By Jessica Spencer

I've been working on a new song called "Lighthouse of Dreams" over the past few months as part of an exploration into where AI can fit into the creative process. Excited to share that the song and video are now done, and I've learned so much along the way that I can't wait to share with you all.

This song was built out of a partnership with Podia, the online learning platform I've loved using for my courses. Podia was looking for creators to sponsor, and my idea of writing a song around Podia and creating a video leveraging AI during the process was selected, which was really cool!

Questions Explored

During the process, I wanted to explore these questions:

• How do I use AI in the creative process and also create something that feels authentic and ethical?

• How do I choose which areas in my creative workflow to augment through AI?

• How do we lead AI when creating and not let it take the reins?

• What are the best AI tools to support my creativity?

How do I use AI in the creative process and also create something that feels authentic and ethical?

I learned that when we can study and understand our current processes and gaps, and then weave AI into those processes, it can be incredibly beneficial. I was amazed at how quickly I was able to overcome creative blocks, and I found myself with more creative energy to take things even further.

In terms of authenticity, we have to think about our personal approaches to creativity. For instance, I'm a bit more analytical in my approach to creativity and see it like putting together a puzzle. This approach lended itself nicely using ChatGPT for idea generation around things like themes, metaphors, rhyming words with specific meanings (takes it a lot further than a simple thesaurus), and even storyboarding the video.

Training is also really important. To keep things feeling authentic, I ended up training a custom GPT on past lyrics I had written so it could get a feel for my tone, cadence, and other patterns that natural language processing technologies do a great job of picking up on.

How do I choose which areas in my creative workflow to augment through AI?

What areas hold you back from

How do we lead AI when creating and not let it take the reins?

Finding ways to be involved at every point in the process makes a huge difference. We can do this by giving ourselves the power of choice (ask for 10-20 ideas around something so you can choose), working in chunks (tackle a single scene from the video instead of creating the full thing all at once), and going through every step in our process (just with AI being a partner to support that process). We have to remember that AI is a tool to help us develop our vision. Thinking about our vision of what we're creating and the reason that we're creating it, early on in the process, is key.

What are the best tools to support my creativity?

This one varies based on your use cases. For me, these are some of the tools that I landed on for "Lighthouse of Dreams":

• ChatGPT - Ideation and support with lyrics

• Soundtrap - AI noise removal tool within the DAW

• LANDR - One-click AI mastering

• Midjourney - Images to build music video scenes from

• Kaiber.ai - Music video (scenes built individually from selected images I created through Midjourney)

I also played around with a bunch of the music generation tools, but found that these didn't work as well in my creative workflow when writing this song. I'm starting to work on some new ones, though, creating the beat with AI first and then building everything else around that. For this one, I liked being fully in charge of the melody, chord progression, vocals, and piano.

I also see a ton of value in music generation tools when doing improvisation to develop new songs. Often, listening to other songs can inspire new musical ideas...the cool thing with these is you can generate an AI music track that can inspire ideas with less risk of sounding like something that's already out there.

Here are a few of the standout features for some of the AI music generation tools I tested:

• Soundraw - Edit instruments and measures after AI generation. Most of the other tools don't give you this type of creative control post-generation.

• Splash Pro (AFFILIATE LINK) - This one is interesting as it lets you create songs using text prompts (most of the others just let you pick options).

• Tuney - Has an interesting AI remix feature that lets you generate a new beat behind any vocals (removes any existing instrumentation from the recording, and syncs the musical key, tempo, and style with the new beat). I sounded like a chipmunk on some of the generations due to key/tempo changes that the platform auto-selected, but you can keep regenerating the output...it just takes some time.

• ChordChord (AFFILIATE LINK) - This uses AI to suggest chord progressions and variations of each chord based on the key you're in. This is a neat one for learning music theory as well to see the patterns of how different chords fit together. Creations can even be exported in MIDI format for use in your DAW.

Where AI Contributed In This Song

During this song, I used AI in the following processes:

Lyrics

My husband and I started talking through some lyrical ideas prior to bringing AI into the loop. I'm a lot slower at lyrics than creating music, and my husband's an incredible creative writer, so it's been awesome partnering with him on songs in the past. Once the first draft was done, I wanted to make it more of a story and work in some more metaphorical language. Here I turned to ChatGPT. I started out by creating a custom GPT that was trained on some of my past lyrics so that it understood my style, tone, cadence, and other lyrical elements. Then I started asking it questions such as the following:

• I'm writing a song about Podia. I love that they are simple to use and learn, you can get your online education business up and running super quickly, the platform looks beautiful, your audience has a single place to come back to, it's easy to communicate with them, and it's made for online creators/entrepreneurs. Help me come up with 10 themes that could represent the entrepreneur navigating the online space and finally finding a home with Podia. - This is where the idea of the lighthouse and sailor's journey came from.

• Give me 5 storyboard options for the song based on the theme of the the sailor's journey to the lighthouse.

• I like this storyboard option [COPY AND PASTE FROM PREVIOUS RESPONSE]. Please give me five options for the first verse. - Based on the custom GPT I trained on my past lyrics, this is where I got to test out if it was aligned with my style. This also allowed me to select what resonated and helped me to remain involved in the creative process.

• I don't like this line: [INSERT LINE HERE]. This is why I don't like it and what I'm trying to achieve: [INSERT DETAILS HERE]. Give me 10 other options.

• What words have a short "u" sound and a similar meaning to hope?

You start building more of your puzzle pieces, then you get to put them together mixed with what's already been created. This made it easier to stay in control of the writing process.

Recording

I used Soundtrap's built-in AI noise reduction filter to remove background noise from my vocals. Although I played around with a ton of AI music generation tools, I ended up not including anything else AI-related when creating and recording the music. That was one of my biggest lessons in this process is that we get to pick and choose where AI best fits, and on this one, it ended up not being needed.

Mastering

LANDR offers one-click AI mastering for audio recordings. It turned out really good, and I could hear a very noticeable difference after running it through the platform. There were few controls for this...it makes nearly all of its own selections when mastering, so definitely limitations here compared to working with a professional audio engineer. This technology made mastering much more affordable and accessible, though. LANDR also highlights on its website that its platform is used by "over 5 million musicians globally," including artists from Def Jam Records, Warner Records, and Disney Music Group.

Music Video

For the music video, I used three AI tools: ChatGPT for storyboarding and character development, Midjourney for scene starter images, and Kaiber.ai for creating the actual video scenes.