My review: Generating a song with AI and releasing it on Spotify

My main gig is writing, but I had a few lyrics I’d scribbled down ages ago for various reasons. They just gathered dust, never becoming actual songs. 😅 Then, I tried Suno AI recently and was blown away by its capability, so I adapted my lyrics into Japanese and created a track. It was genuinely moving to hear my lyrics actually turn into a song. The genre I mostly generated was sweet and dreamy, close to dreampop with female vocals. After putting in all that effort to make a few tracks, I wondered, "Is there a way to distribute this on music platforms?" I succeeded, and now my music is available on various platforms like Spotify, YouTube Music, and Apple Music. Below is proof that the track is live on Spotify.

AI Translated Image

In this post, I’ll share a simple workflow for how I made the song, plus how to distribute your music for free. If you've ever wanted to release a track under your own name, let me tell you—there has never been an easier time than now.

1. Production Workflow. The tools I used were Suno, Midjourney, Stable Diffusion, and Canva. For Suno AI, only songs created while on a paid subscription are licensed for commercial use. (Tracks made on the free plan are copyrighted to Suno.) So, you definitely need to pay for Suno, but the rest are optional. A $10/month subscription lets you create 500 songs. The workflow is simple.

AI Translated Image

➊ Prepare the song form. If you're not sure, you can just throw a song you like at an LLM and ask it to analyze the structure for you. It usually goes something like this: [Intro] -> [Verse 1] -> [Pre-Chorus] -> [Chorus (Hook)] -> [Verse 2] -> [Pre-Chorus] -> [Chorus] -> (Bridge, chorus variation or repetition, etc.) -> [Outro]. The song form is important because Suno uses it as a basis to adjust the melody and emotion. If you just give it stuff vaguely without structure, Suno will also just whip up something random.

➋ Write lyrics that fit the song form. For repeating parts like Verse 1 and Verse 2, you should match the syllable count as closely as possible so the AI doesn't get confused. It doesn't have to be perfect, but similar is best. Obviously, if you’re making an instrumental track, you don't need lyrics.

AI Translated Image

AI Translated Image

➌ Adapt the lyrics into your desired language. If you want to write in Korean, go ahead, but I personally like the J-Pop vibe, so I mostly generated in Japanese. Japanese naturally sounds smoother because it has fewer vowels and no final consonants (받침), and it makes rhyming easier. Just my personal preference (개취), but when you generate using Japanese lyrics, the AI references pre-trained Japanese tracks, which often results in the female vocals sounding relatively higher and clearer than when using English or Korean.

As shown above, I had Gemini and Claude translate the lyrics and then chose the better version between the two.

AI Translated Image

➍ Go to Suno and input the lyrics and prompts specifying the desired genre and sound. I consult the LLM a lot even when crafting prompts. I usually tell it the instruments and genre I want and ask it to write the prompt. I asked for things like electric piano, brush drums, jazzy chord progressions, and a dreamy dreampop genre. FYI, when you feed the lyrics to Suno AI, if you input them like this: [Verse 1] I hold the coral necklace and call your name. The longing in my heart never dries, wetting the transparent sky [Pre-Chorus] How many dreams must I cross to reach you? In the depths of scattered hearts, our promise sleeps [Chorus] The darkness of night surges like a wave, embracing the glittering coral memories, flowing towards you, slowly, slowly — the AI automatically generates the song, excluding the parts in brackets.