I believe that students who recently brush the small red book should be able to see this kind of chance encounter with the stars, the stars across the century photos, some of the characters look very reproducible, tone, light, skin color and so on are very real.
Did a Keanu Reeves and Guan Xi brother, can say Keanu Reeves is basically 100% restore the real, but Chen Guan Xi is still almost meaning, which is related to the amount of Asian picture training.
Artwork Effect
But the photo between the stars is not addictive, to meet the stars by chance to have meaning ah! Yes, Ghibli style was popular a while ago, and "realism" is popular these days, especially with this kind of fuzzy and shaky camera, it can really fool a lot of friends who are new to AI or unfamiliar with AI.
However, it needs to be seriously declared that all these group photos are only for technical demonstration and entertainment creation, and commercial use is prohibited, such as infringement will be deleted.
Production method
Making it is actually quite simple, just go to Gpt4o and type in a section of instructions and you're ok, if you want to make your own andgroup photo of celebritiesIf you still need to use a photo to pad the picture, the next step will take you to do together, learn to figure a fun, maybe after the AI video need to use this skill?
I. Gpt4o generates stars
- Tool: CHATGPT
- Website: https://www.chatgpt.com (magic required)
1️⃣ Open Gpt directly, select the 4o model and enter the command
- You are asked to draw an extremely mundane and uninspired iPhone selfie with no clear subject or sense of composition-it's like a casual snapshot. The photo is slightly motion blurred and slightly overexposed due to uneven sunlight or store lighting. The angles are awkward, the composition is confusing, and the overall presentation has a deliberate sense of mediocrity- like a selfie accidentally snapped while reaching for a cell phone from a pocket. Characters are Keanu Reeves and Chan, 9:16
2️⃣ came out great, TPB this case was ok once.
3️⃣ You can restore to 90% or more if both are foreigners.
Second, a photo of yourself with a star
It's important to highlight here that there are generally two ways that real people can be photographed with celebrities at the moment:
1️⃣ Gpt continues to mat map creation
2️⃣ i.e. Dream Mat Chart Creation
A lot of people say that gpt doesn't look like them, and that imo is a bit more like them. But I think that even though Gpt is "similar", it's really heavy on the AI flavor, and it also requires a lot of card draw.
This is based on the effect of the method on the market in the instant dream:
Adjusting the pad parameters and commands many times, there is still no movie, either Keanu Reeves does not look like, or I do not look like, or I and Keanu Reeves "merged".
So it's recommended to stay with Gpt, but for this need to draw cards multiple times, there's actually a faster and better way to use it - Sora
1️⃣ Click on "Sora" in the upper left corner or open the Sora website directly.
- https://sora.chatgpt.com/explore (magic and gpt membership required)
2️⃣ Click "Images" to generate images immediately.
3️⃣ Continue entering commands:
- You are asked to draw an extremely mundane and uninspired iPhone selfie with no clear subject or sense of composition-it's like a casual snapshot. The photo is slightly motion blurred and slightly overexposed due to uneven sunlight or store lighting. The angles are awkward, the composition is confusing, and the overall presentation has a deliberate sense of mediocrity - like a selfie accidentally taken while reaching for a cell phone from a pocket. The main character is the first picture, with the features to look exactly like the photo, at night, next to Keanu Reeves
4️⃣ Parameter Adjustment
Respectively, Image➡️2:3➡️2 images (at most two for now, the gpt math for four is not enough for now) ➡️None
5️⃣ Upload Selfie
6️⃣ Start Drawing Cards
Sora will spawn a bit faster than Gpt, and without that sudden intermission time limit that Gpt has, it can also produce two maps at the same time (and four later)
7️⃣ As mentioned above Gpt Asians are still weak, but it's expensive to have a realistic style, and after proper card draws, people just choose the one that most resembles them!
Sora still smells pretty good when used this way! It's still recommended to keep using Gpt if you want to keep your characters consistent, so you can create them in an integrated context; but if you need this kind of heavily card-drawn, customized, authentic picture, try Sora!
Well, today's hands-on tutorials are here, if you have gained must point to praise, look at, forward ah! Hurry up to practice it!
Final Thoughts
This way, AI painting realism, daily sense, Gpt / Sora really stand alone ah, another good tool to use!
To do a good job of AI video, AI painting is the most important step, the AI painting tools on the market have their own unique and distinctive styles, I hope that we can all follow along to learn ah ~!