MMM-synthrider
Final Audio Master
AI Song Contest 2022 Entry: monobor x mashmachine - synthrider
Motivation and Approach
Our objective is to critically and artistically evaluate the extent of MMM's musical capabilities.
The question is: "Can MMM be guided to generate excerpts that can musically stand on their own (with minimal to no editing) and eventually make a coherent composition?"
Showcase as transparently as possible MMM and Calliope's capabilities. The MMM model is directed to take control of the full authorship of creating the symbolic musical content of the piece with minimal to no editing. The generated parts/blocks are then arranged sequentially in the best way possible without altering their content (notes or tracks). We do cut the parts to complete earlier where necessary in order to allow a smooth transition to the next part.
Calliope's Generated Examples
We used Calliope, a web application for music composition using MMM, to interact with to the algorithm. We told it which instruments we want and how long a phrase should be. It generates 18-bar MIDI excerpts at a time, out of which we selected the best sounding pieces (shown below) for our theme.
Main Generated Theme:
Generated Variations - Group 1:
Generated Variations - Group 2:
Sound Design and Arrangement
The generated pieces were moved to Ableton for sound design and arrangement. The arrangement focused on creating a narrative since the pieces share the same theme. All sounds were produced using native Ableton synths and samples.
Credits
Music Generation: Multi-Track Music Machine (MMM)
Music Arrangement and Production: Renaud T. Bougueng and Philippe Pasquier
Art Direction: Renaud T. Bougueng and Philippe Pasquier