I Made an Album With AI Music Tools. It’s… Actually Decent

The Setup: One Weekend, Six Tools, Zero Musical Talent

Let me be upfront about my qualifications. I played trombone in middle school. I can identify maybe four chords on a guitar. I am, by any reasonable definition, not a musician.

Which is exactly why this experiment felt worth doing. AI music generators have been making extraordinary promises — full songs from text prompts, studio-quality vocals from a sentence, entire compositions in thirty seconds. I wanted to know whether someone with no real musical ability could produce something a normal person would voluntarily listen to. Not a novelty. Not a joke track you share in a group chat. An actual album.

The tools I used: Suno (v5, the current flagship), Udio, AIVA, Soundraw, Mubert, and Stable Audio. I signed up for paid plans on Suno and Udio, and used free tiers on everything else. Total cost: about $20. Total time: roughly 22 hours across a weekend plus one obsessive Monday evening.

The album concept was called Departure Lounge — ten songs loosely connected by the theme of waiting. Waiting at airports, waiting for someone to text back, waiting for your coffee to cool down. Pretentious? Absolutely. But a cohesive theme would test whether AI could maintain any kind of artistic vision across multiple tracks, or whether each generation would feel like it came from a different universe.

Track by Track: What Worked, What Broke, What Surprised Me

Track 1: “Gate 34” (Suno, indie folk). I prompted for a quiet indie folk song about watching planes take off at night. The guitar fingerpicking was delicate and specific — not a generic strum pattern, but something that sounded like an actual arrangement. The vocals were female, warm, with this slight breathiness that felt intentional. The lyrics I wrote were mediocre (“counting lights on the runway / each one a place I will never go”) but the melody made them land harder than they deserved. First try. Forty-five seconds.

Track 3: “Lukewarm” (Suno, shoegaze). This is where things got interesting. I asked for a shoegaze ballad about waiting for your coffee to cool down, which is an objectively ridiculous concept for shoegaze, and the result genuinely reminded me of Slowdive. The wall of reverb was there. The dreamy, half-buried vocals were there. The lyrics were terrible — “steam rising like the questions I never asked” — but the chord progression and the layered feedback? That was real. I played this one for a friend without telling her it was AI-generated, and she asked who the band was.

Track 4: “Boarding in Ten” (Udio, jazz-hop). My first attempt with Udio, and the workflow difference is immediately noticeable. Where Suno gives you a complete four-minute track, Udio generates thirty-two-second chunks. You build the song piece by piece, deciding whether to extend forward or backward. The jazz-hop result was beautiful — lazy trumpet loops over a boom-bap beat with authentic vinyl crackle — but it took almost two hours. Udio is for people who want to be involved in the process. Suno is for people who want to hear the result.

The Middle Stretch: Where the Cracks Got Interesting

Track 5: “Delayed” (AIVA, orchestral). AIVA does not do vocals. What it does is generate instrumental compositions with structural sophistication the others cannot match. My melancholy orchestral piece had genuine dynamic range: quiet piano building through strings into a restrained climax. The problem is that it sounds exactly like a film score. Technically impressive, but it feels like the moment in a movie where the protagonist stares out a rain-streaked window, not a song on an album.

Track 6: “The Vending Machine Knows” (Soundraw + Mubert, lo-fi). Here is where I tried the background-music generators. Soundraw lets you customize mood, genre, and energy with sliders, but every output feels like a slightly different shade of the same song. Mubert generates ambient textures in fifteen seconds, but the results are aggressively inoffensive — pleasant enough for a YouTube vlog, not distinctive enough to stand alone. I layered a Soundraw beat under a Mubert ambient wash and called it a lo-fi interlude. It works as a transition. It does not work as music.

Track 7: “Three Dots” (Suno, R&B). A song about watching someone type and then stop. This track ate Monday. Suno is phenomenal at R&B — the vocal runs, the smooth production — but it defaults to generic emotional sincerity. I wanted something colder. More Frank Ocean, less John Legend. After about forty generations, I found that adding anti-instructions (“no vocal runs, no gospel influence, restrained and distant delivery”) finally pushed it into the right territory. The lesson: AI is incredible at genre-typical output. Getting genre-atypical output within a genre is much harder, and that is exactly where interesting music lives.

Track 8: “Carousel” (Udio, dream-pop). Dream-pop needs a slow bloom from quiet to overwhelming, and Udio’s chunk-by-chunk approach served this well. I could make each section slightly bigger than the last, adding layers manually. The finished track has a gorgeous crescendo where shimmering guitars stack until they collapse into a sustained chord. Udio’s 48kHz stereo audio is noticeably richer than Suno’s, especially in the high frequencies where reverb tails live.

