Here’s a western metal sound.  It’s the song I used for the webcam intro video on YouTube.  “Suck It” is genuinely great (IMHO).  Ironically, it came out of an attempt to find the sound for the first version of “R Kelly in P Major,” which was a Jpop track.  That song did suck, but this one doesn’t.  Metal won’t be the band’s main sound, but it may show up every now and then.  OM-MINUS uses up most my energy for metal though, and has been getting all of my creative time lately – which isn’t much, but the great thing about making music with A.I. is that it can be done on the side of most computer jobs, without getting in the way.

While I’m on the topic, let me show a screencap of one of my tracks in Audacity –

It’s possible to write a script that pumps out 300 mediocre tracks per day, and it’s possible to get ChatGPT to write lyrics within an established genre, then plug the lyrics into an AI program and get a generic, but decent, track from that genre.  Getting A.I. to make a type of music that doesn’t exist in the material it has trained on is a different matter entirely.  Images too – I wanted an image of a steam engine that derailed and rolled over on its side for the cover of the Way of the Waifu album “Little Engines that Couldn’t,” but I couldn’t get any of the A.I.s I tried to generate a steam engine on its side.  Only upright.  I don’t give up easily, but that one beat me.  Maybe I should have taken one into GIMP, put it on its side, and told the AI to fix the lighting and perspective on my edited image.  That’s similar to what my music making process is like.

The top greyed-out track is the output from Treblo (formerly Sonauto, they rebranded for reasons that make sense to them).

That output was not directly generated by a prompt in Treblo.  It is a few generations of descent away from a not-so-great track generated in Udio before Warner Music Group inserted themselves.  I’ve been trying, and am currently trying, to make a good version of that track in Treblo; and along the way Treblo is spitting out bits and pieces that can be expanded into other tracks (like “Suck It” was).  Growing those tracks might result in a section good enough to also be spun off into another track.  This has been more productive for me than prompting with words.  A.I. in general, and Treblo in particular, are very good at ignoring word prompts.  Treblo can be very good when fed music though, but it usually needs some help from a DAW (Audacity, Soundforge, etc).

The 2nd track is empty space I added to the front of the first track.

Come to think of it, the top track has already been edited in Audacity. I used the envelope tool to fade the end out better, then cut a section out of the intro because it was too long, then cut out the intro and fed it back through Treblo to fix the skip where I cut the unwanted section out.  I made a dozen or more “fix” generations, and took the three best back into Audacity.

The second track from the bottom is the full track with the original intro cut off.

The third track from the bottom is one of the “fix” intros, which was still missing a little something.  The bottom track is a copy of the third track from the bottom.  I shifted the bottom track over because there was a part that could repeat, and used the envelope tool again to silence the parts of the track I didn’t need.

This process can be repeated again and again and again.  The track currently following this one (but only as of this morning, so it could change), is a more extreme example.  It started as a vocal section in a piece of total garbage, that was cut out and regenerated a few times.  It was extended many times, things were cut out in Audacity and replaced with other things, then run back through Treblo, then back to Audacity.  Over and over again.  I can’t even publish my music on Treblo, because Treblo doesn’t recognize it as music created in Treblo.

In the end, A.I. is just another tool in the tool kit.  It can either be used amateurishly, which will show; or it can be mastered – and that will also show.  The right answer isn’t to ignore it and pretend it will go away.  It won’t.  The right answer is to use it to do more.  Don’t make a song, make an album.  Make an album, and the videos, without having to sell out to a studio.  More fully realize a unique personal vision that can’t be made any other way.  There aren’t many artists who get to work without studio interference, or who don’t have to compromise on the vision with the other members of the band.  Now it can be done.

In other news, it looks like Yaoi Hokkaido will be the first member of the team to make it to the real world –

YouTube video link. 

Hmm – the Origin F1 kind of looks like Stuka though.  Elf Xuan doesn’t look like Dreilide, but is an elf robot and is already performing on stage.  But they can’t walk, and Moya can.

I had a dream that one day Way of the Waifu would be a real manufactured Kpop type group, made of humans.  Why not?  Look at how many movies and tv shows started out as comic books.  But, instead of human females, at this rate they really will be repurposed sex robots.  Ironic, but fitting.