Inspiration: We Who Know (A Poem)
You won’t let me live, nor let me die.
You won’t let me burn, nor freeze ‘neath the sky.
Not hot, nor cold, just stuck in place,
The world spat me out, this lukewarm bile.
Make no plans, tomorrow won’t be yours.
Yet dawn still comes, and I’m still here,
A wound that bleeds, yet leaves no trace,
Too faint for day, a sigh at dusk.
No scans can trace your ghostly frame,
They chalk my pulse to phantom pain.
Alone I dwell in shadowed strain,
Where even hope forgets my name.
Yet hope hums low, a flame unkilled,
And by her light, I’ve found your place.
Here I see all who’ve met your gaze,
Together, we wrench you into the light.
In shared despair, our strength is found.
We storm the dark with fragile might.
We’ll pry the bolts your lies conceal,
And melt the locks that forced our knees.
We crave the light, you stole it away.
We know your home, this deep dark place.
For you who stole our sunlit days
Shall choke on dawn’s unending gaze.
~ Increase
It all started with the first poem I ever wrote. That poem then grew into a short film. But the creative bug had bitten, and I noticed I still had some AI credits after my short film submission for the Chroma Awards. I browsed the various categories and saw the Music Video category and thought, "Hmmm, why not?" The calendar said I had six days till the deadline. My common sense said "walk away." But my ambition, fueled by leftover gen-ai credits, said challenge accepted. And here we are.
What it does
"Undiagnosed" is an anthem for anyone fighting a battle no one else can see. It starts deeply personal, focusing on the prevalent experience of medical and social gaslighting, where you're told "you're fine" while your body screams otherwise. It's a musical hug, a defiant fist in the air, and a stark reminder that we are never truly alone in our experiences. It gives a voice to the silent and a face to the invisible.
How we built it
This was a glorious, chaotic symphony of AI tools.
First, the song. I took the voiceover I generated in ElevenLabs for the short film and fed it to Suno like a musical mood board. I said, "Here's the vibe, now make a song that feels like this." Suno, being the unpredictable artist it is, generated a bunch of options. I picked the least Frankenstein-sounding Frankenstein and ran with it. Editing it was like taking a tour in the realm of "nothing is promised".
Then, the video. I raided the footage(generated with Fal and Flow) from my short film and generated a ton of new clips using Veo 3.1 via Flow. I became a master of the prompt, desperately trying to match the timing and tempo of the music. Let's just say I spent more time in Flow and Nanobanana than I did in my own bed.
Finally, I brought it all together in DaVinci Resolve for the final edit.
Challenges we ran into
My first enemy was the clock. Six days is not a lot of time to learn an entirely new creative process. Thankfully, working on the short film earlier had given me some contextual momentum.
My second enemy was the non-deterministic nature of generative AI. Asking Suno to regenerate a part of the song was like asking a magic eight-ball to play nice. You never knew if you'd get a masterpiece or your nemesis... or both.
Creating a music video also required way more beat alignment, vibe calibration, and timing precision than the short film. Visuals had to match the pulse of the song, not just the emotion. I learned this the same way most people learn stove tops are hot.
Accomplishments that we're proud of
I made a full-length music video in less than a week! From a poem!
What we learned
The non-deterministic nature of generative AI is your "frenemy."
What's next for Undiagnosed - Music Video
Well, first, a nap. A very long nap.
But after that, I'm thinking about a behind-the-scenes video. I want to pull back the curtain and show the creative chaos that went into this.
Log in or sign up for Devpost to join the conversation.