Imagine spending eight fkn thousand dollars for the MBP and not getting the fucking AC adapter 👏🏻 who’s brilliant idea was this? “it’s because of the carbon footprint” but I have to buy a separate fkn box to get the fkn adapter to be able to charge my 8 fkn thousand dollars MBP. (chef’s kiss)
Wow! This is crazy 😱 it’s my first ever GitHub sponsor for my work on open-source ❤️ I cannot express how grateful and happy I am about this milestone 🎉
It’s crazy and scary, very scary
I honestly think that it’s crazy that someone could just mark you as something just because they feel like and without even knowing a shit about the other person…
So people can make public lists of others and mark them as something they’re not and ruin them and cause a bad reputation just because someone is just being a prick and does not like that I just wanna use AI to prototype instead of spending thousands of dollars on someone that might even be worst?
Also, GPT Image 1.5 was the most expensive one, Gemini PRO after, and then xAI as the cheapest one :)
Also, xAI was by far the fastest model to output the image, and OpenAI was the slowest one
Worth mentioning that the only one that supported transparent backgrounds out of the box was OpenAI, for Gemini, I had to craft an erosion tool to erase a magenta background, which didn't even come as a single color but in a range 🫠 For xAI maybe I forgot to configure something
Same prompt, xAI, OpenAI, and Gemini, which one looks the best?
I asked for image generation AI models. A random person came in telling me that I should hire an artist, I just replied it’s just to prototype and I’m not hiring any artist just to prototype and I got blocked? 😅 guess people have a very thin skin
Well, this is just for me to prototype games and try stuff so I’m not going to actually hire any artist just to create game prototypes 🤷🏻
What are some good AI models for image generation? I’m working on a game asset generation platform for myself and I’d like to test different models beyond NanoBanana and GPT :)
The inlay-proto.up.railway.app seems to be broken?
Yeah, that’s actually true! I need to fine tune the image resolution based off the prompt and add more settings to adjust output resolution! Thanks for bringing this up! ❤️
Pixel art still has some artifacts, but I think this is due to Gemini itself not creating proper pixel art, because it's creating giant "pixels" in a much bigger image canvas, so that's why it's not looking that good, and I'm getting artifacts. Also, my erossive script is maybe a little aggressive 😅
> Style: Hand-drawn. 2.5D-style armour for a Diablo-like roguelike game. The armour should resemble a demon. You can take inspiration from Diablo's demon hunter and World of Warcraft rogues. Standalone item draw. No character. Avoid: blurry, low quality, poor quality, characters, enemies, npcs.
Wow, check this out! The background removal looks sick when it comes to hand-drawn art style 😱
proceed with the removal of those. I'm learning a lot with this project and burning quite a few tokens and quite a few dollars in asset generation with the Gemini API, but hey, knowledge has no price, unless it ends up getting me broke. At that point I wouldn't care because my wife would 🔪 me 🫠
Gemini to generate the backgrounds with this particular #ff00ff shade", and you know what? Gemini does not give a F$c# about the hex you tell him and it generates a different shade of magenta, so you gotta due hue-based magenta detection across the sprite to identify the magenta shades and then...
receiving the output. But there's also another catch, usually in game assets, designers have this accepted pattern of using this particular shade of magenta #ff00ff to symbolize transparency. It's wildly used as it's a color that's not used at all in game design so I told myself "ok, let's ask...
and the result ended up very promising, but as you can see from the screenshot, there are still pink/magenta artifacts remaning from the image manipulation, because Gemini does not generate real transparent backgrounds, it lacks of alpha channel outputs and I had to manually do it myself after...
artist I decided that "why not try to build my own asset generation pipeline using Google's NanoBanana so I can prototype something more appealing than colored squares?" and there's the result. I've been slowly iterating on generating sprites and spritesheets that allow for transparent bgs...
A fun project that I did over the weekend when my wife and baby were napping, even though I promised myself I would not touch the computer on the weekends 😅 It's an AI tool to generate game assets. Since lately I've been playing around with @phaser.io and @godotengine.org but I'm not a graphical...
Hey @bsky.app I think the OG is broken only here because I’ve been testing on other apps and OG testers and works fine, could bluesky not processing the URL correctly?
Will do :)
And somehow my package gets redirected into a blank page 🤔
Hey @npmx.dev what am I missing on my package to have a better looking OG generation? 🤔
Some www.curl-runner.com analytics for the current period (30d) –by pulsora.co
Well, there's even docs about it now!
Check them at graphdb.alexvcasillas.workers.dev
I've used @cloudflare.social pages and workers and @astro.build to quickly get the docs up and running!
