AI music gets a bad rep, and I understand why. There is a lot out there that sounds really bad. Garbage in, garbage out. The tools are what you make of it.
suno, write me a country song about losing my dog, and my house, and my team losing the big game.Yes, this will create for you a generic country song with all the tropes found in popular country music. But it will have no personality; it will have no heart. And after a single generation prompt, you will probably hear that all too familiar glitch sound in the vacals too.
I put a lot into my music long before touching the music generation tools, and a fair amount after. Those who make good use of these tools will tell you the same thing.
Tools I use:
- ChatGPT – research, brainstorming, making sense of brain dumps and thought explosions, turning chaos into structured lyrics (that are typically terrible) to begin forming what results in a final lyric. GPT also helps build style prompts for suno.ai by describing the instruments and dynamics (this is also an iterative process), the voices, the flow, etc. Also used for image generation.
- Suno.ai – the actual thing you hear, but in the raw. The lyrics and style prompt move to suno where I’ll generate a song over and over until I have something I like. This may take hours or days with lyric and style changes in the process to get something decent. I’ll then have it perform ‘covers’ of the one I settle on so I have alternate vocal tracks. I export the WAVs for the different versions.
- Galaxie.ai for access to the occasional Soro-generated video loop, or converting a still image into a video clip.
- BandLab – Basic Mastering.
- Audacity – Any raw sound editing such as clearing specific frequencies or artifacts before mixing.
- Logic Pro- Bring in the roughly balanced track, split the stems, then I cut out verses, or bridges I want to replace with revisions from the repermances. I’ll add additional effects, some personal vocals or sound effects, balance and eq each stem into a mix I’m happy with, and output a high bit-rate WAV back.
- Spectral Layers 12 – Used when I need a greater break down of the stems, this can split out each drum. This is also fantastic for raw audio editing, like removing unwanted sounds or noise.
- YouLean Loudness Meter 2 – normalize each track to -16 LUFS so Apple get’s precisely what it’s looking for.
- Canva – Great for graphic layout, lettering, etc.
- Photoshop – Need I say anything here?
- Topaz Gigapixel – AI upscaling tool. Images from GPT are low-mid resolution and Gigapixel can upscale them to proper usefulness.