The process behind MsUse (Official Music Video) and proposing new tenets for a new medium

The process behind MsUse (Official Music Video) and proposing new tenets for a new medium

By Sania Bhatia

Some notes by Lisa Lei

MsUse (Official Music Video) is not yet released, this is a short teaser: https://youtu.be/x73SFjRKEnE

It’s not uncommon for artists to be fearful of AI’s involvement in art and with our institutions right now. Before I chose to accept and demystify AI myself, I was also afraid of its impact in filmmaking. I’ve spent my entire education practicing traditional cinematography, so the idea of a machine recreating an image I’ve put effort into composing, sounds like the end of cinema. But there’s a kaleidoscopic cloud of mystification and misinformation obscuring the world of digital creation. The creative potential in these models can shine in a new medium, but it’s asking us to experience and play with it.

A few weeks ago, "Keep AI Out of Tisch" (https://nyunews.com/arts/filmtv/2026/04/29/keep-ai-out-of-tisch/) was published by Washington Square News (Alice Rogers), breaking down what the university, NYU, is currently missing (comprehensive ethics education, deep understanding of copyright law, and the importance of checking partnerships). I’m still a novice in both film and technology, but as an enthusiast of both and graduate of this institution, I strongly agree that these needs be met for future courses. These are necessary checks and balances for a university to stay on track with changing technology. The Hypercinema Lab is mentioned and its initiative to bring in Runway, a controversial AI generation tool involved in copyright controversy (we used Runway for this video). Not mentioned is the new class offered by Interactive Media Arts last semester, Generative Filmmaking in the Age of Hypercinema. Our insights about this class might provide an additional and necessary perspective around what these courses can be in the future, and why filmmakers would actually benefit from being involved in this blend. I've heard many critiques about this initiative on social media, and it's no surprise that filmmakers aren’t happy. I simply want to share our optimism in a time of fear through an alternative use, an already imaginative, playful, rebellious medium: a music video.

The class had two film students, including me, but most were students in IMA or ITP, so I was excited to see how we’d work together with such different backgrounds. Our professor, Gabe, took some time to give us a solid understanding of the impacts, ethics, and various examples of generative artworks before consumer AI. Then, we were introduced to the models: runway, midjourney, nanobanana, flora, gemini etc,. We were given some tokens for a few of the models like Flora and Runway, and spent a few weeks learning how to prompt visuals. I met my co-director Lisa Lei, who truly made this video come alive and was involved in every process since conceptualization.

To test what we could do with the AI, we were asked to use it in a music video. The music video medium is where big, unreal, imaginative ideas are truly embraced and welcomed, so it felt fitting for this test. Because of this freedom, we were able to experiment with the models, and let it guide the flow of the video. This project was going to be less of an experimental music video, and more of an experiment, needing us to keep an open mind to the outcome. We would later discover that blending machine learning with filmmaking would require acceptance of a new intersection between the two, and an openness to a whole different medium.

The Process

Lisa and I spent a while deciding how to use AI in the video. Which specific models do we combine, how much do we let AI roll the dice when it reaches VFX, and most importantly, how much extra work/less work is needed from us? Our whole process until visual effects was kept as traditional and straightforward for a music video as possible. We wrote out scenes, found locations, created a shotlist, and the three of us set out to film around my neighborhood and sneak into the MET. The last thing we needed were original plates of the talent singing from different angles, so AI could basically deepfake them together.

The edit needed to be quickly cut, with the film due 2 weeks from filming. Lisa took over on the VFX side of things and learned the language needed for the different tools to communicate with each other, while I did my best to hide AI’s mistakes with editing. She would input some plates we shot, for example, one shot of the artist's face singing, one of a pool ball, and prompt it (through various models) to make a singing pool ball.

Lisa: “Around that time, Runway Act-2 had just been released. After testing its lip-sync capabilities, we built a workflow that combined text prompts (generated with ChatGPT) into specific models (like Google Gemini), and start & end images in Midjourney to produce large-scale and morphing scenes”

We discovered, and I suppose we've always known, that these tools are truly autonomous. We’d get back rendered shots of visuals we never originally shot or even prompted. Unprompted, the model would make the Quaker Oats guy roll his eyes, or make the pool ball stick out its tongue. These mysterious outcomes really surprised us, so we used some of them in the video. In reality, these renders aren’t mysterious at all. The randomness creates the mystery, coming from a catalog of everything the model has seen before. The catalog is the model’s brain, and it works fast, giving us not only one option, but various different renders from one prompt. I think that’s one tenet of generative art for me, to be open to generating something new, completely by chance, and trusting it to lead the work down different paths.

