Calculations of the vibronic structure in electronic spectra of large organic molecules based on density functional methods are presented. The geometries of the excited states are obtained from time-dependent density functional (TDDFT) calculations employing the B3LYP hybrid functional. The vibrational functions and transition dipole moment derivatives are calculated within the harmonic approximation by finite difference of analytical gradients and the transition dipole moment, respectively. Normal mode mixing is taken into account by the Duschinsky transformation. The vibronic structure of strongly dipole-allowed transitions is calculated within the Franck-Condon approximation. Weakly dipole-allowed and dipole-forbidden transitions are treated within the Franck-Condon-Herzberg-Teller and Herzberg-Teller approximation, respectively. The absorption spectra of several organic pi systems (anthracene, pentacene, pyrene, octatetraene, styrene, azulene, phenoxyl) are calculated and compared with experimental data. For dipole-allowed transitions in general a very good agreement between theory and experiment is obtained. This indicates the good quality of the optimized geometries and harmonic force fields. Larger errors are found for the weakly dipole-allowed S-0-->S-1 transition of pyrene which can tentatively be assigned to TDDFT errors for the relative energies of excited states close to the target state. The weak bands of azulene and phenoxyl are very well described within the Franck-Condon approximation which can be explained by the large energy gap (>1.2 eV) to higher-lying excited states leading to small vibronic couplings. Once corrections are made for the errors in the theoretical 0-0 transition energies, the TDDFT approach to calculate vibronic structure seems to outperform both widely used ab initio methods based on configuration interaction singles or complete active space self-consistent field wave functions and semiempirical treatments regarding accuracy, applicability, and computational effort. Together with the parallel computer implementations employed, the present approach appears to be a valuable tool for a quantitative description and detailed understanding of electronic excitation processes in large molecules. (C) 2004 American Institute of Physics.