One prompt, 7 lines: how to turn your selfie into a GTA 6 cover
No drawing skills and no heist. A selfie, a chatbot and a prompt you build one line at a time.
GTA 6 comes out in three weeks and a lot of profile photos are about to move to Vice City.
Most of them will come from one-click filters, which are fine for a quick avatar. For a poster with your own face on it, a couple shot as Jason and Lucia, or a made-up title across the top, you’ll need to write a prompt. Seven lines, to be exact.
Each of those lines is there because skipping it breaks something specific. Below we build the prompt from scratch for ChatGPT and Gemini, explain what every line fixes, and then put the whole thing together so you can copy it.
Line zero: the photo
It stays outside the prompt and still decides more than any line inside it.
Here’s what a GTA filter does with your selfie. The AI looks at your face, pose, clothes and background, and then redraws all of it in the new style. Roughly speaking, it’s a police sketch artist working from a description, and a blurry description gets you a blurry suspect.
So give it a front-facing shot with even light, no harsh shadows, no blur and a plain background.
Sketchto suggests a profile or three-quarter angle for a more dramatic loading screen. That works for drama and costs you likeness, so if people should recognize you, stick with the front.
Line 1: pick one of the two GTA looks
“GTA style” now covers two different things, and the AI won’t guess which one you mean.
The first is the classic hand-drawn poster. That’s the look of the series’ covers from GTA III onward and the GTA V loading screens: thick black outlines, bright flat colors, cel-shading and a comic-book texture.
The second is “real to GTA”, where you look like a character in a GTA 6 screenshot. Gemini handles this one best, since the whole trend grew out of it.
Write the look as a full sentence:
Turn the person in this photo into a hand-drawn GTA loading screen illustration with thick black outlines, bright flat colors and cel-shading.
Or, for the screenshot version:
Make the person in this photo look like a realistic in-game character from a GTA 6 screenshot.
Full sentences matter here. Gemini ignores tag piles like “gta, 4k, masterpiece, — ar 3:4” and wants the task explained the way you’d explain it to a friend. All that prompt-forum wisdom about tags, ignored. Tough crowd.
Line 2: tell it where you are
The look alone gives you a GTA character standing nowhere in particular. The setting line puts you inside a specific game.
For GTA 6, borrow from Rockstar’s official art: a tropical sunset, palm trees, Art Deco buildings and a magenta-and-turquoise palette. Those are the details the AI picks up when you ask for an authentic GTA 6 look.
Background: Vice City at sunset, palm trees, Art Deco buildings, magenta and turquoise colors.
For older eras, naming the game is the shortcut. Dzine and LightX both respond to phrases like “GTA San Andreas style” or “retro GTA Vice City colors”, and Dzine also suggests “GTA IV aesthetic”.
Dzine even lets you set the strength with words. “Subtle GTA effect” gives you a light touch, and “strong Vice City style” gives you the full treatment. Handy when the first attempt came out a little too cartoonish.
Line 3: protect your face
This is the line people skip, and it’s the reason so many GTA portraits show a stranger wearing your T-shirt.
Without it, the AI happily “improves” you toward a generic GTA character. Somake warns about this on its own site: the AI may simplify or exaggerate details to fit the style. Your nose and your age are both up for negotiation.
Keep their face, hairstyle and skin tone clearly recognizable. Don’t beautify them.
The model matters here as much as the wording. Gemini’s Nano Banana Pro holds facial details, skin tone and hairline more accurately than ChatGPT, which is why people pick it for portraits.
Get Fancies AI’s stories in your inbox
Join Medium for free to get updates from this writer.
If you’d like to compare for yourself, Sketchto lets you run the same GTA preset through several models: Nano Banana, Nano Banana Pro, GPT Image 2, FLUX 2, Seedream and Qwen Image Edit. Some will return you. Some will return a distant relative.
Line 4: lose the smile
AI defaults to friendly. GTA covers are anything but. The characters on those boxes look at you like you’re late on a payment to them, and a big grin ruins the whole mood.
