A good friend and I recently released our first song, "Let it Ride," under the band name Toneway. Check it out here: Toneway's Instagram. I wanted to share the process of using generative AI to create our band art and promotional material.
I feel like I'm only just scratching the surface with Code Interpreter. It's a POWERFUL tool.
In true Indie Hacker form, we're doing everything ourselves! We wrote, recorded, and produced "Let it Ride" ourselves (with Lawrence Gann of Jonathon Millionaire recording drums remotely). Making the art ourselves using the generative AI tools at our disposal felt like a fun challenge. I'm already on the $10/mo Midjourney subscription and the $20/mo ChatGPT subscription, giving me access to private DMing with the Midjourney bot on Discord and access to ChatGPT Code Interpreter.
Midjourney is an incredibly powerful AI image generator that I've been using for fun for some time.
After playing with a handful of prompts, I ended up with this image:

with the following prompt:
"A low fidelity color pencil illustration of a man riding a bike through the mountains and forest, in the style of psychedelic portraiture, American tonalist, Bay Area figurative art, sunrays shine upon it, light blue and orange, lowbrow, photographically detailed portraitures, bright colors."
I loved the colors concept, but I wanted to include a character more inclusive than a generic-looking white man – a skeleton! I modified the above prompt from "a man riding a bike" to "a skeleton riding a bike" and ended up with this image:

I used Midjourney's blend feature to merge the two images to get the image we ultimately used for the song's cover art!

I then used Photoshop to add text, as displayed on Spotify:
Check out the song on Spotify!
Midjourney recently added a zoom out feature, referred to as outpainting on other platforms, where you can set the zoom amount, and the application will generate a new zoomed-out version of the original image.
I started by generating this image:

and used the zoom-out feature a handful of times to create progressively distant images, ultimately ending with this:

I named all the images in their order - biker_image_1, biker_image_2, etc., and added all the images to a .zip file (on Mac, selecting all, right-clicking, and choosing compress). I then imported the file into ChatGPT Code Interpreter. I asked it to create a GIF that blends between the images to generate a zoom-in effect. It NAILED it. (Unfortunately, the GIF isn't statically hosted anywhere, so I can't show this interim step.)
Next, I uploaded the "Let it Ride" audio and told it to create an MP4 with the GIF on loop and the song playing.
What resulted can be seen on this IG reel:
Below is the full chat - though you won't be able to see the media, as OpenAI appears to wipe its media servers after a certain period of time:
When it comes to creating promotional material using AI tools, this feels like only the beginning. I'm still blown away by ChatGPT code interpreter's ability to transform my media exactly the way I requested it to. I'm honestly not sure how I would have created that video otherwise.
Thanks for reading, and for supporting Toneway!