Generate audio with Stable Audio Open Small using ExecuTorch
Introduction
Set up your development environment
Download the Stable Audio Open Small model
Convert the model to ExecuTorch format
Build and run on macOS
Build and run on Android
Next Steps

  Log an issue

  Fork and edit

  Discuss on Discord

About this Learning Path

Skill level:  Introductory
Reading time:  45 min
Last updated:  21 Aug 2026
Skill level:
  Introductory
Reading time:
  45 min
Last updated:
  21 Aug 2026
Authors:Adnan AlSinan, Arm
Pareena Verma, Arm
Arm IP:Cortex-A Cortex-X
Tags: Performance and Architecture Linux Android macOS CPP Python Hugging Face ExecuTorch
Authors:
  • Adnan AlSinan, Arm
  • Pareena Verma, Arm
Arm IP:
Cortex-A Cortex-X
Tags:
Performance and Architecture Linux Android macOS CPP Python Hugging Face ExecuTorch

Who is this for?

This is an introductory topic for developers who want to deploy the Stable Audio Open Small text-to-audio model using ExecuTorch on an Android device or macOS.

What will you learn?

Upon completion of this Learning Path, you will be able to:

  • Download the Stable Audio Open Small model from Hugging Face
  • Convert the Stable Audio Open Small model to ExecuTorch (.pte) format
  • Build the audio generation application for Arm CPUs
  • Run the application on an Android smartphone or macOS and generate audio snippets

Prerequisites

Before starting, you will need the following:

  • A Linux-based x86 or macOS development machine with at least 8 GB of RAM and 50 GB of disk space (tested on Ubuntu 22.04 with x86_64 and macOS with Apple Silicon)
  • A Hugging Face account
  • An Android phone in developer mode with at least 8 GB of RAM and a cable to connect it to your development machine

Summary

AI-assisted

This summary was drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.

Close
?
You’ll convert the Stable Audio Open Small text-to-audio model to ExecuTorch .pte files and run the audio-generation application on macOS or Android Arm devices. First, you’ll set up Python and CMake, download the Hugging Face model files, and export its three submodules. Then, you’ll build and run the app on Apple silicon or cross-compile it for Android, transfer the required files, and retrieve the generated WAV audio.

Frequently asked questions

AI-assisted

These FAQs were drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.

Close
?
Which ExecuTorch installation option should I use?
Install executorch==1.0.0 with pip. Alternatively, clone ExecuTorch, check out v1.0.0, and run bash ./install_executorch.sh.
How do I know the model conversion to ExecuTorch worked?
After conversion, verify that conditioners_model.pte, dit_model.pte, and autoencoder_model.pte exist in the audiogen-et directory.
When building on macOS, what environment assumptions apply?
Create a fresh Python 3.10 virtual environment for the macOS build. Use macOS on Apple silicon (Arm64).
What should I check before running on an Android device?
Use an Arm64 Android device with at least 8 GB of RAM. The prerequisites specify FEAT_DotProd support and optional FEAT_I8MM support. Confirm that adb devices detects your connected device before you transfer the files.
What result should I expect when I run the application?
When you run the app, it generates a short audio sample from your text prompt. On macOS, the .wav file is saved in the current directory. On Android, exit the adb shell and pull the output file from /data/local/tmp/app.
Next