The making of Luccia, the AI music video that looks like stop-motion

Highlights
Lita Bosch set out to make something her friend would love.
Luis Pinágel, a bedroom musician from Costa Rica performing as Naghel, played her his song Luccia. Born from heartbreak and unrequited love, it had the energy of something trying to outrun itself: rock and roll built on Western guitars, cinematic and nostalgic all at once. Lita knew immediately she wanted to bring it to life.
"I need to love the song," Lita says. "If I don't love the song, it's tricky — because I'm going to listen to it at least 10 times a day."
What followed was over a month of working around her day job as a color grader to produce the distinctive AI music video. A two-headed horse. A cowboy. A terrarium that keeps expanding. A spider that never quite catches its prey. All of it rendered in a handcrafted, figurine-like aesthetic so tactile that it looks like it’s physical stop-motion.
Every frame came from AI models, but every decision came from Lita.
"I entered a black hole"
Living an hour outside Barcelona in the Catalonia hills, her background spans sound, cinema, and photography, all the way to color grading. But when she started experimenting with AI video tools, something clicked.
"I entered like a black hole, and I was for days and weeks not sleeping, not eating. AI kind of allowed me to take an idea that I had in my head, and instantly do it. That's what's very addictive."
That speed of translation is what she sees as the fundamental shift in modern video creation and filmmaking.
"It's good, suddenly, to be able to do it," she says. "There's no gap time between your vision and execution. You can have an idea and straight away go and execute it.”
"I want it to look fake"
Before generating, Lita established a strict artistic boundary: no fake humans.
"I'm not interested in creating things where AI recreates humans," she says. "I kind of love to make it obvious that it's fantasy, that it's done with the eye."
The plastic, figurine-like characters populating Luccia are a deliberate stance. For Lita, what excites her about AI is building worlds that would otherwise be impossible on an indie budget. "As a colorist, I work like 10 hours a day. If I want to get into animation, there's no time for me to get into animation right now."
Two heads, dark thoughts, and happy accidents
The main character in Luccia was never supposed to be a two-headed horse.
Lita was building character sheets in Nano Banana, mixing photos of Luis and his mustache with 80s plastic figurine references, when the model errored out, generating a double-headed horse.
"The thing I like about AI is those mistakes that happen suddenly… and that mistake is like something amazing. The AI gave me back a horse with two heads. And I love this character."
That mistake unlocked the rest of the narrative. While Luis wrote Luccia as a heartbreak track, Lita saw a deeper, more psychological story about how sometimes you get lost in your dark thoughts, and those thoughts follow you. To Lita, it was about how a person is always in contact with their fears, and those fears follow you.
Balancing chaos with structure: working with Pablo
Halfway through the project, Lita brought in director Pablo Maestres, whose commercial work spans Netflix, Volkswagen, and Spotify. While Pablo brought years of traditional film directing, he was completely new to AI. Sharing more about the collaboration on this completely AI-generated version of the real interview, you see Lita's style and their dynamic, comically amplified.
"I'm very colorful, even when I don't look for it," Pablo jokes. "Maybe it's because I'm color blind. So that's kind of funny."

The dynamic between the two creators provided the exact tension the project needed. "I'm very messy in my head — just lots of things happening," Lita admits. "And he is the opposite."
Pablo built a visual map, breaking the song into distinct acts to anchor Lita's wild visual bursts. He also brought essential editorial discipline:
"We didn't use any AI to write the story. The ideas were ours. We were working on every single shot—deciding how to tell the story, the aesthetics, everything."
— Pablo Maestres
The technical grind
Luccia took a month of trial and error. Lita’s benchmark was modest: 20 seconds of usable footage per day.
"Some things go super fast," Lita says. "The living room TV scene — that was three days. Sometimes you ask the AI to sit down, and it just does whatever. “Why are you moving the frame? DO NOT MOVE THE FRAME."
The creative process
- Prompt engineering hack: Rather than guessing raw descriptive words, Lita fed reference images to ChatGPT first. "I go to ChatGPT and say: can you describe this image for me? Then from that description I make the prompts. I prefer to always go from words."
- Image generation: Character sheets and environments were mostly generated using Nano Banana.
- Photoshop prep: Photoshop editing was used to clean geometry, remove glitches, and isolate assets before motion generation.
- Video generation: Animating still assets using the AI video generator model, Seedance.
- Color grading and finishing: Polishing footage inside DaVinci Resolve.
Lita admits that her work as a colorist means she pays close attention to these details. The video is heavily color graded, but that's a deliberate personal choice. Often, the final touches can also be applied within the model itself through prompts, style adjustments, and lighting control. The best outputs tend to come from fine-tuning you only develop after generating at scale.
AI myth-busting and the craft
Lita is vocal about pushing back against claims that AI is replacing genuine human creativity.
"People think you can make a video like this in one day. It's a lot of trial and error. We are two creative minds making decisions constantly. AI will never replace our thinking brain, the way we process creative thoughts."
And to anyone who is still on the fence about using AI creatively, she offers a direct response:
"We're not taking anybody's job. We just bring something to the world that otherwise would not exist. If AI didn't exist, we wouldn't do this video.”
Ultimately, Luccia stands out because of Lita, her vision, execution, and a refusal to let AI dictate the final result.



