I was expecting the conversion to use rounding, but instead, it round down.
I encountered this issue across modules for Quest VD, Pico, and VIVE Facial Tracker.
VRCFT version: I tested in 5.1.1.0 and 5.2.3.0

For 1-bit quantized parameters, such as converting:

JawOpen [float] to JawOpen1 [bool]

I expected values below 0.5 to convert to false, and values 0.5 and above to convert to true, like this:

• JawOpen = 0 → JawOpen1 = false
• JawOpen = 0.6 → JawOpen1 = true
• JawOpen = 1 → JawOpen1 = true

However, in reality, JawOpen1 only turns true if JawOpen is exactly 1 or higher:

• JawOpen = 0 → JawOpen1 = false
• JawOpen = 0.6 → JawOpen1 = false
• JawOpen = 1 → JawOpen1 = true

The same issue applies to higher bit quantization, but it’s most noticeable with 1-bit, where it can result in up to a 99% discrepancy.

This causes extreme cases where a value of 0.99 is still considered false.

For example, on the Pico 4 Pro, the TongueOut parameter only goes up to 0.99, meaning TongueOut1 from VRCFT is always false.