The free audio editor now gets a suite of free AI tools from Intel, some competing with expensive paid subscription services. That covers useful stuff like noise suppression and transcriptions and music separation, plus more experimental generative features.
Included features (with Intel’s documentation link) – all running locally on your computer (so local data only):
- Noise suppression for noise removal
- Transcription using Whisper.cpp with translation and export functionality
- Music separation (though that feature is becoming more commonplace all around)
- Music generation (from text prompts) and Music Style Remix (musical style transfer), based on Stable Diffusion / Riffusion
This is all effectively a collection of sound-specific tools built on the OpenVINO, an open-source toolkit built for running machine learning on CPU and GPU on Intel desktop and cloud platforms. (The usual open source-from-a-big-ventor thing – it’s free and open, provided you use specifically our products. And it’s not as the name might make you think an open-source database of pinot noir.)
The generation/remix features are still a little questionable in their value, but while it’s less hyped, the other features have quietly made AI audio processing something a lot of us use daily.
The catch here is which OS you’re on – and the mention of Intel should be a clue. The pre-built binaries are Windows only. There are instructions for Ubuntu (I caught this news from OMG Ubuntu). The builds require a huge number of dependencies, though, and on macOS there are no instructions at all. Maybe some intrepid CDM readers want to try collaborating on getting the Mac build to work? (It’d be terrific to have this stuff work natively on Apple Silicon, though, which I suspect might mean going back to the original libraries and skipping Audacity.)
For my part, I’m just going to try running the Windows stuff on my PC. It runs locally, so the other question will be how effective this is on Intel platforms.
This brings a larger point, though. With so much machine learning stuff publicly available as research, it may be a better value to run free tools than the commercial ones to stay current, even if you have the budget. That’ll be task-dependent, of course, but I’d be slightly concerned about sinking a ton of money into a subscription or software investment. (Of course, provided you don’t have an annual contract, month-to-month payments could be a better deal since you can always jump ship if something better comes along.)
This was released only early last month, so I expect sorting builds on other platforms is doable. Feel free to sound off in comments.
Introducing OpenVINO AI effects for Audacity [Audacity Blog]
Oh, and if you’re bothered by the look of Audacity, here is your next stop – maybe more important than these plug-ins, even:
Audacity Themes [medienmarmela.de / German – dark theme pictured at top]
Audacity 3.4 introduced a lot of improvements, as well.
I have a big backlog of interesting AI stories, so stay tuned for AI week very soon.