AI music generation

The amount of control you have varies by tool, but most music generators let you influence the output in several ways.

Genre and style are the most straightforward. Describing a genre ("indie folk", "90s hip-hop", "orchestral film score", "lo-fi bedroom pop") gives the AI a strong starting point. You can combine genres too, and the results are often surprisingly coherent: "bossa nova with electronic elements" or "Celtic punk" will both produce recognisable interpretations.

Mood and energy shape the feel. Words like "upbeat", "melancholic", "aggressive", "dreamy", or "anthemic" push the output in different emotional directions. These interact with genre in interesting ways; "melancholic electronic" produces something very different from "melancholic country".

Lyrics can either be written by you or generated by the AI. Most tools default to AI-generated lyrics if you don't provide your own, and the quality varies. AI lyrics tend to be competent but generic: they rhyme, they fit the metre, they make surface-level sense, but they rarely have the specificity or emotional precision of good human songwriting. Writing your own lyrics and letting the AI handle the music is often the more interesting approach, because it puts your ideas at the centre of the creation.

Instrumentation and production style can sometimes be steered through your prompt. Specifying "acoustic guitar and strings" or "heavy synthesisers and 808 drums" will push the arrangement in different directions, though the AI ultimately makes its own choices about how to realise your description.

Current music generators are much better at producing a plausible complete track than they are at giving you fine-grained control over specific musical elements. You can't easily say "make the second verse quieter" or "add a guitar solo at the 1:30 mark" or "change the key from A minor to C major". The AI generates the whole thing as a single piece, and if you don't like something about it, your main option is to regenerate and hope for a better result.

This is changing. Suno has introduced a Studio feature on its paid tier that offers more granular editing, including stem separation (isolating vocals, drums, bass, and other instruments into separate tracks) and timeline-based editing. But on the free tiers you'll be using for this activity, the workflow is: describe what you want, generate, listen, and regenerate if needed. Think of it less like using a music production tool and more like directing a very fast, very capable session musician who interprets your brief and plays you something back.

Suno (suno.com) is the most popular AI music generator and the best starting point for this activity. It's the tool that, for many people, first showed that AI could produce complete, listenable songs.

You type a description of the song you want (or provide your own lyrics alongside a style description) and Suno generates two versions for you to choose between. Each is typically around two minutes long on the free tier, with full vocals, instrumentation, and production. The results are often better than you'd expect, particularly for pop, rock, folk, and electronic genres. It handles vocals especially well, producing singing that sounds natural and emotionally appropriate to the style.

The free tier gives you 50 credits per day, enough to generate roughly 10 songs. Credits reset daily and don't roll over, so use today's credits today. You'll need to create an account, but the sign-up process is straightforward. Free-tier songs are for personal, non-commercial use only.

The paid Pro tier ($10/month) adds commercial usage rights, access to the more advanced v5 model, and more credits. The Premier tier ($30/month) adds Suno Studio (a DAW-like editing environment) and more generation capacity. For this activity, the free tier is more than enough.

Udio (udio.com) is Suno's main competitor, founded by former Spotify AI researchers. It produces music that many listeners consider higher-fidelity than Suno's, particularly for complex arrangements and genres that demand precise audio quality. Where Suno tends to be the easier, more approachable tool, Udio appeals more to people with existing musical knowledge who want more control.

Udio's free tier is more limited than Suno's: you get 10 credits per day plus a monthly bank of 100 credits. That's enough to try the tool and generate a handful of tracks, but you'll hit the limits more quickly. At the time of writing, Udio is in a licensing transition period, and some features (including audio downloads) may be temporarily restricted on the free tier. Check the current situation when you sign up.

The paid Standard tier ($10/month) and Pro tier ($30/month) unlock higher-quality downloads, stem separation, and more generation credits. If you want to take AI-generated stems into a traditional audio editor, Udio's approach will appeal, but for this activity Suno's free tier gives you more room to experiment.

Riffusion (riffusion.com) takes a different technical approach: it generates music by creating visual spectrograms and converting them to audio. It's more limited than Suno or Udio but interesting as an alternative method, and it's free to try.

AIVA (aiva.ai) specialises in orchestral and cinematic music. If you need a film score or classical-style composition rather than a pop song, AIVA is worth exploring. It has a free tier, though the outputs tend to be instrumental rather than vocal.

Beatoven.ai (beatoven.ai) focuses on background music for video content. It produces royalty-free tracks designed to sit behind spoken content. It's less flashy than Suno or Udio but more targeted for practical professional use. It also distinguishes itself by being Fairly Trained, independently certified as having trained its models on licensed music rather than scraped content.