Let $G$ be a finite group of order $g$. A probability distribution $Z$ on $G\ $is called $\varepsilon$-uniform if $\left\vert Z(x)-1/g\right\vert \leq\varepsilon/g$ for each $x\in G$. If $x_{1},x_{2},\dots,x_{m}$ is a list of elements of $G$, then the random cube $Z_{m}:=Cube(x_{1},\dots,x_{m}) $ is the probability distribution where $Z_{m}(y)$ is proportional to the number of ways in which $y$ can be written as a product $x_{1}^{\varepsilon_{1}}x_{2}^{\varepsilon_{2}}\cdots x_{m}^{\varepsilon_{m}}$ with each $\varepsilon _{i}=0$ or $1$. Let $x_{1},\dots,x_{d} $ be a list of generators for $G$ and consider a sequence of cubes $W_{k}:=Cube(x_{k}^{-1},\dots,x_{1}^{-1},x_{1},\dots,x_{k})$ where, for $k>d$, $x_{k}$ is chosen at random from $W_{k-1}$. Then we prove that for each $\delta>0$ there is a constant $K_{\delta}>0$ independent of $G$ such that, with probability at least $1-\delta$, the distribution $W_{m}$ is $1/4$-uniform when $m\geq d+K_{\delta }\lg\left\vert G\right\vert $. This justifies a proposed algorithm of Gene Cooperman for constructing random generators for groups. We also consider modifications of this algorithm which may be more suitable in practice.