Ambient music from Flex Council

Flex Council sprang to life by chance – a YouTube interview between Andrew Ng and Ian Goodfellow revealed the vast creative potential of generative adversarial networks, a type of AI algorithm based on competition.

What followed was a two-year research and production process which involved learning Python and machine learning theory, surveying the top algorithms, creating datasets, rendering audio, and finally, putting it all together in Ableton Live. The finished tracks reflect the genres of the datasets they use, namely slowcore, vaporwave, indie rock, folk, hip-hop, and classic rock.

Commenting on the track, the artist offers, “December Song is a departure from my other Jukebox-based tunes in that it derives from multiple songs rather than a single song. The key to getting everything to fit together was matching the tempos and key signatures, in this case 75 bpm and G Major (or a related key.) I was also looking for chord sequences similar to the “‘0’s progression’, which is I-VI-IV-V, like in Earth Angel and Stand by Me, a progression which most people would characterise as sentimental or melancholic.”

Adding, “This was useful to narrow down the emotional scope of the track. For lyrics, I consulted an online lyrics generator. This particular generator allows you set the tone of the lyrics, which I set to very sad.”

“I also used a source separation algorithm called Spleeter to isolate the vocals in the Jukebox runs. Audio restoration software like Steinberg’s SpectralLayers and Zynaptic’s Unchirp/Unfilter helped to reduce some of the sonic artefacts. The Jukebox models are from Cocteau Twins, The Cure, Ashanti, Tricky, Stevie Nicks, and The Weeknd.”

With this innovative exploration of computer music, Flex Council delivers a deeply atmospheric, ambient listening experience. Sparse textures weave between percussive shots as the instrumentation delicately swells in a lullaby ebb and flow.

Listen to December Song below

More from Music

---

Editor Picks

---