Demo page for the ES-Instruct Master’s thesis project at KTH Royal Institute of Technology, in collaboration with Epidemic Sound.
ES-Instruct is an instruction-tuned text-to-music editing model that takes an input audio and a natural-language prompt and performs targeted ADD or REMOVE edits while preserving unrelated structure.
The demo audio files are based on music from MoisesDB and Epidemic Sound.
All demos are converted from WAV to MP3 for web demo loading purposes.