Just because I get confused/forget every.single.time.
Once and forever I would like to commemorate the relation between two parametrizations of the Negative Binomial distribution: the one, using size=n and prob = p (and used in R as the default version), against the one where explicitly known are the mean $\mu$ and and the overdispersion parameter $k$: \begin{equation} NB(y \mid k, \mu) = \binom{y + k -1}{y} (\mu+k)^y \left(\frac{k}{\mu+k}\right)^{k}, \end{equation}
The relationship is as follows: \begin{equation} n = \frac{\mu p}{1-p}, \quad \mu + \frac{\mu^2}{k} = \frac{n(1-p)}{p^2}. \end{equation}