Identification of natural compounds, especially secondary metabolites, has been hampered by the lack of easy to use and accessible reference databases. Nuclear magnetic resonance (NMR) spectroscopy is the most selective technique for identification of unknown metabolites. High quality 1H NMR spectra combined with elemental composition obtained from mass spectrometry (MS) are essential for the identification process. Here we present MetIDB, a reference database of experimental and predicted 1H NMR spectra of 6000 flavonoids. By incorporating the stereochemistry, intra-molecular interactions, and solvent effects into the prediction model, chemical shifts and couplings were predicted with great accuracy. A user-friendly web-based interface for MetIDB has been established providing various interfaces to the data and data-mining possibilities. For each compound, additional information is available comprising compound annotation, 1H NMR spectrum, 2D and 3D structure with correct stereochemistry, monoisotopic mass as well as links to other web resources. The combination of chemical formula and 1H NMR chemical shifts proved to be very efficient in metabolite identification, especially for isobaric compounds. Using this database, the process of flavonoid identification can then be significantly shortened by avoiding repetitive elucidation of already described compounds.