Moises’ AI Studio DAW arrives with a built-in session musician — here’s my first impression

Moises has just upped the AI ante. The company behind the popular stem separation and chord detection app—65 million users at the time of writing, plus Apple’s 2024 iPad App of the Year trophy on the mantel—has today unveiled AI Studio, a browser-based DAW with a synthetic session musician on call.

Is this just another suffocatingly simple AI prompt machine? Not quite. Moises AI Studio is all about generating stems.

READ MORE: Loopcloud’s Flip Sample instantly transforms loops and one-shots into new variations – and it’s a free update for existing users

Alongside the many ethical and artistic critiques levelled at ‘full-song’ AI generators, there’s also a much more practical complaint: they’re just not that useful. Spitting out prefabricated tracks is fun for a hot minute, but trying to extract instrumental parts that you can actually edit and mix has previously required multi-step workarounds.

This is where AI Studio stands out. Not only am I able to conjure unique instrumental parts on a dime, but Moises claims that the musical data has been ethically sourced, licensed, and copyright-cleared, so I had no qualms when hitting that ‘generate’ button.

The first feature to impress me is its “context-aware” audio generation. Record or upload some audio—it can be a solo musical part or a mix with multiple instruments—and Studio AI will analyse the tempo, genre, harmony, and rhythm. It’ll then generate an accompanying instrumental part that matches the groove of what you’ve already created. It’s a workflow that feels less like Skynet ghostwriting your song and more like a bandmate who actually plays in time.

At launch, there are seven instruments to choose from—guitar, bass, drums, and a smattering of keys and strings—and each one can be generated with a preset genre to shape the results. There’s a satisfying number on offer, covering everything from punk to trap, dubstep to bossa nova.

Alternatively, you can generate stems based on Moises’ AI matching technology to produce parts more personalised to the audio you’ve uploaded – in practice, I find this method is the most successful.

It’s all pretty fast, too, with three minutes of audio taking about 30 seconds to generate. In two minutes, I’d added unique drums, bass, e-piano, and violin to my initial acoustic guitar track – not too shabby. However, over time, that 30-second lag does start to add up, and, as unique audio is being generated each time, it’s not possible to audition presets before you render them. Even after a short time, I find myself losing creative flow as I render stem after stem, looking for the right sound.

You can also generate stems based purely on text prompts, you can upload audio reference tracks to guide the generated output, and you can even combine both techniques to get something that retains the feel and playing style of an existing musical section but takes it in a new direction. There’s also a set of additional creative controls which allow you to define how closely the generated stems should match your uploaded audio or the selected genre preset.

It’s worth restating that all of this comes packed into a web browser. I find the barrier to entry practically non-existent, with an uncluttered UI, no additional hardware needed to get going, and an easy-to-follow workflow. It’s also worth noting that all the new features in AI Studio sit alongside the existing Moises AI tools, which include stem separation, a suite of 50+ AI vocal models, a prompt-based ‘lyric writer’, and automated mixing and mastering capabilities.

Of course, pro-level producers are unlikely to make release-ready tracks without a few additional creative steps—though I may well be proven wrong on that—but AI Studio has big potential for remixers, content creators, and anyone looking to quickly work up a demo track from an initial sketch.

My first impression is that it’s pretty powerful as a space for rapid ideation, allowing you to hear your musical ideas in a range of genres and styles with minimal friction.

The biggest drawback is audio quality. Compared to a basic sample pack, you’re still looking at a noticeable step down. Depending on your production style, this may not be an issue, but when pairing generated stems with clean acoustic recordings, the difference is obvious to me.

AI Studio is available to all existing Moises users, and everyone gets a free monthly allotment of credits that can be used to generate stems. The Free tier gets you 60 generative-AI credits, a £5 per month Premium subscription will net you 100 credits, the £25 per month Pro tier comes with a hefty 600 credits, with more available for purchase if you manage to burn through your monthly stash.

With AI Studio, Moises is staking out a position on what generative AI audio can be, showing that it is possible to build ethically trained models that still produce musically coherent results with reasonable audio fidelity. Most important of all, at least for the casual user, the company’s conception of stem generation feels at home in the neighbourhood of sample libraries and software presets. These are tools designed to enhance productivity without supplanting creativity.

You can try it yourself at moises.ai.

Clovis McEvoy is a writer, researcher, and composer. His work has appeared in MusicTech, MusicRadar, Future Music, and the UN’s WIPO Magazine. He is currently an Affiliate Researcher at the University of Greenwich, investigating immersive music and media. He is an award-winning sound artist and founding member of the multidisciplinary group Rent Collective.