Thanks for liking my song!
My process to make it is this:
- I write the lyrics myself, no AI anymore, only my first four songs (out of a growing number, almost finished wiritng lyrics for the 24th) still contain some AI text.
- Since composing the music is a skill I do not (yet) have, I use Suno to generate a prototype of the music. This is somewhat akin to using the services of a ghost composer / music ghostwriter, but with (somewhat) more control, faster turnaround, and is considerably less costly!
- I split the stems (separate vocals from instruments) so I have one track that is vocals only, one that is instruments only.
- I remake the vocal track in Synthesizer V, which allows me, among others, to fix glitches from Suno that are difficult to fix within Suno, and also to get voice consistency accross all tracks from the project. I want them all sung by the same (virtual) singer, as if they were all from the same band.
- In my DAW, I do post-processing on the vocals track (compressor, drive, reverb, delay, multiply, EQ, etc) and mix with the instrumental track from Suno (which I use as-is).
This is the step at which the songs I posted are so far.
Eventually, my plan is to also replace the Suno instrumental track with a human-made one, so that the final version of the song is a human-made cover of the prototype generated with Suno.
Now, as for my process within Suno, I just don’t write a plain prompt, hit the button and take the first result. The prompt needs to be correctly crafted to get good results. There is an old saying in the computing field (I am a computer engineer) that is still valid today, especially with transformative AI: garbage in, garbage out. I know the feel/vibe/type of sound I want from the song (including band/album consistency accross multiple songs), and it usually takes a lot of tries (often dozens) before I get the one that best resonates with what I want from it.
I like very much symphonic metal (Nightwish for instance) so I like gettign either orchestral instruments, piano or violin mixed in my metal songs. Those that have violin, the idea is inspired from Lindsey Stirling whose music I like.
To answer what you’re wondering, Suno made the arrangement and I chose the one that gave me the feeling of “this is the one” from a lot of versions. This “this is the one” feeling comes when it fits what I had in mind, and I listen to it and just want to keep it on repeat over and over and over again, and then more.
So, while Suno made it, the energy and mood are what I was seeking from the start, the kind of sound and vibe I want out of the entire project (all the while having the songs sound different from each other while belonging together). For the whole project, several hundreds of unfitting versions sit at the bottom of the trash bin, witnesses of my attempts to get the right one for each song.
This song has something different from my previous ones though: Suno just released a new version 4.5 model about 2 weeks ago, this song is generated using v4.5 while the others were done using v4.
As for Latin vs English, yes understanding the original language can help, but personally, I like songs in other languages even when I don’t understand, and my brain treats the vocals as a highly sophisticated instrument, catching a few meaningful words here and there.
This one, along with Filia Fraudis (The Daughter of Lies) and the upcoming Fata Geminata (Twinned Destinies) are a series of medieval prophecies forming the backbone of my trilogy, so it makes sense to have the prophecies written in Latin - and have the songs in Latin as well.
Most of my songs are in english though! Got 3 in Latin, one in German, one in French (I’m french native) and the rest is in English.
French will be hard to get right in Synthesizer V (get rid of any foreign accent) but I’d like to try that challenge!