It’s fine it’s just extra crispy.
2 Likes
Spraypainting bread dough is perfectly safe right?
(Will this get me punched to death because “Naked Robloxians?”)
https://i.imgur.com/TEqLoGL.gif
https://i.imgur.com/TEqLoGL.gif
2 Likes
Technically you’d just be painting your clothes.
I WANT A FACELIFT MACHINE TOO!!!
I can confirm this.
1 Like
WOW this is amazing 
2 Likes
That’s awesome but there’s so much B L O O M-so much that the image looks blurry. Drop the bloom a bit and let the detail shine through! 
1 Like
Actually, there isn’t much bloom. Its mostly Blur with a bit of bloom.
Heres a shot without any blur to see the difference. (I actually forgot to tune down the blur a while ago so thank you for saying that!)
4 Likes
That is awesome. Amazing map my dude.
2 Likes
Put some more work into this flying game.
It’s probably not going to go very far. simply because I wanted to have more realistic controls compared to other flying games on roblox.
5 Likes
More neural net stuff! This time the objective was to allow for more high-velocity drifting. Most of these guys are going over 200studs/s (some over 300!) and still manage to make the turn.
Here’s a graph that updates in real-time to show how each generation scores. Score is based on how accurately and quickly they follow their path. 100 extra points if they make it all the way to the end. (Hence that jump from 33 to 165)
Click for nerdy explanation of some stuff
The graph just scores them per generation. Green bar inside of the main bars is representative of how many of the attempted mutations failed (produced less points than it’s pre-mutation state).
You can loosely correlate green bars with how many mutations are about to be made in the next generation. (AKA the more green bar the more random mutations it’ll try)
The system starts out slow (aka all those 0 scores at the start) by randomizing all inputs. Once it has a car that scores points (completely by chance) It’ll start to mutate from that. If said mutation is producing a positive result (more points) then it’ll continue to mutate in the same way. Eventually it’ll reach a local low for the variable (picture a ball that’s rolled down a hill) and it’ll move on to other variables by random.
Steady growth tends to imply only 1-2 variables being tweaked, spikes represent a new mutation type.
Side-node: I’ve noticed the best growth stems from adding a slight chance of mutation even when the current mutation path is producing good results. Ex:
Gene 3 is mutating at a rate of +10% each generation. No other genes will mutate until this one reaches it’s most effective value, but I add a 1 in 4 chance that another gene will begin mutating with it. This tends to give me better results in the long-term.
5 Likes
Recommend not using a black background for these kind of renders, hard to see any detail. Instead use a darker grey background.
1 Like
Finally wrote a GOAP AI that is structured in a way I approve of and can manage w/ much more ease.
Its pretty simple at this moment, all the AI does is find a weapon, get ammo for the weapon and attack its target. In the near future I’m going to expand on its locomotion logic to find ideal locations to shoot from, and make a second AI class for the Target which will make the Target try to flee from its Attacker.
3 Likes
There will also be a system where they will find the best possible way to get toward these multiple objections and combat correct? Lets say that the Ai has a melee, would the Ai decide to either melee or stay afar and shoot?
1 Like
Making a TrelloAPI. It’s a side project for me right now. Anyways, I have plans for the future to make a parkour fighting style game or just update my current games.
1 Like
I made the images with transparent background, but the utility that generated the GIF just splashed a black background, and I found myself unable to change it.
I wouldn’t spend 15 minutes to render it again with a background so 
1 Like
At the moment there’s no pathfinding, but I’m wanting to have paths take in account what would be the safest way to get to each objective, but yes the AI does try to find the best combination of actions to achieve its goal. Its goal at the moment is to kill a target, but in order to kill a target, the AI needs a weapon, and to get a weapon it needs to find a weapon. Once it has a weapon it checks if it has ammo, if it doesn’t then it finds ammo. Once the AI has both a weapon and ammo, it will move to attack the target. If the AI runs out of ammo mid-attack, then the AI will run to the nearest ammo stockpile and restock before re-engaging the target.
2 Likes
I’ve just been working on showcases this year, trying to build up a portfolio. Right now, I’m working on a horror themed chain of games, one leading to another.
Here’s a screenshot from the first one:
5 Likes