How to use AI to generate an "ordinary" iPhone selfie that goes viral online?

Everyone on the internet is playing this lately.AICelebrity photos!

Jay Chou makes low-key appearances in various places, Elon Musk and Jensen Huang drink milk tea by West Lake, Ultraman and Einstein take a selfie together...

ThatpracticalGPTWith a temperature of -4°, you can easily create fun group photos with a "casual snapshot" feel!

The following is a step-by-step tutorial.

After comparing several tools, I think...GPT-4o has the best grasp of realism in these "casual snapshots"~

No need for complicatedPrompt wordsThe natural language generation is very effective, and it can be continuously adjusted through multiple rounds of dialogue.

We open ChatGPTWebsite, selectGPT-4o model (freeUsers can try it out; the Plus version offers higher credit limits.

Simply copy the following template and modify the content within the brackets:

"Generate an extremely ordinary, unremarkable iPhone selfie with no clear subject or composition—like a snapshot taken on the go. The image includes a sidewalk, a green hedge in the background, slight motion blur, overexposure, an awkward angle, and a chaotic composition, all deliberately mediocre overall. The characters are [Jay Chou + Putin], the background is [the Great Wall], and the aspect ratio is 9:16."

For example, Jay Chou posing for a photo with Vladimir Putin:

Lei Jun and Steve Jobs together:

The realism is absolutely perfect!

When these images are generated, there are also someTips,for example:

If we want to generate Asian celebrities (such as Jay Chou), we can add "Authentic and natural style"",Fine skin"",Avoid cartoonization"This type ofPrompt words.

If you want to mimic the "candid photo" feel, you can add "Passersby blurred"",Backlight silhouette"waitPrompt words.

We can also make photos more lively by changing people's expressions. For example, inPrompt wordsAdd "Smile"",sad"",Awkward"etc.

Many at a glanceAIThe photos all have this common problem:

Professional composition, exquisite lighting, skin so smooth you can't even see pores...

Everything was "perfect," but it always seemed to lack a certain "human touch."

At this point, "imperfection" becomes the soul of realistic photography.

We can add it to the requirements.Motion blur (e.g., shaky hands, motion trails), uneven exposure (overexposed/underexposed areas), awkward composition (unconventional angles).These elements trigger the model's ability to capture the randomness of life.

Based on the imagined scene in which the character would be situated, and the character's selfie habits, the concept of "awkwardness" was specifically extracted and added.Prompt words.

For example, add a prompt based on the original:Uneven lighting at night, use the front camera to focus on the face + relaxed feeling of travel..

You can then get a straight man's selfie version of a cat traveling:

Please draw an extremely ordinary, unremarkable iPhone selfie, lacking a clear subject or composition, like a snapshot taken casually. The photo has a slight motion blur.Uneven lighting at nightThis resulted in slight overexposure. The angle is awkward, the composition is chaotic, and the overall effect is one of deliberate mediocrity, like a selfie accidentally taken from your pocket. (Using...)Front-facing camera, focusing on the faceThe main character is a tabby cat.The relaxation of travelThe background shows the Eiffel Tower illuminated at night in Paris.

It is important to note thatPrompt wordsYou need to determine the breed of the cat.If you directly write that the main character is a cat, you might get a really straight-man photo:

Good visuals need to tell a story.For example, the recent viral trend on Xiaohongshu:

A middle-aged Lu Xun poses for a photo with Peking University, accompanied by a caption expressing his feelings about leaving his job;

A durian fell onto the grass, and an angry Newton angrily rebuked whoever had placed the durian on the apple tree.

While the viral posts on Xiaohongshu are certainly brilliantly written, unfortunately...AIIt's still hard to understand why people would chuckle when they put durian and Newton together.

If directly in the existingPrompt wordsWrite"The main character is Newton, and the background is a durian falling on the ground, showing Newton's angry expression after being hit by the durian."

All that was produced was a picture with a slightly absurd and humorous feel:

If you don't watchPrompt wordsLooking at the picture alone, it doesn't feel like a joke about Newton at all.

However, by simply adding a slight restriction on the background, the effect becomes completely different:

Please draw an extremely ordinary iPhone selfie, lacking a clear subject or composition, like a snapshot taken casually. The photo has slight motion blur, and uneven lighting or indoor lighting causes minor overexposure. The angle is awkward, the composition is chaotic, and the overall effect is a deliberate mediocrity, like a selfie accidentally taken while taking your phone out of your pocket. The subject is Newton, and the background is a durian fallen on the grass under a tree, showing Newton's angry expression after being hit by the durian.

The personality and identity of celebrities are generally fixed, so as long as the scene and interaction details are set up, a photo that matches our imagination can be generated.

If you want to generate a photo of cats and dogs that looks natural and matches your imagination, you need to define the roles they play.

For example, this time, I originally wanted to generate a Flying Kitten with Soaring Ambition, but I only got a Cowardly Cat Passenger.

If we are in the original diagram abovePrompt wordsAdd some descriptions to the basics, such as:The photo has slight blurring due to vibrations and motion from high-altitude flight."",The main character is a tabby cat pilot who feels the urgency of aerial combat, while the background shows a view from the cockpit of a group of fighter jets engaged in combat."

The art style immediately changed:

Referring to the previous approach to creating raw images, we can also use basic templates.Prompt wordsBased on this, with slight modifications and limitations, incredibly realistic "photo scams" are generated.

For example, using the recently popular bean bun input method...Prompt wordsTo create, choosePortrait photography style.

The image style is "Portrait Photography." Please draw an extremely ordinary iPhone selfie in front of a mirror. The subject is a cute girl dressed in a JK-style cosplay outfit, taking a snapshot with the rear camera in front of a full-length mirror in her carefully decorated, cute room. The photo has a flash on, slight motion blur due to insufficient shutter speed (if blur is required, it should be a slight overall blur), a messy composition, and emphasizes the room environment reflected in the mirror (from a first-person perspective), highlighting the blurred light and shadow relationships. The overall impression is one of ordinariness and everyday life, with a aspect ratio of 9:16.

Doesn't the final product look great?

It's exquisite, yet still feels realistic.

To some extent, it can satisfy the needs of users who prefer high-quality visuals.

Let's go back to the original template, finalize your protagonist and background, and let the colorful world in your heart bloom in "reality".