Entry 227

Mutation's field voices, food and skill words

2026-09-27volumes

Filescrates/piney-gen/src/manifest.rscrates/piney-data/src/tables/voice.rscrates/piney-audio/src/driver.rscrates/piney-audio/src/lib.rscrates/piney-data/src/sound/voice.rscrates/piney-game/src/mode.rstools/sound_tables.pycrates/piney-audio/tests/voice.rs

Reported from play: on Mutation, the party members' lines were silent in game (Black Rose had text but no voice). Worklog 226 fixed the event voices. The field voices, the food's and the skill words still came from Infection's tables on Infection only, which tools/sound_tables.py wrote by name. They are now piney-gen's voice group for every volume, in Japanese and English, and the Python generator no longer writes any voice table.

What piney-gen reads

• field: ccVoiceRequest's cases.
  • The dispatch is a run of li r, group / beq $a0, r.
  • In each case, the lb of saveData.voice (+0x842c) and the bnez after it mark where the English branch starts.
  • Each lui/addiu pair builds a table. Its file is the first voiceFile store after it.
  • A case has one Japanese table and one English table, which set the same file.
• food: ccVoicePgFood's table, read the same way. It is the same on every disc, so it is written once (voice::FOOD).
• skill: skillVoicePlay's tables by charTbl row, up to spcVoiceData's NULL. The game gives up on a character with no file before it looks at the rows.
  • Mutation's voiceData has three more rows than files, which the game cannot reach.
  • A skill table ends with its (0, 0) row, which is kept.
• files, files_e. These run to a NULL or the next symbol. Mutation's evVoiceFile runs straight into evVoiceFileE, which gains MIAE.BIN (20).
  • The names are written as paths on the disc (VOICE_E/PARTY_E.BIN), not as the IOP's cdrom0:\..., so the port no longer strips them at run time.
• Row bounds. A table runs to the next table built in these functions, or to the next symbol if that comes first. A row is a line (a whole-sector offset and a positive size) or none (-1, -1).

On Infection, every field, food and skill table comes out the same as the old tables, row for row (checked by a comparison test before the old tables were removed).

Mutation's differences

• Mia's English lines. The party, talk and present groups call a new getter of talkNum (MUT 0x0017abf0, 18-20 in the extension) with 1. While talkNum[1] is set, English messages 3-5 (party), 2-3 (talk) and 5-9 (presents) play rows 0-2, 3-4 and 5-9 of one 10-row table from MIAE.BIN.
  • piney-gen reads this as VoiceAlt: the two slti bounds, the getter's index, the addend to the message (-3, +1, 0), the file and the rows.
  • Event::VoiceOptions now carries talkNum for rows 0-20 from the save (mode::voice_options). The field menus' requests send it fresh each time.
• Table sizes. The Japanese party, talk and present tables grew to 63, 42 and 105 rows. The English ones kept 54, 36 and 90. An English message past its table reads the next table's rows in the game; the port plays nothing, as before.
• Loading. ccVoiceRequest and ccWordsPlay return at once while ccSnd +0x139 is set, from ccFileListLoad's first file to its last. This is not ported, because the port loads between frames.
• Skill words. skillVoicePlay and its row rule (the jump table's cases and bases) are Infection's.

BOSSTALK.BIN (Fidchell, -40) is on neither Infection's disc nor Mutation's.

Checks

• Every field, alternative, food and skill line on Infection's disc and on Mutation's lies inside its file, in both languages (every_field_line_is_in_its_file, every_mutation_field_line_is_in_its_file).
• Mia's English party, talk and present lines come from MIAE.BIN only while talkNum[1] is set, and never in Japanese (mutation_mia_speaks_from_miae_once_talked_to).
• The eemu voice fixture (every Infection row and group, both languages, the skill words) still matches, with its file names read as disc paths.