One key bottleneck of solid-state NMR spectroscopy is that ¹H NMR spectra of organic solids are often very broad due to the presence of a strong network of dipolar couplings. We have recently suggested a new approach to tackle this problem. More specifically, we parametrically mapped errors leading to residual dipolar broadening into a second dimension and removed them in a correlation experiment. In this way pure isotropic proton (PIP) spectra were obtained that contain only isotropic shifts and provide the highest ¹H NMR resolution available today in rigid solids. Here, using a deep-learning method, we extend the PIP approach to a second dimension, and for samples of L-tyrosine hydrochloride and ampicillin we obtain high resolution ¹H-¹H double-quantum/single-quantum dipolar correlation and spin-diffusion spectra with significantly higher resolution than the corresponding spectra at 100 kHz MAS, allowing the identification of previously overlapped isotropic correlation peaks.