Stop Prompting, Start Directing: How to Speak Midjourney's Language

I still remember the first time I generated an image with Midjourney. The prompt was simple, something like “an astronaut riding a horse.” When the grid of four images materialised a minute later, it felt like genuine magic. A door to a new universe had just been kicked open. I, Rohan Mehta, a writer and editor with zero drawing talent, could suddenly create visual art. The power felt immense.

That initial euphoria lasted about a week. The magic quickly faded into a familiar, frustrating pattern. I wasn't trying to make surreal art anymore; I was trying to use these tools for my work at Pulse AI. I needed specific images for articles, social media, and presentations. But everything I created had this sterile, uncanny sheen. The people were too perfect, the lighting too flat, the compositions too centered. It was the visual equivalent of elevator music – technically proficient but utterly devoid of soul. It all screamed “made by AI.”

My breaking point came while working on a long-form piece about the future of work in India. The narrative was personal, focusing on the quiet ambition and hustle of a new generation navigating careers from their homes in bustling metros. I wanted an image that captured this mood: a sense of focus amidst the chaos, a bit of melancholy, and a lot of hope. I typed in what I thought was a descriptive prompt: “A young Indian woman working on her laptop in a modern Mumbai apartment, looking focused.”

Midjourney returned exactly what I asked for, and it was terrible. The woman looked like a stock photo model who had never known a moment of self-doubt. Her apartment was a minimalist fantasy from an interior design magazine, not a real, lived-in space in a city where every square foot is precious. The lighting was bright and even, erasing any sense of time or place. It was a cliché. The image didn't support my story; it undermined it. I had described a scene, but I had failed to convey an emotion.

That’s when I realised my entire approach was wrong. I was talking to Midjourney like it was a search engine, a glorified Google Image search that could invent pictures. But it isn't. An AI image generator is a creative partner. More specifically, it’s your own personal, infinitely patient, slightly literal-minded film crew. You aren't meant to just 'prompt' it. You're meant to 'direct' it. You have to learn its language, and that language isn't English; it's the language of cinematography and art direction.

I decided to scrap my old prompts and start over, this time imagining I was on a film set. My goal wasn’t just to describe the 'what' but to dictate the 'how'. This shift in mindset changed everything.

The first thing I tackled was lighting. My initial prompt had no lighting direction, so the AI defaulted to its standard: bright, commercial, and boring. This time, I thought about the mood. I wanted something intimate and a little dramatic. I added “chiaroscuro lighting” to my prompt, a term from art history referring to strong contrasts between light and dark. I also specified the source: “light from a single desk lamp illuminating her face.”

Suddenly, the scene transformed. The background fell into shadow, creating a sense of intimacy and focus. The subject was no longer just a person in a room; she was the center of her own small universe. The generic apartment became a mysterious, personal space. I was getting closer.

Next, I thought about the camera. Who is telling this story? Where is the viewer? My first image was shot from a dead-on, eye-level perspective, which feels neutral and uninspired. For my new version, I wanted to convey a sense of strength and determination. I changed the virtual camera position. I added “low-angle shot” to the prompt. This simple change made the subject feel more powerful, monumental even, as if we were looking up to her. It’s a classic cinematic technique used to make characters feel heroic, and it worked wonders for the AI.

Then I considered the lens. A standard AI image often feels like it was shot with a default phone camera lens – everything is in focus, and there’s no sense of depth. Photographers and filmmakers use different lenses to create different feelings. A wide-angle lens can make a space feel vast or distorted, while a telephoto lens compresses the background, creating a sense of intimacy or surveillance. I wanted a natural, film-like feel, so I added “shot with a 35mm lens.” This is a classic focal length for street photography and cinema, known for providing a field of view that feels human and grounded. The result felt less like a render and more like a captured moment.

This is where it gets really fun. Now that I was thinking like a director, I could get obsessed with the details. Most AI images are too clean, too digital. Real life, and especially real film, has texture. It has grain. It has imperfect color. So I started specifying the film stock. This is perhaps the most powerful piece of vocabulary you can learn.

Instead of just saying “photorealistic,” I told Midjourney to emulate a specific type of photographic film. I prompted for “shot on Kodak Portra 400.” This is a legendary film stock famous for its beautiful, warm skin tones and fine grain. The effect was instantaneous. The image developed a soft, analogue warmth. The digital sterility vanished, replaced by a subtle, pleasing graininess. The colors shifted beautifully. Now it didn't just look *like* a photo; it *felt* like one. For a different project, I might ask for “Fujifilm Velvia,” known for its high saturation, to capture the vibrant colours of a market in Jaipur, or the stark, moody look of “Ilford HP5” for a black and white shot.

Finally, I addressed the composition and the environment. The “modern apartment” was the source of my stock-photo problem. Real apartments, especially in a place like Mumbai, are full of life and clutter. So, I changed the prompt to “cluttered, lived-in apartment with books and papers on the desk.” I also added a classic compositional rule: “rule of thirds.” This instructed the AI to place the subject off-center, making the image more dynamic and visually interesting than a simple, centered portrait.

I even added details about the world outside the window. Instead of a generic cityscape, I specified “the blurry glow of neon signs and passing autorickshaws seen through the window.” This not only grounded the image in a specific cultural context but also provided a beautiful secondary light source, adding depth and color to the shadows.

My final prompt was a monstrosity of a sentence. It looked something like this: “Cinematic film still of a young South Asian woman, focused, working at a cluttered desk in her moody Mumbai apartment, low-angle shot, dramatic chiaroscuro lighting from a single desk lamp, the blurry neon glow of the street seen through the window, shot on Kodak Portra 400, 35mm lens, grainy, atmospheric, rule of thirds composition --ar 16:9.”

The image that came back was everything the first one was not. It had a soul. It told a story. The woman felt real, the space felt authentic, and the mood was perfect. It wasn’t a picture *of* a woman working; it was a picture *about* ambition, focus, and the quiet solitude of modern work. I had finally created the image I actually wanted.

The lesson here is transformative for any non-artist trying to wrangle these incredible tools. Stop describing a noun and start directing a scene. Your vocabulary is the single most important asset you have. Don’t just say “dark,” say “chiaroscuro lighting” or “shot during golden hour.” Don’t just say “from the front,” say “low-angle shot” or “Dutch angle.” Don’t just say “realistic,” say “shot on Kodak Portra” or “style of a still from a Satyajit Ray film.”

Learning this new language gives you control. It allows you to build a consistent aesthetic across multiple images, creating a coherent visual identity for a project instead of a random grab-bag of AI-generated styles. It’s the difference between being a passive requester and an active creator. Whether you're in Bangalore trying to illustrate a blog post or in Boston creating a marketing campaign, the principle is the same. The AI is your studio, and it's waiting for your direction.