Lisa's major, Interactive Media Arts, is known for conceptualizing and involving new technology in art. Bringing our skills together felt rebellious and new, and helped me redefine the relationship between humanity and technology. Lisa: “I’ve come to see how environment and exposure shape one’s response to new tools. Clearly our professor’s intent in pushing us to engage with them and form our own opinions. In my Interactive Media Arts program, experimentation with AI is widely accepted and even encouraged, though opinions still vary. With my background in visual design, I’ve typically seen its usage for early-stage ideation and prototyping. For this music video, that role shifted, where we developed the storyboard first and used AI to enhance footage we shot ourselves. As the project progressed, my sense of what was executable expanded. I was enthralled with the possibilities AI could create. I think this also comes from my limited background in film and TV. In that sense, the project became a gateway into that world, where Sania is more experienced. Seeing how much effort goes into even the shortest scenes, and how grueling the editing process was, particularly when integrating AI and hiding its mistakes - made this one of the highlights of my undergraduate experience. MsUse may appear highly machine-generated, but it involved as much, if not more, human conceptualization and effort than any other project I’ve made here at NYU so far”.

Making the MsUse video was a heavily involved process for us, creatively. This process of playing with it, subverting, and converging it with film felt like an experiment, as all of our opinions morphed and shifted. Ultimately, the AI in our video looks machine-made. I truly believe a trained VFX professional can do this with higher precision and control, because humanity is quite evident in art. On the first day of the class, we were asked to handmake a collage and prompt AI to create an image based on what we had made: https://www.instagram.com/p/DOPOtj_juCn/?igsh=MWlpeDhvbDh1dGs1cw== . We familiarized ourselves with the truth and practised it right away. My classmates made some beautiful collages with their own hands. Some AI recreated it well, but the artworks that stood out were made by our hands, magazines, glue, and scissors. You can limitlessly prompt the models for the perfect results with AI, of course. With a machine, those can be generated infinitely, but without a machine, we have a limit or a cutoff to resources, marking a project's end. AI can be a good resource in art, because it has the capacity for a higher limit. This collage project stayed with me while creating MsUse, sort of as a moral rulebook, a reminder that there are no limits in generative media until you create them.

In February, Film and TV professor Harry Winer hosted a small discussion at Tisch to elaborate on our experiences, and our classmates had some wonderful insights about demystifying AI. I wish I had the resources and platform to share their perspectives as well, but their projects were extraordinary, and their thoughts even more. The discussion we had reinstated for me, that certain understandings of ethics around technology are immovable and within us. We clearly need better boundaries with tools going forward, like Rogers brings up, but from this discussion alone, I am more certain that artists know and feel how to do the right thing to protect human creativity. Capital requires efficiency and artists need fantasy, imagination, and freedom to create. AI’s large umbrella covers both efficiency and fantasy, making it more difficult to vet the machines we create with. Dissonance between creativity and efficiency has always existed in the university and beyond (Rogers), and the grit of finding the middle ground is often left for us to figure out. I personally found I needed to reconnect with creation again to find my middle. I picked up crocheting (learned that it's one of the few crafts that can't be replicated by a machine!). It was hard, I realized I suck at counting. I started sewing (not well), making weird things out of trash, and attempted to learn the basics of creative coding. I embraced my greatest joy, cinema, and rewatched the movies that made me a filmmaker. Instinctively, we know we don't need to destroy to create, making it more important to embrace the technology we already have, to remember and to reimagine new mediums.

I don’t use AI in my filmmaking, other than in this video. The copyright, environmental, political, and equality concerns are still undeniable. I would also probably go broke buying tokens to create a whole video with AI on my own. But ethically, through attempting to make something with a tool I was previously afraid of, I’ve eradicated some fears I had around the extinction of artistic humanity. I enjoyed this project, I was able to think bigger than usual. We could make any face sing the song, from candy bars to nickels. The wrong people might misuse this ability but the least we can do is have fun. AI is the largest mirror we’ve held up to us technologically, yet it's a black box. It learns from us, but we also learn from other artists. Does that also make it a creator? Or a co-creator? Does it need us? There’s a whole future for us to experience with its involvement in art, and the story is incomplete. But because of this experience and discussions with open minds I've met along the way, I am okay with not knowing the answers to these questions just yet. I am open to seeing this technology sustainably and safely exist as a resource in the new medium. Asking artists, filmmakers, and institutions like NYU to consider the potential of these resources, because underneath this large umbrella, there’s plenty of room to build a safe playground with some guardrails.

Out of about 150 shots, 20 shots are rendered with AI in “MsUse”.

Lisa’s instagram: https://www.instagram.com/lisallei/

Link to Alice Rogers’ piece: https://nyunews.com/arts/filmtv/2026/04/29/keep-ai-out-of-tisch/2

Artist/Talent- BunnyMonroe (Gus Ralston)

Directed by Sania Bhatia and Lisa Lei

Edited and shot by Sania Bhatia

VFX rendering by Lisa Lei

Title graphics by Addison Rieke