From Prompts to Premiere: Our First AI-Assisted Short Film

In 2024, we decided to try something that, at the time, still felt experimental: producing a short film using artificial intelligence. The idea came from a simple motivation. A film festival deadline was approaching, and instead of approaching production in the traditional way—with cameras, locations, and a full crew—we wondered if it was possible to build an entire visual narrative using the new generation of AI video tools that were beginning to appear online.

So we tried.

At the center of the experiment was a Discord-based application that allowed users to generate short video scenes through text prompts. The system worked in a way that felt almost like directing through written language. Instead of setting up lights or moving a camera, you described what you wanted to see. The machine would then attempt to transform that description into moving images.

It sounded simple. In reality, it turned into a fascinating creative process.

Writing the Story First

Like any film project, everything started with the script.

Before touching the AI tool, we spent time writing the story the traditional way. Characters, scenes, atmosphere, and pacing all had to be imagined clearly on paper. In fact, this step became even more important than usual, because the clearer the narrative structure was, the easier it would be to translate it into prompts later. The script had to do more than describe actions. It needed to communicate mood, visual composition, and even camera movement. Every detail would eventually influence the prompts used to generate the footage.

Once the script was finished, the real experiment began.

Translating Cinema Into Prompts

The next step was something completely new: converting the screenplay into prompt descriptions.

Each scene had to be rewritten in a way the AI could interpret. Instead of dialogue and stage directions, we created visual instructions. A simple line from the script might turn into a long prompt describing the environment, lighting conditions, time of day, camera perspective, character appearance, emotional tone andvisual style

It was less like writing dialogue and more like painting with words. And it took a lot of iteration. Sometimes the results were surprisingly good. Other times the AI produced strange interpretations that forced us to rethink the prompt entirely.

Scene Generation and Selection

Once the prompts were ready, we began generating the scenes.

This part of the process felt a bit like working with an unpredictable collaborator. You would submit a prompt and wait for the system to render the clip. Sometimes the result captured exactly what you imagined. Other times it created something unexpected—but occasionally those surprises ended up being better than the original idea.

Because of this, generating the footage was not a one-shot process. We often produced several versions of the same scene and selected the one that best matched the narrative.

Gradually, piece by piece, the film began to take shape.

Editing the Film

After generating enough scenes, we moved into a more familiar stage: editing.

All the clips were assembled into a timeline where we could begin shaping the rhythm of the story. Cuts were adjusted, transitions were tested, and the visual flow of the narrative started to emerge.

This stage felt surprisingly traditional. Even though the footage was generated through AI, the editorial process remained very human. Decisions about pacing, timing, and storytelling still depended entirely on creative judgment.

Sound Design and Final Touches

The final stage was sound editing.

We added music, ambient sounds, and subtle audio cues to give the film emotional depth. Sound design plays a powerful role in storytelling, and this was especially important in a project where the visuals had been generated digitally.

Audio helped unify the scenes and give the film a sense of continuity.

By the end of the process, we had something that felt like a real short film—built through an unusual collaboration between human creativity and machine generation.

The Festival Result

We sent the film to the festival.

We didn’t make the final selection.

That didn’t feel like a failure. The experience itself was worth it. We had explored an entirely new production method and learned how storytelling could evolve when new tools enter the creative process.

At the time, AI-generated video was still in its early stages. The results were interesting but sometimes inconsistent, and the workflow required patience.

Still, it felt like a glimpse of something bigger.

Two Years Later

Now, looking at the landscape only two years later, the difference is striking.

AI-generated video has improved dramatically. Today it’s increasingly common to see AI-assisted commercials, experimental films, and social media campaigns. What once felt like a technical curiosity is gradually becoming part of the normal creative toolkit.

Just as digital cameras once replaced film in many contexts, generative video tools are finding their place in the production pipeline.

What seemed experimental in 2024 now feels like the early stage of something that will likely become routine. And that’s what makes the experience memorable: being there at the beginning.

Watch the Film

If you’re curious about the result of that experiment, you can watch the short film here:

Watch it on YouTube (The Meltdown - Shortfilm - Final Cut 2024): https://youtu.be/zl2eSNkWb8k

It may not be perfect, but it captures a moment when storytelling and technology were learning how to work together in a completely new way. And sometimes, that’s exactly where the most interesting stories begin.