Mouth closed, not smiling.
Four words, and the result gets noticeably closer to the canon.
Line 5: the title and the logo problem
A cover needs a title, and this is where ChatGPT earns its place. It writes letters inside images more accurately than the others, so “WANTED” signs, mission names and your own names come out readable.
There’s one catch. ChatGPT refuses anything with the GTA logo, the word “Rockstar”, official artwork or real car brands. It guards that logo better than Rockstar once guarded the first GTA 6 trailer from leaks.
So describe the style and invent the title:
In the style of a GTA 6 cover, with the title “GRAND THEFT AVOCADO” in bold cover-style lettering. No real logos.
You lose nothing by this, since the AI has no trouble with the palm trees and the sunset. And art without the real logo is also the safest kind to post.
If you need exact wording, like the date and address on a party invite, it’s often easier to add it afterwards. LightX lets you put game-style fonts, a health bar and a mini-map on top of a finished image.
Line 6: set the shape
ChatGPT often hands back a square unless you say otherwise, and a square fits neither a wall poster nor a YouTube thumbnail.
Vertical 3:4 format.
Use 3:4 for posters and Instagram and 16:9 for YouTube. Put it in the first prompt along with everything else, so the first attempt already has the right shape.
Line 7: the second message
The last line goes into a follow-up. Gemini works best as a conversation: generate, look at the result, then ask for changes in the same chat. “Make the sunset more orange.” “Darker sunglasses.” “Move the title higher.”
And generate more than once. The first result is rarely the best, so a couple of extra attempts are part of the process.
The full prompt
Here’s everything together, ready to paste into Gemini or ChatGPT with your photo attached:
Make the person in this photo look like a character on a GTA 6 cover. Background: Vice City at sunset, palm trees, Art Deco buildings, magenta and turquoise colors. Keep their face, hairstyle and skin tone clearly recognizable. Don’t beautify them. Mouth closed, not smiling. Add the title “GRAND THEFT AVOCADO” in bold cover-style lettering. No real logos. Vertical 3:4 format.
For the hand-drawn version, swap the first sentence for the loading screen one from line 1. Then go to line 7 and keep talking.
Two people, one cover
Jason and Lucia are a couple, and almost all of Rockstar’s GTA 6 art is built around the two of them. So a couple shot is the obvious next move, and it’s also where many filters give up, because they’re built for one face.
Gemini can merge several photos into one. Upload both and say in the prompt which photo goes to which character, for example: “Use the first photo for the man on the left and the second photo for the woman on the right.” Then add the face line for both of them.
If you want your names as the title, ChatGPT is the better pick for the lettering. And if you already have a photo of the two of you in one frame, LightX handles shots with several people too.
When you don’t feel like typing
Some evenings a prompt is too much effort, and that’s fine. Somake, Fotor and insMind do the whole thing in one click.
Somake lets you choose the aspect ratio before generating. Fotor adds a skyline with palm trees and then suggests turning the result into a poster. insMind works on the web without an account.
Somewhere in between sit the tools with ready-made prompts you can edit. Media.io has presets like “Vice City Neon 80s” and “San Andreas Comic Art”, plus a page of GTA 6 poster prompts, which is a good place to borrow lines for your own. Bylo AI lets you pick GTA 5, San Andreas or Vice City, or type something like “80s neon skyline”, and can put a health bar and “WASTED” on top.
Before you hit Generate on any of them, check what the service does with your photo. Some delete it after 7 days. One gives itself a perpetual license to publish it. We went through all of that, plus who allows commercial use, in our breakdown of 10 GTA AI filters.
One last setting
If you used Gemini, it adds a visible watermark by default. Switch it off in Settings > Media Watermark before you post. The invisible SynthID tag stays either way.
Long story short: one decent photo and seven lines of text. Post it, print it, or set it as your avatar until November 19.
Fancies writes about AI art for people who make it for fun. More guides like this at fancies.ai.







