codex for musicians

codex for musicians
last month i applied and was accepted into the @OpenAIDevs codex ambassadors program. i am a lifelong working musician, composer, and teacher who loves making creative spaces for learning. over the next year, i will be organizing local monthly meetups in the city i currently live in (slc, usa).
there is so much that goes into making a career as an independent musician that is not music-making. here are some of the things i am excited to share with the community of musicians interested in using codex:
- owning your own content delivery network (cdn) has been the key to every successful working musician’s career for thousands of years... yes, way before the music industry. with codex, it is possible to create websites that organize, distribute, license, and manage subscriptions.
- access to music education from every culture and civilization across time in codex. you can build daily automations that surface whatever you are working on in music. these can be simple one-sentence facts or ongoing research into extremely discrete and nuanced areas of music.
- ai assisted composition is less scary than folks might think at first. when i was working at my first analog recording studio, i had 1 to 2 hours of prep work for each day’s session: setting up mics, zeroing the board, demagnetizing the tape heads, and writing out all of the cue sheets and even some charts for guest musicians. lots of that setup work is now digital and can be managed by codex, especially as you teach it more about your working style.
- audio analysis has been out of reach for many musicians, even though most of the software has been around for a very long time. codex can use python libraries like librosa to analyze my entire catalog of compositions and released recordings. from learning the tempos, keys, meters, and timbres i have chosen over the years, codex can take notes and apply that context as we work on other projects.
- local model training for new music on only my music can be used as an alternative to generative models trained on other people’s music. many generative models making whole tracks are using variations of generative midi, but raw audio is still compute-intensive for that. codex has helped me train models that generate small samples of sound that i can make new software instruments from.
there is so much more that crosses into running a small business and building internal software tools that i am excited to explore and share with folks! thank you to all of the people at @openai who are supporting so many artists and creatives right now. (big thanks to @gabrielchua and @reach_vb for all of the work going into growing this amazing community.)
still a bit nervous about ai stealing all of music and putting us musicians out of work? try this in any room around the world with any number of people: stand up and ask, “raise your hand if you think it is okay for folks to steal music?” nobody will raise their hand, and music has earned that global sentiment over thousands of years. other domains can look to musicians, as we have already learned how to work, adapt, and thrive through the kind of fragility they may now be facing.