A PC app that captures live gameplay video and audio, runs a real-time AI model to understand in-game events, then generates and streams advanced DualSense haptic waveforms to the controller so unsupported games can feel like they have native PS5-style haptics
Hyaa!
This is the first devlog for the project I’m starting.
The goal is simple: make DualSense haptic feedback work in racing games
that do not support it.
A lot of racing games already give you useful information through the
screen and sound. You can see the car slide, hit curbs, lose grip, brake
hard, or drive over different surfaces. You can also hear engine load,
tire slip, bumps, and collisions.
I want to use that information to predict what the controller should
feel like in real time.
The plan is to train a multimodal model on racing games that already
support DualSense features. The model will learn from:
Then, I want to run that model on games without native DualSense support
and generate haptics live while you play.
The first version will probably focus on regular rumble. After that, I
want to scale it toward more advanced feedback, including adaptive
triggers and haptic patterns.
This is still early, so the first step is building a recorder. I’ll keep
you updated as I make progress.
Have an amazing day :)
