This depends on what services we are to provide. The rule of thumb for straight audio is that recording will take three to five times runtime, depending in part upon whether a complete listen-back is required for QA purposes. But when one records to video, the coefficient goes up steeply depending on the service.
For UN-style voice-over, one should expect the recording to run at least eight to ten times runtime. In the case of lip-sync recording, the factor increases to at least 15 and possibly more, depending on how closely one wants the mouth movements and voice to be synchronized. In all cases, one must take into account the difficulty of the material and the quality of the translation. As to the latter, we will only guarantee our estimated recording and talent time if we are engaged to do the translation.