Arm on-device real-time audio generation with Stable Audio Open model and Ableton Live

AI Impact Summary

Arm-based on-device sound generation demonstrates end-to-end inference using the Stable Audio Open model, executed with PyTorch and TorchAudio on CPU without GPUs. The workflow writes generated .wav files directly into the Ableton Live project, enabling a seamless prompt-to-audio loop that preserves privacy and minimizes latency. While the approach highlights efficient CPU-based diffusion parameters and thread utilization, production deployments must assess model licensing, memory footprint, and cross-DAW compatibility for broad adoption. This could unlock new edge-first creator workflows and potential cost savings by eliminating cloud inference for many use cases.