I've been playing around with an online deep learning/neural network tool called SpecVGAN, and ran some low-quality Aleona's Tales sound effects through it, with arguably decent results. Since Wyrmsun uses a few sound effects from AT, I decided to contribute these:

• Lumber Mill -- reconstructed version (SpecVQGAN produces two variants of a sound, "original" and "reconstructed")
• building construction -- reconstructed version
• placement success -- reconstructed
• building explosions -- reconstructed

I think these came out quite well and are an improvement over the low-quality originals.

SpecVQGAN outputs WAV files in 22 kHz, I converted them to 44 kHz with Audacity to match the rest of Wyrmsun's sound files. This has not affected the quality of upscales in any way.