Funky Afro Diva Androids: Riffing With GenAI Tools
Earlier this year I created a series of images using the previous version of Midjourney (v5.2). The initial prompts came from an ChatGPT-generated poem about “funky afro diva androids.” The poem consists of several stanzas, each group of lines are read together and united around a theme or idea. I tried out including different stanzas as prompts in Midjourney, along with a combination of art styles and booster words. For example, I used this stanza as a prompt:
In the neon glow of a future night, Where circuitry hums and sparks ignite, Funky Afro Diva Androids come alive, In a realm where rhythms and circuits thrive.
In a sense, Midjourney was riffing on stanzas from the poem. The word riff (ex. to riff or riffing) has many connotations but, historically and traditionally, it refers to repeating chords or lines in music. In jazz, blues, and other musical genres, riffs are short rhythmic, melodic, or harmonic figures that are repeated to establish the framework of a song. Riffs are often paired with call-and-response to create cyclical patterns or iterations. Take, for example, the song “Boogie Chillen” by John Lee Hooker:
To “riff” means to improvise on a subject by extending a singular idea or inspiration into a practice or habit. Using generative AI tools to develop a concept such as “funky afro divas” is what I refer to as techno-riffing. In December 2022, Mubert used the open-source Stable Diffusion model to turn descriptive text into music loops. In January 2023, Google published a paper on their own text-to-music generator called MusicLM. In 2022, two developers fine-tuned Stable Diffusion to create a neural network called Riffusion that generates music using images of sound (spectrograms). In other words, it uses text prompts to generate images that can be converted into sounds.
This week I used the online Riffusion app to riff on the “funky afro divas” idea. I cut and pasted the first stanza of the ChatGPT-generated poem into the “lyrics and sound prompt” field (see above). The app recommends 5–20 words but my stanza has a few more and it still worked.
Next, Riffusion requires that you describe the sound, so I clicked on “Prompt Genius” and went with that. Then, I clicked “Riff” and the app generated a project title (“Neon Circuitry Funkadelic”) with three options, or image thumbnails, each with a different sound and vocal style. Below are the options I chose. You can listen by clicking the links below.
Here’s the same stanza from the poem translated into images and sounds:
There’s a remix option once the output images and sounds are displayed. Also, I can save them as videos, download them as MP3s, and even split them out as stems, which are a type of audio file that breaks down a track into individual mixes. This will allow me (or someone else) to control each of the particular mixes for production. Stems tend to break down into four tracks, usually covering the melody, instruments, bass, and drums.
Note: I address the concepts of remixing and riffing in my book Techno-Vernacular Creativity & Innovation. In the book, I address how these concepts or modes of TVC activity apply to computational action and thinking tools such as visualization, pattern recognition, body thinking, empathizing, dimensional thinking, modeling, and playing. I look forward to exploring this further in future courses I teach.