The player object now has arms that the grapples visually attach to. They smoothly move from the rest position to the correct position when grappling and rotate towards the correct point.

I've had to decide that the player can't turn around while it is grappling, because I'm separating the grapples by an offset along the z axis and if the player turns, I'd have to move the positions that the grapples were attached to, so as is you can only turn around when you don't have either grapple extended. Will see how much of a problem this is when a proper model is in place.

The arm system is set up so it provides a quaternion for each arm in model-local space which is then translated with the world matrix when the model is rendered, so I'm hoping once I have a rigged, skeletal animated model as the main character, it should be easy enough to apply these to the model's arm bones rather than the separate meshes I'm using for the arms and body here.

The arms will point in the correct direction regardless of what world matrix is applied to the player. E.g. you can rotate the player model around the Z axis and the arms still line up correctly.

The grapples are still actually attached to the offset where the shoulders are in terms of the physics, then once the arm is positioned I obtain the world point of each hand to pass to the grapple-rendering code to make it look like it is attached to the hand. One reason for this is that if the grapple starts being fired out from the hand position, it can tunnel through walls unless I made the player physics shape large enough to cover the extended arms too, which feels too big. Thankfully the difference in positions is small enough you can't really notice so this seems to work okay.

Not sure what's next. Perhaps getting a simple model made and rigged so I can replace the temporary player object with a proper mesh and start looking at blending the computed arm positions with some standard skeletal animations.

Thanks for reading.