Artifact 87: や すら わで (Ya sura wade) feat. Hatsune Miku (folk-rnn v2 + Déguernel)

This artifact is contributed by Society Member Déguernel. He writes: There is something about a cappella airs that I find particularly moving. Maybe it’s the courage it takes to sing one, maybe it’s the intimacy it creates. But more likely it is the solitude it evokes: from a serene peacefulness to a profound loneliness. A

Week 83: Don’t Ask Me Why (folk-rnn v40atonal + Déguernel)

MUSAiC postdoc Ken Déguernel asked a special folk-rnn model to generate something in 12/8 with a dorian mode. That special model is v2 (trained in Irish traditional tunes from thesession.org) but fine tuned for 40 epochs on thousands of algorithmically generated twelve-tone row compositions. (He didn’t know this.) Ken embraced the craziness that resulted and

Week 79: e-clipse (Tradformer + Sturm/Déguernel)

Here’s a different one. Tradformer created a nice jig, and then I made it a 5/8 “clipped jig”. Then Ken comes along and plays with the timing but on longer time scales in both parts. The title comes from Ken, probably to emphasize that the root is E.

Week 78: Scherzo (Sturm)

We now reach the first scherzo of “The shadow still lay where he had been standing” – a semi-autobiographical opera for YouTube, exploring ways of co-creating with a variety of AI “partners” — now recasting other operatic work on YouTube. In the present case, the AI partners are snapchat filters and nothing else. Future episodes

Week 76: Arthur goes to the church of The Electrician (Music Transformer + VQGAN-CLIP + GPT + UberDuck + Sturm)

These are the third and fourth scenes of “The shadow still lay where he had been standing” – a digital multimedia semi-autobiographical opera for online streaming in which I am experimenting with co-creation using a variety of AI “partners”. The third scene summarizes where we’ve been so far, and provides a bit more of Arthur’s

Week 70: The Electrician (Sturm + Music Transformer + VQGAN-CLIP + GPT + UberDuck)

This is the second scene of “The shadow still lay where he had been standing” – a long-form composition experimenting with co-creation using a variety of AI “partners”. The melodic material of this scene comes from an output of Music Transformer given a musification of morse code for “electrician”: ./.-.././-.-./-/.-./../-.-./../.-/-./. Dots are one note, dashes

Week 69: The shadow still lay where he had been standing (folk-rnn v2 + Music Transformer + GPT + VQGAN-CLIP + Sturm)

In “They are everywhere” and “A slow waltz” I began forging partnerships with a variety of machine learning models: folk-rnn to create melodic material, GPT to create topics and texts, and VQGAN-CLIP to create imagery. In this piece I add Music Transformer to the mix to create an arrangement of the melody, and then score

Week 66: A slow waltz (folk-rnn v Swedish + GPT + VQGAN-CLIP + Sturm)

In “They are everywhere“, I explored using GPT to create a text to accompany my playing. I ended up with a charming/disturbing experience that I had no intention of creating. This piece continues my explorations with a script created by iterating with GPT, and then generating visuals VQGAN-CLIP using prompts from the script, e.g., “She

Week 64: They are everywhere (folk-rnn v2 + GPT + Sturm)

This tune comes from folk-rnn (v2), which I feel is a jig created from an old aire. So I decided to query GPT for some text. Then it all came together very quickly. It used to be me.The little girl who reads at bedtime,under the covers with a flashlight,against my parents’ wishes. I liked the

Week 42: Heading back home to finally bury mom (folk-rnn v2 + Sturm)

I learned this piece from folk-rnn (v2). It appears as no. 28984 on pg. 4293 of 58,105 Irish Style Double Jigs. I tried it as a fast jig, which was alright, but I felt I needed to play it as a kind of air – one describing my feelings over finally heading home (this week,