We prove well-posedness for doubly nonlinear parabolic stochastic partial differential equations of the form $dX_t-\text{div}\,γ(\nabla X_t)\,dt+β(X_t)\,dt\ni B(t,X_t)\,dW_t$, where $γ$ and $β$ are the two nonlinearities, assumed to be multivalued maximal monotone operators everywhere defined on $\mathbb{R}^d$ and $\mathbb{R}$ respectively, and $W$ is a cylindrical Wiener process. Using variational techniques, suitable uniform estimates (both pathwise and in expectation) and some compactness results, well-posedness is proved under the classical Leray-Lions conditions on $γ$ and with no restrictive smoothness or growth assumptions on $β$. The operator $B$ is assumed to be Hilbert-Schmidt and to satisfy some classical Lipschitz conditions in the second variable.