Running conditional variational auto encoder on non image data

I have been working on conditional variational auto encoders ,and I would like to use it to generate new data samples.

My existing data is in a json file,which has 218 parameters that vary between -1 and 1.

Let’s call this X

Apart from this the conditional data is a numpy array of dimension 33x3.(which is c),these are basically contain pose landmark information obtained from Mediapipe.

Now as per my understanding, the decoder will calculate P(X|z,c) ,and from what I have seen,usually c/label is a one-hot vector

So,now the question is ,how can I use a conditional Variational auto encoder to generate a new X conditioned on c.Also is it possible to use a CVAE for my case,if not then what other generative models can I use?

I know that in input the input to the encoder will be a tensor of dimension 218,but I am unsure of how to use the condition variable .