Lead With Love

I'm glad you liked it. :)

Of all the versions of music I got made for this, I think the "Epic" is my favotire.

The plan is to continue writing an album of emotional songs based on my life experience over time.

So, here's how it works:

Suno makes music with vocals and everything. That's all the AI, so it's an incredible tool for writers--one we didn't have at our disposal until these past couple years.

My process is as following:
First I write a song, then I have Dall-E (through ChatGPT) generate images based on my appearance, with an appropriate setting for the musical genre/style I want, Suno generate the music, and finally Headliner to create the videos for YouTube.

In this case I used the following prompt for Suno to get the exact feeling I was looking for:

"Epic, atmospheric, cinematic, deep resonance, soft percussion, heroic synths, majestic cords, ethereal pads, dynamic swells and crescendos, haunting undertones, rich harmonic overtones, adventurous swagger, clear and melodic bass vocals, hypnotic and mysterious. Arrival to Earth style."

Getting the images and music generated just the way I want it is sometimes incredibly frustrating and time-consuming, because there's a lot to learn about how to communicate what you want to an AI. With music as well, I really had to learn more about all the different musical genres, subgenres and instruments. You simply have to learn by experience. I did it the hard way myself, and learned it all through ChatGPT. Pro tip: The best lesson I've learned is to never listen to ChatGPTs incessant "help" when it's basically just making up busy-work to keep retain user interaction. I would ask it for pointers like: "How can I get this image to look more like me?" ChatGPT would full on walk me in circles, and if you allow it, it will change your prompt for Dall-E (the image generator behind the scenes) in all kinds of stupid and pointless ways. I would then ask ChatGPT to check it's own work and ask it to remove unnecessary fluff and redundancies, and finally it would admit that the original prompt I'd written was good enough... and that it was just inventing more busy-work.

Anaxios, for instance, took me 4 days just to get the image right, because of the excessive amount of specific elements I wanted: Make it look like me, in a specific pose and lighting angle, with the armor design and spires in the background based on my own Sci-Fi work, as well as the Axios character's specific look (eyecolor, hair and facial hair). After 3 days of that, it took another full day of making the Axios/Anaxios split image where half of it is the demonized fiction of him in the story: https://www.youtube.com/watch?v=V7iA_AwIk4c

Key thing here is to keep the amount of specific elements limited so it's easier for Dall-E to give you what you want. Usually I just focus on something easy, like: "Make it look like me, as a Scottish man in the highlands in the 1800s." I'll put most the emphasis on getting the face to resemble me. Sweet spot is defining it as "based 90% on my facial features" (from uploaded pictures), and then add the best description I can muster of my features in the prompt (shape of brow, eyes, nose, mouth, cheekbones, jawline and chin). Keep in mind that Dall-E isn't allowed to generate images based exactly on real people (if you ever wanted to go this route). In this specific case I settled for the picture you see in this video, which is quite a bit more defined/wrinkly than me, but it still resembles me, which was good enough for me: the: https://www.youtube.com/watch?v=6lQ9KuJsOag&list=PLiqpjAMwHGsZGtFdl9abtT...

Suno is much, much easier to work with, but it's easy to settle, because it generates a lot of good music. My advice to anyone who wants to try it out for the first time is to get a clear "picture" of what you want first. Think of something specific (song or instrumental) and lock in on the feeling and the instrumentals. If you don't know the specific words for what you want, that's something ChatGPT is really good at helping you with. Since I'm not a musician, I will often go to ChatGPT and ask it to define the instruments of a certain piece of music or genre(s) and have it generate a prompt I can use for Suno. I'll then break it down piece by piece, and have the AI explain to me what "driving percussion, bass guitar slides or acoustic guitar twang" (for example) actually means if I need to. I won't use words or language for Suno when I don't even understand what is, and I never trust that ChatGPT is 100% on the mark for what I want.

It ends up being a lot of experimentation and sitting around listening to endless versions of the same song, or just waiting for Dall-E to generate images. I often end up doing both at the same time, but when I do, I forget to listen carefully to the music. Suno will make mistakes sometimes, where a single word can be garbled, rushed and pernounced wrong. It's easy to miss after being in the fog of it for hours/days on end. I've had a couple mishaps where I only discovered the mistake after I already put the video out on YouTube.

And then there's the whole YouTube thing, for those interested. Creating your own channel feels much like creating your own website using a CMS like WordPress. Just go through everything, take your time, check all the options, learn what they are by experimentation. For new videos, always remember to define tags, and add an endscreen and cards that serve your channel. Also make playlists once you have enough content to warrant it and use card to refer to them. I use the endscreen to feature the next song I think people might be interesting in hearing. I originally chose the "Best for Viewer" which doesn't help your own channel much, so I don't recommend that--even though a lot of channels on YouTube recommend that option.

I hope this helps you, and anyone else that might wanna give it a shot. It's a whole new era for writers. :)

All I do is write and use AI tools to create the music and images. What you're hearing is actually all Suno (AI) made music, the instrumentals and vocals. I'm glad you liked it, though. :)