Free carnitine is essential to mitochondrial health by buffering the free acetyl-CoA pool and thereby maintaining energy production. It is also responsible for transporting long-chain fatty acids into the mitochondria for oxidation. Almost all the body’s carnitine is in muscle, and plasma concentrations do not reflect tissue content, but there are as yet no non-invasive techniques to assess muscle total or free carnitine. Here we describe a novel non-invasive postprocessing method, using standard 1 H magnetic resonance spectroscopy data, for quantifying muscle total and free carnitine concentrations, which includes an orientation-visibility and spectral fitting component, and consideration of interfering metabolites. We demonstrate the importance of the orientation correction even within one muscle group (accounting for up to 1.9-fold difference within one muscle group and 2.9-fold difference in signal between muscles), show its good reproducibility (CoV 8-12%), and validate the results with mass spectrometry measurements in muscle biopsy samples. We apply this method in a group of patients with genetic mitochondrial disease, to investigate the relationship between mitochondrial dysfunction and muscle lipid accumulation. As predicted muscle total and free carnitine were lower in patients with disease and correlated with the degree of mitochondrial dysfunction and lipid accumulation. Further, robust spatial correlations of total carnitine and muscle lipid imply heterogeneity in mitochondrial function. Our findings suggest that increasing muscle carnitine stores could ameliorate the metabolic effects of and disorders related to mitochondrial dysfunction. Furthermore, it has not usually been known in supplementation studies whether l-carnitine actually reached the target tissue. We suggest that this novel method has significant potential for informing on physiology and pathophysiology, and as a biomarker in monitoring treatment response, investigative drug discovery, and personalised medicine.