Video wird geladen...
Video konnte nicht geladen werden
yeah this is peak
48,829 Aufrufe • vor 2 Tagen •via X (Twitter)
16 Kommentare

Back then, we deliberately chose to build Beat Saber beatmaps by hand, as the quality of generated content simply couldn’t match it. Human touch will always have its own magic, and I wouldn’t give up on that. But I can see the tech getting to a point where I’d actually want to test some generated maps based on my preferred style.

oh this is sick! i have an eval suggestion: can you put in something like kanaria - king and see if it does something reasonable for the "left side, right side" part? they messed this up in pump but it's quite good in maimai

what llm are you using? for a start this looks scarily promising

which am i using to help build/train this or which am i finetuning for this?

Both!

the model itself is a qwen3 0.6b with a *lot* of conditioning and audio input layers stacked on (MERT, custom audio projection layer) and heavily tuned. chose this over a pretrain from scratch since i can inherit the attention patterns qwen learned in pretraining, plenty of useful geometry can carry over to mapgen. music adherence (where all the custom model design comes in) is still its weakest point, idt the current iteration will be what ends up shipping my collaborators are multiple instances of both fable 5.1 and gpt6 astra. one of the sessions iterated for 6 days mostly unsupervised, most others were heavily collaborative - i prefer being in the loop and guiding design decisions

there might be a case for scaling to up to 3b or something, but the llm capacity doesn't seem to be the limiting factor and id rather improve other parts of the stack before scaling the actual generator

Yep makes total sense. strong foundation will really push this forward and tbh given how its trained for ~6 days it looks like great progress. good shit!

most of the actual training time has gone into testing myriad different RL approaches to see if i can comfortably hill climb a whole suite of metrics first, the actual training for the checkpoint in the vid took ~2.5 days across 3 runs. turns out the hard part is r&d still

have u seen this

WOAH this map feels crazy readable. insane results

just wait it'll get better :D

What is this fruit fly’s name?

RELEASE WHEN?????

omfg!!!! excited about this

Check dm's!
