A method for film mode detection is disclosed. The method generally includes the steps of (A) detecting if a plurality of fields in a video signal are in a 3:2 pull-down pattern to control a plurality of first flags based on a plurality of statistics gathered from the fields, (B) detecting if the fields contain moving interlaced text to control a second flag based on both (i) the statistics and (ii) a repeat-field flag of the first flags that indicates repeating consecutive same polarity fields and (C) deciding among a plurality of inverse telecine processes to de-interlace the fields based on all of (i) a 3:2 mode flag of the first flags, (ii) a 3:2 direction flag of the first flags and (iii) the second flag.