Participants 2024
About the TEAM
Meet two brothers combining science and creativity. Carli Nistal, a renowned hip-hop producer from Spain’s north coast. A versatile bass player, sound engineer, and DJ, Carli’s eclectic production style spans local and international collaborations with artists like Steele and Buckshot. His decade-long career reflects his boundless creativity and talent.
His brother, Javier, a former musician turned sound signal processing and machine-learning expert, comes from a research-oriented family. He merged his love for music and technology, specialising in AI-driven music tech, and is the brain behind innovations like DrumGAN and Diff-A-Riff. His journey took him from Madrid to Barcelona, then to Manchester and Paris, driven by curiosity and research.
Together, these brothers blend technical expertise and artistic flair, making them a formidable team in the AI Music contest.
About the SONG
Introducing 'Toxic,' a vibrant blend of disco and house pop crafted by two brothers with a shared passion for music and technology. Leveraging the cutting-edge capabilities of Diff-A-Riff, an advanced AI model, they seamlessly fused classic rhythms with modern grooves. Impressively, the team and the AI co-created every main track in 'Toxic,' showcasing the power of artificial intelligence in music.
The track kicks off with an infectious drum pattern, synth bass, and piano chords, laying a solid foundation. An electric guitar riff adds a touch of soul, while a funky bass line weaves through the beat, creating an irresistible groove. Layers of rhythm and arpeggiated piano tracks, along with a tambourine, build up the energy, culminating in a rich, polished arrangement.
Drawing inspiration from the 70s, the brothers added strings reminiscent of Barry White and Isaac Hayes, infusing "Toxic" with an epic disco vibe. The result is a dynamic, toe-tapping track that merges nostalgic influences with state-of-the-art technology, showcasing the magic that happens when creativity and innovation collide. Enjoy the seamless blend of old and new, brought to life by a duo dedicated to pushing the boundaries of music, all through the power of AI.
About the HUMAN-AI PROCESS
To create "Toxic," we employed Diff-A-Riff, an accompaniment generation model that adapts to user-provided contexts. Diff-A-Riff is a latent diffusion model with state-of-the-art audio quality, enabling various control mechanisms such as text/audio prompting and loop generation.
We started with a simple loop inspired by mainstream disco/house pop, consisting of a drum pattern, synth bass, and piano chords, providing a rhythmic and harmonic foundation. Using Diff-A-Riff, we prompted it with an electric guitar sample we liked as a style reference.
Next, we created a drum loop using the guitar, bass, and metronome as context and a house drum loop as the target. This resulted in a clean and polished drum track.
For the bass line, we used a funky electric bass as the target reference, and after several tries and some editing, we produced a satisfactory one. We observed that Diff-A-Riff often produces similar results with the same context/target combination, which is helpful for layering but requires careful parameter adjustments for variety.
With drums, bass, and guitar in place, we generated rhythm, arpeggiated piano tracks, and a tambourine track using loops from our library. We combined these elements and sequenced a polished arrangement, adding transitions, FX, and extra drum and bass layers for added depth. The track evoked a 70s disco vibe, so we incorporated strings inspired by Barry White and Isaac Hayes. The generated string tracks had an epic disco feel despite minor glitches, which were unnoticeable in the final mix.