it's all alphabetised.
Once that's done, I simply do a search/replace operation to remove the "categories" from the prefix of every "diffusemap" line (so as to not affect the header) and I'm left with MTR scripts that have headers based on categories, but still uses the original file and location of the editor image, so I don't end up with another copy of the image for every category version.
So now in Radiant, I have all the originals but I also have extra "zzz_category" folders underneath, where mappers will be able to find what they're looking for based on substance and feature, instead of the original location and filename, which isn't often very intuitive. All that's required is to run a script on the .MAP file to remove the category prefix from the headers and return it to the original header before compiling/exporting. I don't find that Radiant takes so much longer to load or anything, and more importantly the size didn't increase by more than half a MB.
