So last time I wrote about creating video and music with AI. Well, I’ve gone whole hog and created an entire album with AI. It is set for release on BandCamp on November 21. Giving it a little headroom allows me some time to promote and hopefully generate a small of amount of interest before it goes live. I did set up 3 tracks that are previewable. https://justinvail.bandcamp.com is the link to check out the album.

The process I somewhat covered in my last post but since I’ve gone this far, I’ll go over in detail how I got from total newbie to releasing an album of 11 songs.

My main tools are ChatGPT, Suno, and Audimee. With ChatGPT i worked out lyrics first. I’d write something, give it ChatGPT to refine, add my edits and so forth. Sometimes little changes after a few rounds, sometimes quite a few changes. In the end I wound up with lyrics I was happy with and can honestly say are my creation.

Then it was over to Suno for the music. After a few trys I wound up using ChatGPT to write prompts that were more specific than what I would manually write. I’ve since discovered this is a fairly common practice for music and video. ChatGPT provided prompts are not perfect, but in general do better than what I can write. It is still necesary to read over the prompts to make sure nothing extraneous was added. ChatGPt was designed to be creative. It will sometimes, based on previous prompts or what it “thinks” I want, give extras that may not be desired. In every case, because my singer is a baritone, I specified a baritone lead vocal from Suno. Doing that made sure I didn’t wind up with a song that was out of my singer’s vocal range.

Having achieved what I think will be a desireable prompt, lyrics are added, then the prompt, a title and then let Suno do its thing. After a few minutes two variations are presented. Assuming I’m happy with one of them, I then download the stems, those indiviual parts of the song. Stems are then loaded into Logic Pro and lead vocal isolated in a mono track file and saved. With the isolated vocal I then headed over to Audimee AI. I really liked the vocal on Fire and Rainbows which was the first AI song I was able to produce that “spoke” to me. I had Audimee clone and train a vocalist to sound the same. With that accomplished I had a consistent AI voice I could use on all my tracks. With the isolated vocal of whatever song from Suno I was working on, I could then run that through Audimee and have my trained voice sing the song. Once that was complete, usually within a couple minutes, I downloaded the new vocal and added it to my Logic Pro project. Once I’m happy with what I’ve got in Logic Pro, I saved the tracks as m4a ( iTunes ), I work on a Mac so the m4a is preferred. The wav file is for uploading to Bandcamp.

When I had originally started thinking about producing an entire album, I began researching what it would take to get it up on Spotify. The short answer is that the path to Spotify is through a distribution service. I’m not going to go deeply into all that because I ultimately decided not to try and get on Spotify. At least not yet.

There is a lot of information out there that suggests that Spotify is fine with AI generated music. And that at least on the surface appears to be true. However I’ve seen quite a few complaints about AI generated or AI assisted music being removed. I can’t cite a reference, because most of what I was seeing was from YouTube and I didn’t carry the research to the point of getting responses from Spotify or any distribution service. So this potentially falls into the category of hearsay.

I just wanted to get the music out there and Bandcamp seems for now to be my best bet. Joining was pretty simply and uploading though a bit tedious was not difficult. Naming, tagging and adding extra bits to each song like lyrics took a few minutes. Even before the upload, I created an album booklet that included lyrics and notes about each track. I also included some images I had created for the two music videos. Yeah, I have an AI singer name Justin Vail™. Yes I applied for a trademark.

Justin Vail™ is the visual front of this whole endeavor. You’ll see him in the music videos, it is his “voice” that sings. And yeah, he is entirely AI. Assuming I create a follow-up album or continue to make music videos it seemed like a good idea to not only create a public personna for him, but to protect that personna that I’ve spent quite a bit of effort in creating and developing.

So, that is where I am now. I have a album that is releasing in a week or so, for a AI character I’ve created and 11 songs that I think are pretty good. Whether anyone else does and is willing to shill out some cash to buy remains to be seen. Whatever happens, my creative side was fully engaged through this whole process.