I've seen a fair amount of amazing artistic work developed with AI. With the help of Midjourney, I have even created images for Techlicious that I'm pretty darn proud of. But this music video released by Daniel Eckler of Spacecadet really blew me away with its incredible graphics and creativity. And everything in it is 100% AI-generated.
The journey began with a serendipitous output from Midjourney, of C-3PO sitting on a throne. From there came the fundamentally absurd concept of a gangster rap video starring C-3PO and Childish Gambino (Donald Glover). Actual production of the video involved an intensive 28-step process for Eckler and his colleagues, using a wide array of AI tools and platforms. If you want the details, you can learn more below.
Even though the music needs some work (Ok, a lot of work – AI generated music is well behind AI's visual capabilities), I believe this video serves as a powerful example of how AI will revolutionize the creative industries. Though, not everyone is in agreement. There were plenty of negative reactions on X (formerly Twitter), including one commentator who described it as, "Deeply stupid, immensely tacky, a fantastic example of mediocre AI slop."
Wherever your opinion may lie, as AI continues to advance at a rapid pace, we can all expect to see many more (and far better) projects like this emerge. So give it a watch and tell us what you think in the comments.
The process
Lyrics were generated using ChatGPT, with hundreds of options produced and the best ones selected. Rough cuts of the song were created using AI music tools like Suno, which helped to quickly establish the overall feel and flow. For the visuals, Midjourney was used to generate over 1,000 images, with the most striking and consistent ones chosen. Figma was used to organize and refine these generated assets.
Consistency was a key challenge, addressed through painstaking prompt engineering to ensure a coherent visual style. The team also researched extensively to find ways to smooth over the rough edges inherent in AI-generated video. Characters' outfits were designed using Midjourney, and a rough storyboard was put together using AI tools like Katalist.
Read more: AMD Unleashes Radeon RX 7900 GRE Worldwide: An Affordable Powerhouse
With the building blocks in place, the real assembly began. Photoshop's generative fill tool was used to clean up and adjust images. Runway ML was heavily employed to animate the characters, often requiring multiple runs to achieve the desired effect. An initial cut of the video was put together using another AI editing tool from Fable.
For the vocal track, human rappers were brought in to provide a base performance, which was then transformed using AI voice models from Resemble AI and Eleven Labs to sound more like C-3PO. Face swaps were done for human characters using $NSFW.
The editing process was highly iterative, with constant refinements and redos using various AI tools. Lip syncing was a particular challenge, with different AI models like Synclab and D-ID used for different shots. Upscaling of footage was done using Topaz Labs' AI-powered tools.
Read more Artificial Intelligence Can't Replace Headshot Photographers Yet
Disclosure: In keeping with the theme of this article, portions of it were created using Claude.AI.
[Image credit: C-3PO/Donald Glover mashup via Midjourney]
Josh Kirschner is the co-founder of Techlicious and has been covering consumer tech for more than a decade. Josh started his first company while still in college, a consumer electronics retailer focused on students. His writing has been featured in Today.com, NBC News and Time.
From Stewart Pan on April 29, 2024 :: 4:03 pm
Pretty stupid attempt. I’m probably biased because I hate that music anyway.
Reply