why?
Extintor Incendiandose
Creator of
Recent community posts
One of the things you've probably noticed is that the app may not work sometimes, this is because the rate limit of the API in use is only 50 queries per day, In this new version, apart from changing to a model with a higher query limit (I saw that Gemini 2.5 Flash Lite has a limit of 2K per day) In addition, I will probably now be forced to add support for load an apikey from the SD card so that anyone can use their apikey, so when mine stops working they can change it for their own that does work.
I've been testing by modifying the general prompt, and I think I've achieved a huge reduction in censorship simply by adding a short sentence.
Although this adjustment within the personality prompt doesn't have much effect, it definitely needs to be made to the general prompt, so I'll have to implement it in an update.
The previous API didn't have any censorship, unlike this one. I'd really like to go back to the old API if I could.
Although I don't know if modifying the general prompt could achieve a similar effect, I'd have to test it and see what I achieve for the next update.
What I'm sure of is that with enough conversation the character will open up more so that you can have a very explicit, uncensored conversation.
I'm sorry about that. It really caught me off guard when, out of nowhere, the API I was using suddenly stopped working, and support didn't respond to me.
With these changes, old characters won't work, but it should now be easier to create new ones, especially with 64x64 pixel PNG support.
And with the new API, the responses are more accurate to what the characters would say.
Well I've been thinking a bit (a lot) about adding the ability to create characters within the application, including the ability to generate images and...
I almost hit the wall, I didn't understand almost anything about how to render a png image on the screen (since as far as I know you can't create t3x files within the console) and this was the first step to be able to connect to some image generating API
The number of times I had to use artificial intelligence to be able to do this with my few skills was monumental, and even then I couldn't even manage to show a distorted texture on the screen
In the end I was able to do it but using libraries that were discontinued 8+ years ago (sf2dlib and sfillib), with that and updating the libraries a bit (on my own and with the help of AI to understand which function I should replace the obsolete ones with) I was finally able to render a 32-bit RGBA PNG image on the screen at a resolution of 48x48 pixels placed in the sd card root.
I then tried connecting to an ai image generating api and it worked a bit... the resulting image was 512x512 pixels, exceeding the screen size by quite a bit, although shortly after that could be fixed by changing the scale so there's just one small big detail left
The price, ai image generating apis generally have a relative cost depending on their capabilities, the api I was thinking of using already includes a few free credits to use but once they run out... I don't want to get into debt.
So I was thinking of developing a paid version that includes:
- image generation with AI
- use a more advanced LLM with a longer text length
with a small BIG limitation
- It can't be used forever, only until the credits per user are exhausted
Approximately $1 can be used to generate 200 images in the API that I've been testing for a while, and with $0.4 you can get approximately 1 million tokens in responses using the Google Gemini API (I don't know the prices of other APIs)
So I would really appreciate comments regarding this idea to see if you think it's prudent or if it would be worth a version of the app that allows you to do this type of things and at what price for the amount of use (taking into account that RAI Loader gets fewer downloads every day)
Well, it depends on whether you are going to do a free public personal project or something serious and commercial.
You could use Unity but that brings piracy and performance problems, a good option is Octave or creating your own "engine" for your game (That's what I did with devkitpro and C++, creating the "engine" for RAI Loader)
Unfortunately Godot for 3DS still has a long way to go to be usable.
it's pretty easy to create a .t3x file, you just need to have devkitpro with the tools for 3ds
i'll tell you something curious? RAI Loader allows you to have multiple images for the same character, they are chosen randomly every time you load the character.
This little guide made for Universal Updater works perfectly to make .t3x files for RAI Loader (just remember that the .t3x file must have exactly the same name as the JSON file)
https://github.com/Universal-Team/Universal-Updater/wiki/sprite-sheet
Regarding the character limit... it cannot be increased, I have used a very limited free api, Maybe it would be possible if I added the ability to use the GPT chat API and have everyone enter their API key, but I would have to make big changes.
Another thing would be to create a paid version with what this one lacks (save chats and use an API with longer response length) But I don't think it's necessary
Yes, you can create your own characters, you just have to edit the JSON file following the same syntax, the content with the role ai is a system instruction while the content with the role assistant is the first message from the AI
The .t3x file is the character image, it's optional but it looks pretty :3
participants will have to make a game inspired by... Animal Crossing

your game can be inspired by any aspect of the game.
example: color palette, character design and many other things
you can be inspired by any game of the animal crossing saga.






















