I regret nothing except reading the comments

After 10 million views, I've learned that nothing brings people together like arguing about a fictional princess.

Author

PJ Ace
January 09, 2026

Welcome to the newsletter. Buckle up, we’re getting straight into the prompts as promised:

2×2 Generation prompt.

Below is the prompt but here are a few important tips:

Have a script or shotlist and tell Chatgpt/Gemini to fill in details about your Scene, Characters, action, camera & composition, lighting and tone.

This prompt is more of a broad-strokes thing that you need to have ChatGPT customize based on your story.

The reason I’m all in on 2×2 is because it upscales with 95% hit rate as opposed to 3×3 which needs a lot of work and reprompting to get right.

Once you’ve determined the flow of your story through a script or shot list, have an LLM break it up into 4 shot sequences and specify those 4 shots in this prompt. 

You don’t need to individually write it, but tell your LLM that you’re looking for four shots in this sequence, for instance, “Hey chatgpt, use this prompt structure, please fill out the details but make sure to specify the four shots in the 2×2 grid are 4 big moments in Link jumping up, stabbing and landing next to the dead Lynel.”

You can get super specific or let the LLM do all the work if you just want to stay big picture.

The reason I prompt 2×2 instead of an image at a time is that it keeps spatial consistency between the images. So if you’re generating a room or two characters talking, they will be in correct relation to one another.

Remember to add @char1 or @img1 to reference characters or locations.

Here’s what it looks like in Freepik:

To 5,000 neckbeards that commented: YES, I KNOW THAT LYNELS ARE CENTAURS, NOT MINOTAURS. I’M SORRY, THERE’S NOT A LOT OF TRAINING DATA ON 18-FOOT TALL MORNING GLORY MILKING FARM PROTAGONISTS

Generate a photorealistic cinematic 2x2 grid of still frames from a live-action Legend of Zelda: Breath of the Wild film, depicting Link killing a Lynel in a single continuous battle moment.

The imagery must feel grounded, brutal, and physically real. No animation style, no painterly rendering, no exaggerated fantasy glow. Real weight, real violence, real consequences.

Remember to use @char1 as a reference for what Link looks like and @img1 for what the Lynel looks like.

Each frame represents a different angle or distance from the same sequence of action, capturing fragments of the moment where Link jumps onto the Lynel, drives the blade down, and the creature collapses.

⸻

SCENE & ENVIRONMENT

(add details here)

⸻

CHARACTERS

(add details here)

⸻

ACTION & CONTINUITY

(add details here)

⸻

CAMERA & COMPOSITION

(add details here)

⸻

LIGHTING & ATMOSPHERE

(add details here)

⸻

TONE & FINISH

(add details here)

Upscaling Prompt

After you generate, select your favorite images, crop them, and upload them back into a platform like Freepik to upscale.

Select Nano Banana Pro and put in the following prompt.
(I like to run 2 at a time to compare options)

Me after another hand-drawn furry art connoisseur explains why AI is tasteless

Preserve the exact composition, framing, camera angle, color grade and subject placement — do not alter or add new elements

Increase resolution to true high-end cinematic clarity, with natural film-grade sharpness (no AI oversharpening).

Apply blockbuster–style cinematography:

– grounded realism

– soft, motivated lighting

– natural contrast

– restrained highlights

– deep but clean shadows

– subtle atmospheric depth

-Keep the same color temperature and color tone

Texture pass should feel physically real: skin pores, fabric weave, dust, stone, metal, wood — all enhanced without plastic smoothing.

Maintain cinematic depth of field consistent with the original image (natural lens falloff, no artificial blur).

Do not redraw. Do not stylize. Do not beautify. Only enhance realism and resolution.

That’s the prompting process in a nutshell. Remember to use style and character references and then just take each scene in your script or shotlist 4 shots at a time.

A strong style reference really goes a long way to a cohesive aesthetic.


As I’ve shared before, use something like Figma or Milanote to organize your shots.

Technically, this scene should be numbered in a shotlist, but for more “vibey” projects, I like to just generate a bunch of shots and arrange them later like a puzzle.

That’s the gist of my latest prompt technique.

(If you haven’t seen the Zelda X post, it has a few more details in the breakdown. Also, I made an 80-second overview video on IG.)

———


I spend most of my time making ads at Genre.ai but I definitely will do another video game adapation or two.

Let me know which one you think I should adapt!

-PJ

Fair Use Legal disclaimer: This is a fan film, not affiliated with Nintendo or Sony Pictures