DMAV.txt

DMAV:
0-3: "DMAV" magic number
4-7: 1 (version?)
8-11: Persona ID (<Persona_id> in character's voice table, voc_sb_line_sit.xtbl, and voice_dynamic_dialog_events.xtbl, <wwise_id> in audio_personas.xtbl)
12-15: Voiceline ID (<Voiceline_id> in voice_control.xtbl, <wwise_id> in audio_line_tags.xtbl)
16-19: "subtitle offset", offset to subtitle data (from end of dmav header), same as 28-31
20-23: ?, usually 0
24-27: length of subtitle data, including "2" value before start of subtitle headers
28-31: "subtitle offset", offset to subtitle data (from end of dmav header), same as 16-19, not sure why it's in here twice
32-35: length of soundclip in milliseconds

SOME DATA (at offset DMAV header size [36]), subtitle offset bytes long:
Not sure what this is yet, could be lip sync data or something.

SUBTITLES (at offset DMAV header size [36] + subtitle offset):
0-4: if not equal to 2, then this is subtitle data, with two entries total (only seems to happen in one unused brute subtitled soundclip, in voc_brute_mission_media.bnk_pc, so this may not be the correct behavior). if equal to 2, it means 28 subtitle entries FOLLOW this. 0 means no subtitles and the DMAV section is over.

ENTRIES (either 2 or 28 times):
0-3: length of entry
4-7: offset of entry, starting from the first entry (after the subtitle metadata)

SUBTITLE ORDER (list repeats itself once; leftover of dropped dialect support?):
English
Spanish
Italian
Japanese (requires a different lookup table)
German
French
Dutch
Unknown (Always 0 length, 0 offset)
Unknown (Always 0 length, 0 offset)
Czech
Polish
Korean (requires a different lookup table)
Russian
Unknown (Always 0 length, 0 offset)