A spectral parameterization method with discretization parameters will meet a tradeoff between the design accuracy and computational effort when designing a variable digital filter. Thus, a continuous spectral parameterization method is presented for obtaining the WLS optimal coefficients of variable fractional delay filters. Separated non-negative weighting function and some characters of matrix and matrix operations are applied to derive a closed-form solution, and Cholesky factorization is adopted to avoid the numerical problem of computing the inverses of a matrix. Its high accuracy and considerably reduced computational complexity of the proposed algorithm are experimentally verified. The given FIR filters can found their applications in high accuracy dynamic time compensation.