WVY-Transition-30m
OSLIM "Object Swap Layered Interpolation Mapping"
A tiny video model trained from scratch to create intentional transformations
these are the generated outputs + trained references for our new mechanism OSLIM demonstrated with a small dataset .. now we just need to train a larger dataset .. upgrade coming this weekend
if this works we will be able to do text/image/video-to-video like @ByteDanceSeed_ Seedance 2.5 but with a much smaller dataset & limited transformations
im intentionally not doing the side missions for lucia to fuck up the relationship cuz im tryna see if they have a side mission to cheat on her w mandy
i just murdered two women at the beach then put one of them in the bushes and i put the other one in the trunk of my car .. now im driving her across the city on autopilot while listening to travis scott .. gta next level .. omg the cops just realized shes in the trunk ! fuck they literally arrested me as i was typing this shii i shoulda been paying attention
If this video doent solve your AI psychosis then your doomed for eternity
AI is over bro .. no model as of today can achieve what this .. it will be years before u can actually build GTA .. this is the best game to ever be created
i been playing it all day havent came across a single bug either .. omg i love this game so much
GTA 6 players have apparently found a way to LEAVE REALITY.
A bizarre in-game item seems to transport your character into an entirely different dimension.
As if Florida wasn't strange enough already.
Writer: Oliver
@amasad https://t.co/RBtuSs3iwg
good luck to him lol .. im pretty sure it gona go just like photocraft though .. that repo has 4k issues and it keeps getting worse
@ingi_erlingsson im using your posts to create an image/video reference model for object swapping .. im going to update you on the results after i finish training it im using millions of videos for the real version .. this is the prototype trained on about 10 videos and 100 images
WVY-Transition-30m
OSLIM "Object Swap Layered Interpolation Mapping"
A tiny video model trained from scratch to create intentional transformations
these are the generated outputs + trained references for our new mechanism OSLIM demonstrated with a small dataset .. now we just need to train a larger dataset .. upgrade coming this weekend
if this works we will be able to do text/image/video-to-video like @ByteDanceSeed_ Seedance 2.5 but with a much smaller dataset & limited transformations
@ingi_erlingsson i been training a video generation model from scratch using your videos + curated data in attempt to recreate your style
im really close to figuring it out i just need more references & pre-training data .. aiming for a 1b parameter model
This video was generated by WVY-Transition-30M