AudioLDM
The AudioLDM family uses latent diffusion for text-to-audio, audio-to-audio, text-guided style transfer, super-resolution, and inpainting across speech, ambience, music, and short effects; AudioLDM2 continues the research line with public weights.
Why use it
It is selected because it supports structured experiments around an existing sound instead of only returning a complete clip: designers can preserve an event while changing texture, repair a gap, increase resolution, or compare random seeds to test footstep materials, ambience layers, and abstract ability sounds cheaply, while researchers can inspect the actual model and inference code.
Where it fits
Use art, animation, audio, levels, and narrative to make the game easy to read and its world convincing.
What to check
Official repositories and weights use CC BY-NC-SA 4.0, permitting only non-commercial use and applying share-alike duties to adaptations; the original project also says its AudioSet, Freesound, and BBC Sound Effects demo relies on a UK academic-research copyright exception, which cannot support a commercial-clearance claim. It remains research software with older environment and GPU requirements, so commercial projects need a differently licensed model.