A downloadable game for Windows and Linux

Download NowName your own price

Please see the devlog for new and updated features!

WHAT IS IT?

Mutate 212 animals from all across the animal kingdom and make them fight in this fully local, offline, LLM-generated narrative battle simulator. Your monsters! Your commands! Their actions!

This game utilises eight customised LoRA adapters carefully constructed by author Mitch Davis (https://akickintheteeth.com/). These are sets of words and phrases and ways to use them assembled into rich datasets that, somehow, manage to underpin SLOP FIGHTER's unique, dynamic, and most of all dramatic narrative battle system. There are thousands of combinations of species and types, and tens of thousands of varied, expressive responses for every situation. You can even feed your monsters between battles. 

You wield a team of three monsters against either CPU or PvP opponents and use fully LLM-generated movesets entirely unique to your monster against them. The battlefield is yours to command. Command EARTH monsters to shake the ground or AIR creatures to strike from the stratosphere. Utilise directional movement to your advantage by giving custom commands. The only limit is your imagination!


EIGHT TYPES

Each of SLOP FIGHTER's eight mutation types has its own personality:

FIRE Passionate and intense.

WATER Fluid, patient. Like the tide.

AIR Light and playful.

EARTH Steadfast and immovable.

MUTA Broken, unhinged, and grotesque.

TECH Calculated and precise, like a machine.

COSM Detached and ethereal, both cosmic and mystical.

SHAD Hollow and deathly, nightmarish.


212 ANIMALS

SLOP FIGHTER draws from the rich well of the kingdom Animalia and all its phyla. Mutate random animals from all across it, like bears, squid, tigers, and vultures. Mutate almost any animal you can imagine (including the bobbit-worm!). Every animal in SLOP FIGHTER is defined by its species' traits. Imagine a bird's keen senses, a wolf's territorial instincts, or a mole's burrowing ability coming into play. Every animal shows its teeth, or its claws, or its tail, in its responses. That is what truly gives SLOP FIGHTER its unique edge - every response is weighted by animal traits that define its battle style. Both predator and prey have full opportunity to use everything to their advantage in SLOP FIGHTER's robust battle scenarios.


HOW DOES IT WORK?

The game runs entirely windowed at 400x240 resolution. It loads from your computer's status bar, either on the right or left side of your screen. It does not divert attention from anything else you have onscreen. It's entirely unobtrusive, and play is casual. Change the options for loading side in settings. You play the game entirely via keyboard. There is no mouse either required or necessary. Use arrow keys to navigate menus, and type commands where required.

PvP play occurs entirely over local Bluetooth. There is no internet server cloud-based dial-up wifi support or LAN play. SLOP FIGHTER uses Bluetooth Low Energy (BLE) or traditional/manual RFCOMM to communicate with other systems. As such, PvP play requires both players be within Bluetooth distance of each other.

At battle start (for PvP) a rough BLE RSSI signal measurement is taken, which works as a VERY rudimentary distance measurement. There are three distances: CLOSE, MID, and FAR. This measurement defines the position of both monsters at battle start, so pay attention! Your monsters can use RANGED attacks from FAR distance for better effectiveness, but if you want to start CLOSE, you'd better be ready to brawl.  You can even move through the battlefield! Just give direction via a custom command.

Once the battle begins you're off to the races, and now you're using your monster's unique mutations and biological heritage to its advantage to defeat other weird, mutated biologica. The game uses status effects, misses, and critical hits for even battles, and despite the simplicity of the game's display, a score of complex calculations are made under the hood for each individual animal and mutation type and altogether will define damage dealt.


So have fun! Share your monster's wildest responses with others by commenting below! Tell your friends about the game! 


Check out more of my work at https://akickintheteeth.com/


Yes, this is an entirely vibe coded video game. I would hesitate to say that it uses AI. I think the world got carried away with the wording there. I am an author. LLMs were made both for and by people like me (not Sam Altman). What I've done is more like what disc jockeys do mixing and chopping sound bites than building artificial intelligences. I appreciate the support I've had utilising LLMs, and do entirely believe there is more value to them than, well, warfare against citizens. Look closely and SLOP FIGHTER might show you what I mean.


Uses Google Gemma4 2B for inference

Fine-tuning provided by Unsloth

Runs with pygame and llama-cpp-python

Updated 13 hours ago
StatusReleased
PlatformsWindows, Linux
Rating
Rated 5.0 out of 5 stars
(1 total ratings)
Authorquarter2
GenreFighting
Made withpygame
TagsAI Generated, Casual, Indie, Monsters, Multiplayer, Pixel Art, PvP, Retro, Turn-Based Combat, Word game
Asset licenseCreative Commons Attribution_NonCommercial_ShareAlike v4.0 International
Average sessionA few minutes
LanguagesEnglish
InputsKeyboard
AccessibilitySubtitles, High-contrast
MultiplayerLocal multiplayer
Player count1 - 2
LinksBlog, Ko-Fi, GitHub
AI DisclosureAI Assisted, Code, Text

Download

Download NowName your own price

Click download now to get access to the following files:

slopfighter-linux.zip 4.9 GB
Version 3
slopfighter-windows.zip 4.8 GB
Version 2

Development log

Comments

Log in with itch.io to leave a comment.

Actually a fun concept, but obviously the Gemma4 2b makes it quite heavy for lots of devices. 

It didn't work correctly on smaller LLM?

(1 edit)

Gemma4 2B is the smallest of the Gemmas. Google made it with "efficient architecture", also, which I am told makes it more effective on less powerful devices, like phones. The difference is about 3gb of filesize, and from what I can tell from using it it is only marginally slower than Qwen3 1.7b (but way more capable). It still runs fine on Raspberry Pi 5.


It did not work correctly on a smaller LLM, but without deep insight into how the LLMs are constructed it's really hard to tell what they're fully capable of. It's actually crazy how different Qwen and Gemma are in inference styles. I really like Qwen, it's a very straightforward sort of LLM. Gemma4 is somewhat rambling. I'm certain that in future we could get this sort of thing down to a sub-1B model, just not right now.