Portrait de Skygge

Skygge, inspiration accelerator

Ben­oit Car­ré was invited to Trempo as part of Slash to extol the vir­tues of arti­fi­cial intelligence.
Along­side the musi­cian, the eight artists of this European train­ing pro­gram dis­covered how the tech­no­logy could help cre­at­ors in terms of inspir­a­tion and com­pos­i­tion. Under the monik­er Skygge, he remains one of the pion­eers in this field – so much so that pres­ti­gi­ous artists such as Stro­mae or The Pirou­ettes are rush­ing to col­lab­or­ate with him.

No one ever knows what to expect before meet­ing Skygge. All they do know is that he’s one of the apostles of “arti­fi­cial intel­li­gence applied to music”, and that he works with Spotify’s CTRL-labs. When he walked into Trempo’s offices on that frosty Janu­ary 18th, the man did not fit the ste­reo­type of a mad sci­ent­ist who escaped from some under­ground work­shop. The 50-year-old, with his salt-and-pep­per hair (well, mostly salt), imme­di­ately cor­rec­ted us: no, he is not a research­er, but a musi­cian. He had attained suc­cess with his former group Lilicub and their hit Voy­age en Italie that any­one over 40 has heard. “I star­ted mak­ing music with a rel­ic: the Com­modore 64. I also wrote a lot for oth­er people: Françoise Hardy, Imany, Michel Sar­dou, Johnny Hally­day… Par­al­lel to my career as a song­writer, I star­ted to spend time at the Sony CSL Lab, where eight research­ers were design­ing tools based on AI. They liked the way I was com­pos­ing and wanted my feed­back. I espe­cially like the har­mon­ic sur­prises.” The Flow-Machines pro­ject led by François Pachet and his team of research­ers is now hos­ted by Spo­ti­fy with the goal of mod­el­ing and repro­du­cing music­al styles. “For three years, I par­ti­cip­ated in the devel­op­ment of this tool until I was com­pletely immersed in it.”

“You take your musical fantasies and the machine offers suggestions” 

His­tor­ic­ally, A.I. was born in the 1950s when math­em­atician Alan Tur­ing wondered if a machine could “think”. Today, we are still in the early stages of how this tech­no­logy can be applied to the music. And Ben­oit Car­ré is cer­tainly a pion­eer, who chose “Skygge” for his stage name (“shad­ow” in Dan­ish), in ref­er­ence to a Hans Chris­ti­an Ander­sen tale where a scientist’s shad­ow becomes human when it comes into con­tact with poetry. In 2018, he opened his Flow-Machines stu­dio to adven­tur­ous musi­cians, includ­ing Stro­mae, The Pirou­ettes, Laurent Bardainne, Médéric Col­lignon, and Michael Lovett, the gui­tar­ist-key­board­ist of Met­ro­nomy. With this merry band, Skygge com­posed the album Hello World in 2018, which uses arti­fi­cial intel­li­gence. “I wrote for myself, then invited people from very dif­fer­ent worlds. I was able to indulge in new exper­i­ments and explore a lot of things. For example, you can infuse your music with har­mon­ic arrange­ments. It’s then ana­lyzed and trans­formed by the machine: you have the begin­ning of a song, you add your music­al fantas­ies and the machine offers you dif­fer­ent choices.”

“We remain in charge of the process, because we’re the ones providing it with content”

Does this mean that, with these new tools, stu­dio musi­cians can now go and have a drink at the bar next door while the machine works for them? “No,” retorts Ben­oit Car­ré. “It’s an inspir­a­tion-accel­er­at­or, but it will nev­er write a song for you. It just offers ideas that can be integ­rated into a musician’s work. It gives you hori­zons, but it doesn’t take any­thing away from your work at all. Besides, as things cur­rently stand, the res­ult is zero at 80%. At 18%, it’s inter­est­ing. And only the oth­er 2% are really great. I would say that, on the con­trary, you have to be much more present. It’s a jungle of sound and har­mony. Yes, I’d say I carved my way through a jungle to find my path.” He con­tin­ued this explor­a­tion in 2019 with his EP Amer­ic­an Folk Songs. “It’s an inter­est­ing example,” explains Skygge. “We were devel­op­ing a pro­to­type for orches­tra­tion that worked with sheet music. I took Pete Seeger’s voice in the track Black Is The Col­or. I just had a melody, then injec­ted bossa nova har­mon­ies I liked and it cre­ated beau­ti­ful har­mon­iz­a­tion. We’re always in con­trol of the pro­cess because we’re the ones adding the mater­i­al. We remain fully involved at all stages. Arti­fi­cial intel­li­gence is not cap­able of cre­at­ing a two-minute song for you in the long run. To get ten seconds of music, it takes three hours of ren­der­ing, and very low qual­ity mono sound will be the res­ult.”

“Studying the reactions of other artists”

How­ever, we can’t res­ist ask­ing him the ques­tion: with A.I., won’t mediocre com­posers just end up design­ing hours and hours of elev­at­or music? “There will always be the same num­ber of people mak­ing crap – that won’t change. It won’t kill cre­ation. But, yes, it will prob­ably take up mar­ket share.” For now, this tool isn’t main­stream, and Skygge imme­di­ately accep­ted Trempo’s offer to work with the artists of Slash. “I’m inter­ested in see­ing how oth­er artists react, and how they use the tech­no­logy. After this event, I’ll leave them the access codes to the lab so they can con­tin­ue. I’ll give them the tools and the deal is to get feed­back. Maybe one day they’ll release a track that they partly cre­ated with these tools.” Giv­en the enthu­si­asm gen­er­ated by Ben­oit Carré’s event at Trempo mid-Janu­ary, there’s a good chance he’s right…

Inter­view : Sylvain Chantal
Photo : Elod­ie Daguin

Slash Program

Slash is a Europe-wide train­ing pro­gram build-up by Trem­po with the great sup­port of SACEM. It aims to train emer­ging pro­fes­sion­al musi­cians in their career devel­op­ment. It is co-fin­anced by the European Uni­on through its Cre­at­ive Europe pro­gram (Music Moves Europe).

The European Commission’s sup­port for the pro­duc­tion of this pub­lic­a­tion does not con­sti­tute an endorse­ment of the con­tents, which reflect the views only of the authors, and the Com­mis­sion can­not be held respons­ible for any use which may be made of the inform­a­tion con­tained therein.