Participants 2024

Rabbit Hole, the compelling tale of toxic internet misadventure, includes both operatic song interludes and rap verse. Assembling the track took several weeks as a collaborative effort between all three team members Don Bones, Katherine Canoville, and Harry Coney. The topic was comprehensively researched using input into Chat_GPT to suggest relevant material to develop into suitable lyrics, the social message of the track as important as its production. Each member of the team had specified roles in track creation, however the composition of the lyrics were a team effort, regular meetings over Zoom, determining the course of the finished article.

Due to issues of unauthorised mimicry, plagiarisation, and cultural misappropriation, AI rap has understandably developed a “bad rap” over the last few years. We as Botuoso sought to put the medium to fair use in our project, using ethical sources and positive intent to portray the voice of a radicalised young impressionable man from England.

Our aim, as a team working in tandem with artificial intelligence technology, was to make the vocals seem ultra-realistic, while simplifying the creative process using tools which anybody can access with ease, inspiring others to follow suit. Rather than training AI programs to emulate our own voice patterns, we made use of the pre-trained voices provided by online text to voice programs.

Eleven Labs (elevenlabs.io) provide a vast array of worldwide voices, including a varied selection of informal UK voices impossible to find elsewhere, especially created by actors for sound generation. The particular voice we utilised, fittingly portraying the naivety of the character, the rap sections of Rabbit Hole, a confession of how he became radicalised through internet conspiracies into full-on extremism.

The voice emulator wasn’t a rap generator, but rather, we adapted ordinary spoken sentences into the rap format, where necessary, alternating pitch and adding echo and reverb to give the impression of continuous flow. Line after line of vocals were created separately and painstakingly coordinated with the beat to give the impression of a genuine performance. Some words and phrases required speeding-up, while others were reduced in tempo, the result, a human-AI collaboration which sounds like a purely human rap, which is a significant achievement in itself.

The operatic song sections were provided from another text to song application called Voicemod ( https://tuna.voicemod.net/text-to-song), a Welsh tenor vocalist “Joe” (a paid professional voice actor), ably singing the chorus of Rabbit Hole from basic text input, providing eerily deep vocals between the spoken word.

The music video we especially created for the track, was a human-AI collaboration in itself, making use of Magic Studio’s Art Generator (https://magicstudio.com/ai-art-generator/), frames edited and linked to create a memorable rabbit-themed accompaniment to the music.