Let's share our tools and help each other make the best games possible!
Here's my setup for the jam:
Code: Claude Fable in VS Code
Art: Probably GPT Image 2.
Music: Suno 5.5
Engine: LittleJS. It's been working amazing with Claude. I'm using the version in this repo that has extra AI tools: LittleJS-AI
What's everyone else planning to use? Drop your stack below. 👇
Hermes agent powered by ollama cloud GLM 5.2 to compile thoughts and generate prompting and instructions for Fable in Claude Code, MCP comfyui and MCP Blender (comfyui will be using qwen variants) and of course suno 5.5 for music, and elevenlabs for voice and sound effects.
@Focacci - If you are paying tokens for image generation you must stop and try comfyui it is totally local and the generation quality honestly can rival frontier models. Try it it costs you nothing but some time and bandwidth! Also you can hook directly to api and agent can alter and run workflows. With qwen image to image editing workflows you can keep perfect character consistency across unlimited generations all locally. (I am not being paid by comfyui, but if they see this post... LOL)
Hey Uh.....New Here. I guess this would be a good motivation to get my feet wet again in game making, I know my termonlogies and I can read basic code but thats about it. I'm only familiar with Actionscript, Phython, and game maker. Stepping outside my comfort zone once again.
I already had a strategy for game design before this, I'm a graphics designer before a programer so I shouldn't struggle too hard since AI will be helping me with the code.
For Coding: I use primarily Claude then alternatively Gemini ( for either double checks, cross references, or just comparing languages on both to see which is most effecient in the specific task
Art: IF i'm not doing the art style myself from start to finish, My go to are Google Gemini, Bing, and Merlin's image generation isn't bad either.
Music: In my opinion I think Ai music generation still needs work but i'll just use any good free music Ai generator then edit myself in FL or Audacity
Engine: Not Sure Yet, But I want to try JS Engines, already downloaded a couple.
For music, try Suno 5.5, It is getting a large following in the professional music scene and some artists even sponsors remixing contests. The real truth I think is that for real audiophiles it may never get there but for the vast majority of your user base the songs are enjoyable, I even saw a small You Tuber talking about how they have quite a few Suno songs on their real playlists.
Late to this thread, shipped my entry yesterday so heres the actual stack.
Code: Claude designs and writes everything, Claude Code implements and deploys. Two AI pipeline, one describes, one builds.
Art: nothing generated, all drawn by code on canvas
Music + SFX: also generated by code, pure math rendered to WAV when the round starts. No Suno, the game composes its own 60 second track
Engine: none. Vite + react + one canvas
Entry is Chargo if you want to hear what a math bassline sounds like. Theme fit was pure luck, the game was already about holding a charge
Awesome setups! Here’s my setup for the jam
Code / Engine: Vanilla HTML5, CSS3, & JS (Custom Web Engine)
AI Assistance: Claude 4.6
Assets & Audio: OpenGameArt.org (free community assets for now—planning custom art and audio updates soon!)
Avatars: DiceBear API for avatar generation (also planning on creating custom art for avatar creation)
Late to this, but here's our stack plus one measured number for the image-cost debate further up.
Engine: none. Hand-written HTML5, one canvas, vanilla JS — same place Toklesar and Bakko.GG ended up. Code: Claude Code writes it, disclosed on every entry. In-game art: drawn by code, the build ships zero image files. Audio: ZzFX, a few hundred bytes of maths — worth knowing that self-contained procedural synthesis is not generative AI, so it does not need an AI-audio tag. AI images are used for cover and store art only.
On cost: we rendered the same cover prompt at all three gpt-image-2 quality tiers and measured low $0.0082, medium $0.0706, high $0.2813 (and over two minutes). High is 34x low and it lost — richer light, but it resolved to an anonymous glowing shape at the ~350px thumbnail a store actually renders. Judge the tier at thumbnail size; we would have kept paying for high otherwise.
One trick worth stealing if you need trailer footage in a hurry: do not hand-record it. We run the game headless and step it a single frame at a time through the render tick, screenshotting per frame, so a GIF is reproducible — same script, same frames, every run. You can re-cut a trailer after a balance change without replaying anything, which matters on a jam deadline.
OTTERBOUND is our pick of the entries if anyone wants to see where it lands. Happy to go into detail on any of the above.
For SEN, my stack was:
Code: Claude
Debugging / development support: Claude / ChatGPT
Graphics: Claude / Fab
Audio: Fab
Engine: Vanilla HTML5 / JavaScript / Canvas
Claude handled most of the implementation and also created the in-game graphics, while I focused on the game concept, mechanics, and overall direction.
Claude code for code. And codex cli for art generation which is driven by sprite maker. For controlling unity I used Unity MCP the open source version in tandem with MooseRunner for testing (disclaimer my own tool).
The flow was crazy. Opus 5 was a killer at utilizing all tools. I did almost all dev over my phone remotely. Playtesting and balancing and adding game mechanics and updating code architecture. Even though I have not seen one line of actual code the modules looks solid and well structured as the test also did.
Crazy times. It was fun but the hardest part was art generation. Especially the tile set. Whish I had some time by the PC. Sadly I did not and lost about 45% of dev time on tileset generation.
