The sound of crungus
Music and speech models from Replicate's music collection and a curated selection were given “crungus” and “the sound of a crungus”. We request twenty-second clips where supported; duration otherwise depends on the endpoint. Clips are loudness-normalised for playback. Audio consistency measures similarity within each model's clips using CLAP. 310 clips from 16 models.
tune in to crungus radio → every clip, shuffled, over every image
The sound archive
order by