Introducing AI Audio Effects to Audacity: Featuring Voice Transcription

Intel has rolled out a series of AI effects for the open-source sound editor Audacity.

Intel’s OpenVINO AI plugins for Audacity, as they’re officially called, are designed to function offline and locally, so there’s no need for cloud servers, registrations, or subscriptions to use them, you just need sufficiently good hardware!

Among the included are time-saving effects designed for editing spoken word audio, such as podcasts, audio books, and interviews, as well as effects for making/producing music (further enhancing Audacity’s growing DAW capabilities):

  • Noise Suppression, intended to reduce unwanted noise
  • Transcription, meant to transcribe or translate words and output to label track
  • Music Generation, used to create new music from a prompt
  • Music Style Remix, for generating new music based on existing music
  • Music Separation, to separate a song into its vocals and instrumentals

OpenVINO stands for ‘Open Visual Inference and Neural network Optimization’. It’s a new, open-source toolkit developed by Intel to run “AI inference” tasks “locally” on Intel hardware.

The Transcription effect is powered by Whisper.cpp, while the Music Generation and Music Style Remix use Stable Diffusion and Riffusion.

These new AI effects sound useful but there is a ‘bum note’ to using them on Linux: they have to be manually compiled from source code — both the OpenVINO AI effects and Audacity. No pre-built Linux packages are currently provided.

Helpfully, OpenVINO has a step-by-step guide that walks through the entire build process, start to finish. Better yet, it’s written against Ubuntu 22.04 LTS (though may work with later versions).

The Linux guide appears to be thorough, but I have not personally tested it. The compilation process necessitates a myriad of dependencies, some of which are several gigabytes in size and would exhaust my inconsistent 3G data. Furthermore, considering how sluggishly my laptop ran the GeekBench ML benchmark tool

I’ll leave the rest to you 😉

Windows users are taken care of with ready-made packages that contain all of the aforementioned OpenVINO AI effects. These can be downloaded from the OpenVINO GitHub releases page. However, these effects do insist on having the latest version of Audacity for Windows installed.

Go to the Audacity blog announcement for more detailed information about the OpenVINO AI effects.


Posted

in

, , , ,

by

Tags: