Realistic Voice Gender Conversion (VGC) requires independent scaling of the glottal (pitch) and vocal tract (formant) related features of the input speech signal. We present a VGC algorithm which has two novel features. Firstly, an efficient frequency scaling algorithm is presented. Secondly, we use this to scale all frequencies in the input signal by the desired formant scaling factor. We then deconvolve the glottal contribution using a standard linear predictive analysis and frequency scale it further such that the desired pitch scaling factor is equal to the product of the two frequency scaling factors. Finally, we resynthesize the converted speech. The female-to-male results were excellent while the male-to-female results sounded synthetic.