itch.io

A new version of Everchanging Quest is available in a Hugging Face Space :
https://huggingface.co/spaces/Jofthomas/Everchanging-Quest

Hi, again (it's been quite some time). But the game is available in a space :
https://huggingface.co/spaces/Jofthomas/Everchanging-Quest

Yes I'll be working on a full fledged version of that game in Godot4, hence hopefully the points you raised will be fixed 

I watched your profile, it seems you are interested in the dungeon generation part, this is honestly, the part that work the least in my project. 

I wanted to function calls for object placement, but it raised a lot of errors and manual guardrails. As it was made for a game jam, once I did commit with this idea, I had to stick to it.

alright, sorry for the delay. I reuploaded a new version, I had to switch the voices, hence it might break again.

I'll be working on a version as an hugging face space now

oh yeah, the api key has been revoked since then. I'll build a new one version and comment back here when that's done ( in the week )

Happy you liked it !

Indeed, I think generative AI will have a huge impact in games, and the biggest change for me will be NPC agents ( from the transformers library, not RL agents ) but it would have been way too complex and unstable to implement in 2 days.

Hey again !  I have no problem with the implementation of the game and don't worry I understood what needed to be done. EVen more considering that all the controls are on screen.

when I said "  I don't have any feedback on what I type like what I see in your video " I meant I can't see the input field which made the game quite harder with no visual feedback of what I type. And resizing the window once the game started does not work as a solution.

I just gave it another try and in the end I have to resize the window before playing.

Have nice day !

Here is a representation of my screen with the game started :

Happy you made it to the top 10 ! Congrats. Well deserved, you are probably my favorite game of the Jam. Even if we'll be competitors, I wish you the best of luck ! ;) 

Happy you made it to the top 10. Gave you full grade for the efforts.

Alright, that's my bad. Sorry I'll follow the steps.

But I'm curious, about somthing. I just noticed the "server" you call is on LocalHost IP adress on port 8000. What kind of Backend is it supposed to call ? Because if it's for testing purpose I can understand. But on prod, maybe you forgot to change for the actual adress of your server ?

Sure !

I'll let you know :)

Indeed relaxing.

Now I'm curious how you animated generated images.

Was not meant to flex. Yes indeed harder on the last few ones, no did not get all of them, clicked too fast on one and got one wrong. so 11/13

Interesting concept, litterally going for the theme. 

I gues it might have been a lot of back and forth to find the correct prompt to have 6 kind of an object and also the previous image. ( but this might probably be generative fill untill satisfied )

As french I find it funny. Very clever use of diffusers.

A bit too easy, the right choice is a bit too obvious, but the overall game is refreshing and fun to play.

Too bad no one came ahah at designated hour . +1 for effort

Did you delete your game ? Can't find it. Can be very interesting in games

I also had the bug where there is no ennemies on the canvas. I died a lot due to me being bad.

Overall enjoyable experience.

Thank you !

I was pretty surprised myself how well it worked too !

I tested it and you have nothing to do with the bad performances of this game. Unfortunatly, GPT2 unless finetuned is a bad model. GPT-3/GPT4/Falcon-7B or even Flan-T5-Large or XL would have been better alternatives ( but some would probably not run/take forever to load if run locally ) :) 

Thank you ! Happy you liked it 

I like your idea, it was bit hard to start playing, I'm on PC and I don't have any feedback on what I type like what I see in your video. Your use of AI is still cool :) 

I don't see my open ai usage incresing, so I think something is going wrong in the backend.too bad I would have liked to discover generated maps

a working recipe, to fight for your cause. Well done

Hi again ! I'm trying the downloaded version, it's been 5 min that the room 1 is loading and still nothing ? Is it normal or should I wait more ?

Thanks ! It is a very usefull tool :) but the core about the game is function call. Acheck it out. I shared the code

I like your game, but the theme seems a litlle bit forced to fit.

Can you tell me a bit more about your pipeline ? it seems well executed :) 

ahah I can guess so...

alright, nice little game.

Very good game overall !

Controls are a bit hard to master, so I got beaten by your agents. Nice work, having the time to code this game AND train RL agent in such a short time. Deserves more visibility 

Alright, I'll give the downloaded version another try tonight :)

Ah,  that's a nice pipeline, it seems indeed very noise sensitive as any unprecented output in the early stages gets amplified at the end of the pipeline. in your case if the guidance toolkit was working properly ( chich is not the case, I tested multiple times ) it would have been interesting to have stable prompts :https://github.com/microsoft/guidance

But I judge by results, and in your case it worked well.

Seems interesting. I'll try to reconnect at 21 ( today 11/07/2023 at UTC+1 Paris time) and will probably wait for about 10 mins to see if someone is interested to play. 

Thank you for testing it ;) I'm a data scientist so my idea  was just to demonstrate the potential use of LLM. I actually never made any fancy games prior to this jam, so the gamplay is FAR from perfect. it's not meant to be hard or anything. just demonstrate something new.

I agree with you analysis, I tried to explain it, but if I had more time I would use transformer agents with falcon 7B and use diffusers tools to generate assets. Or another approach that I considered would have been to use a ClipSeg tool inside an agent to find a usable asset out of a big spritsheet of objects. In the end, I did not get the time to produce what I was thinking about.Thus this is only a POC

Good job :)  The game was cool.

Come check, my own implementation of the use of LLM in my game. I mostly used function calls from GPT4-0613 to generate content/interactions. I also shared the code to the LLM calls on github. It might be interesting for you .

nice. playing on a laptop with no mouse so i'm not really enjoying the experience to it's fullest ( it's hard this way ahah )

Hey ! the project seems ineteresting. I also tried to go with AI generated content ( more about stories and objectives on my side ), so I would have really liked to test it, but I can't make it work. Is there a bug that you're aware of ? trying my  luck with the downloadable godot version next.