Track 9: “Final Call” (Stable Audio, post-rock). Stability AI’s entry into music generation occupies an odd middle ground. My post-rock attempt captured the quiet-loud-quiet structure I wanted, but the guitars sounded synthesized in a way Suno’s do not. For sound design and ambient textures, Stable Audio is excellent. For anything that needs to feel performed by humans, it is not there yet.

Track 10: “Departure” (Suno, ambient). I ended where I started. A four-minute ambient piece about silence echoing through an empty terminal. Layered pads, distant piano, what sounded like field recordings of a large empty space. The perfect closer — quiet, patient, slightly lonely. Five tries. When the last note faded, I had an album.

The Honest Comparison: Which Tool for Which Job

Tool	Best For	Pricing	Verdict
Suno (v5)	Complete songs with vocals, any genre	Free / Pro $10/mo / Premier $30/mo	Best all-rounder by a wide margin
Udio	Precise control, layered builds, hi-fi audio	Free / Standard $10/mo / Pro $30/mo	More work, more control, great for producers
AIVA	Orchestral, cinematic, instrumental	Free / Standard $15/mo / Pro $49/mo	Unmatched for classical and film scoring
Soundraw	Background music for content creators	Creator $16.99/mo / Artist $29.99/mo	Functional, not for standalone music
Mubert	Ambient textures, quick background audio	Free / Ambassador $14/mo / Pro $39/mo	Fast and pleasant, lacks distinctiveness
Stable Audio	Sound design, ambient, experimental	Free tier + credit-based paid plans	Clean output, more sketch than song

The gap between the top tier and the rest is enormous. Suno and Udio are making music. The others are making sounds. That is not a criticism — Soundraw and Mubert serve real needs for people who want background audio quickly — but if you are trying to create something you would put your name on, the conversation starts and ends with Suno and Udio in 2026.

Between those two, the choice depends on your personality. Suno is for people who have a vision and want to hear it realized immediately. The trade-off is that you sacrifice granular control — if the chorus is perfect but the verse needs work, you regenerate the whole thing. Udio is for people who want to build the song section by section. The results can be more intentional, but the process takes five to ten times longer.

One thing worth noting: the legal landscape shifted significantly in late 2025. Warner settled with Suno, and Universal made a deal with Udio. Both platforms now operate under licensing agreements with major labels, which means the “is this legal” question that hung over AI music for years has largely been resolved for these two platforms.

What I Learned About AI Music (and Music in General)

The album is actually decent. I keep hedging that statement, but I mean it. If you put Departure Lounge on a playlist next to human-made indie music, most tracks would not be obviously out of place. Three or four of them — “Gate 34,” “Lukewarm,” “Three Dots,” and “Carousel” — are songs I would genuinely listen to again by choice. That is a strange thing to say about something I made in a weekend with no musical training.

But “decent” is not the same as “great.” Every track on the album is, at its core, an averaging of influences. The shoegaze track sounds like Slowdive because the model learned what shoegaze sounds like by processing thousands of shoegaze songs. It cannot make the next evolution of shoegaze. It cannot surprise itself. The output sits comfortably within established genres but never pushes against their boundaries in the way genuinely memorable music does.

AI music generators are extraordinary at competence. They produce technically proficient, emotionally appropriate, genre-consistent music on demand. What they cannot do — at least not yet — is produce music that feels like it had to exist. The melodies are pleasing without being memorable. The production is clean without being distinctive.

Here is the counterargument I keep coming back to: most human-made music is not great either. The vast majority of songs released every year are competent and forgettable. AI music generators are already better than a lot of what appears on streaming platforms daily. The question is not whether AI can make music as good as Radiohead. The question is whether AI can make music as good as the background of a coffee shop, and the answer is an unambiguous yes.

What surprised me most was how much the process felt like actual creative work. I was making artistic decisions constantly — about mood, pacing, emotional arc, when to push a prompt further and when to accept what the AI gave me. The tools removed the barrier of musical skill, but they did not remove the need for taste. If anything, taste became the only thing that mattered. For working musicians, these tools are instruments, not replacements. A songwriter using Suno for demo arrangements will work faster. A content creator who needs background music should use Soundraw instead of paying for a license. A film composer sketching ideas in AIVA is not cheating — they are being practical.

The album exists. It is on my hard drive. I have listened to it four times, which is more than I can say for some albums I have paid actual money for. Is it art? I honestly do not know. But the weekend I spent making it was the most fun I have had with technology in a long time, and the fact that the result is listenable still feels like something close to magic.

Frequently Asked Questions