This is what I love about AI images. You can easily explore super creative ideas like that
Also, beautiful results, you and Hylian
This is what I love about AI images. You can easily explore super creative ideas like that
Also, beautiful results, you and Hylian
The AI generated the image, and then you corrected its mistakes right? Belongs here, I’d say.
It would be different if it were the other way around
(Nice one, by the way)
Nope. The character I found on the net and inserted into the overlay. And did some adjustments.
Anyway, here are two overlay+presets for Castlevania if anyone’s interested: Castlevania overlays
I see, so a good amount was made by you. This is what’s interesting about this stuff: at what point would it not be considered AI content anymore?
I’d say since AI did initial heavy lifting it belongs here, but you’re free to post anywhere you want
Personally I think most AI images benefit from some amount of editing, so I’d include any non-overhaul edits here. I’ll change the main post to reflect this.
Heads up:
There seems to be a new free image generator that does a lot better with text
Still may struggle with text it has to make up, but any text you specify is written correctly.
Front view of a CRT television at night. The screen is turned off and reflects the room. The TV is labeled “Toshiba”. (the prompt is auto-enhanced on this site)
Perhaps any gibberish can be avoided by specifying the button labels, or lack thereof.
Reve was updated and it’s one of the most advanced for making Overlays for free now.
Some tests I’m doing for Spooky TVs: All and these Spooky Tvs plus uborder presets
The prompt:
A vintage 1970s television with a heavily worn wooden cabinet faces perfectly straight ahead, positioned centrally in the frame in a decaying, pitch-black room. The TV is photographed head-on, showcasing its symmetrical design with the convex screen dominating the composition. The aged wooden frame displays deep scratches, water damage, and peeling veneer in perfect detail, while its dust-covered screen shows age-related discoloration. Tarnished and partially broken decorative knobs and channel selector are clearly visible below the screen. Scattered across the rotting floorboards around the TV lie three partial human skeletons in chaotic disarray - ribcages separated from spines, scattered vertebrae, and detached limb bones, all draped in thick, gray spider webs that glisten in the moonlight. Several skulls rest at unnatural angles, partially obscured by debris and webs. Moonlight seeps through broken windows behind the TV, creating an ethereal backlight that perfectly outlines the television's weathered silhouette. Cobwebs stretch across the screen's corners and entangle the cabinet's ornate details. Cold blue moonlight illuminates floating dust particles around the television, creating an unsettling atmosphere. The room's peeling wallpaper extends outward from behind the TV, while the edges fade into impenetrable darkness.
Good result, but… What a prompt! lol
Remember you can have a simple prompt and tell an LLM like ChatGPT or DeepSeek to make it super-detailed!
(Additionally, some image generators like Reve can auto-enhance your prompt)
So, ChatGPT recently released what’s probably the best (publicly available) image generator yet. It seems just now they’ve also made it free of charge for everyone, and this is my first attempt:
A CRT television at night, sitting on a wooden desk. The TV is turned off and facing the camera.
It lets you keep prompting it for edits, so I did just that:
Extend it so it’s 16:9
Add a big pile of PS2 game cases on the wooden table, next to the TV, and a small action figure of Snake on top of them. The games are at an angle and each game is labeled with its name. Include Shadow of the Colossus, Metal Gear Solid 3 and Resident Evil 4
(I tried with Sonic but it didn’t let me…)
The image looks good, but I need more focus on the TV. Please make the TV perfectly face the camera and have it more in the center. Make the games and Snake smaller so everything can fit!
Has its errors, like making Snake Eater a separate case, but still, the rate of progress is nothing to scoff at. More could probably be fixed/added if I kept prompting it.
AI Image + SOQUEROEU overlay:
Link if image is blocked: Link To Image
Link if image is blocked: Link To Image
Astonishing. It keeps getting better and better. Forgiving classical AI mishaps (like “The Legend of Zelda” poster starring Mario dressed as Luigi), everything else is cohesive and aesthetically pleasing. Bizarre outputs are ceasing to be.
These are really cool! I’m not against A.I. I just don’t want it to entirely replace art done by people. The A.I. that I used for some bezels had to be touched up. A.I. gets it wrong most of the time. For me A.I. is a good tool, but it still needs that human touch to make it look right.
Yes, I agree. I had to combine the awesome work from SOQUEROU with AI output using GIMP. The usage of AI enhance the results in some ways and degrade in others, so you need to pick the good portions.
This is crazy good!!
Are those 100% AI or did you need to tweak them in Gimp?
GIMP was used only to increase the TV useful areas. The characters are 100% AI.
I’m not an artist and I have no talent for touching anything other than simple geometric shapes.
Great work, as usual. Perhaps you have not the practical skills, but you do have a good artistic sensibility. Your shaders and overlays are always aesthetically pleasing, you can’t do that with technical abilities alone. There are plenty of content (lots of them professional) that fall flat because they’re just cold skill, but without a purpose.