Entry 95

The field's voices and the skill words: ccVoiceRequest, ccWordsPlay and skillVoicePlay

2026-09-25audiotest

Filescrates/piney-audio/src/driver.rscrates/piney-audio/src/lib.rscrates/piney-audio/tests/voice.rscrates/piney-audio/tests/voice_ee_fixture.txtcrates/piney-data/src/sound/voice.rscrates/piney-data/src/sound/mod.rscrates/piney-data/src/sound/inf.rscrates/piney-game/src/mode.rscrates/piney-game/src/main.rscrates/piney-game/src/world.rscrates/piney-game/src/area.rstools/sound_tables.pytools/sound_ee.pydocs/engine/sound.md

The user reported that the skills had no voice lines. The audio driver played the event dialogue's voices, but it marked everything below event -1 as not ported. That meant these were silent:

• the field's own voices, including the party members' greetings on joining and leaving;
• the skill words, which piney-battle's skill::Request.words already asks for.

This port lives in piney-audio, so the combat agent only has to wire the calls.

The field's voices

ccEvVoiceRequest hands every event below -1 to ccVoiceRequest (0x0017eeb0). That is a switch over 20 groups; each has a Japanese and an English table in gcmn.prg, and a voiceFile:

• the Grunties (-2 to -12) use file 0;
• the dogs (-13 to -16) use 5;
• the fountain (-20) uses 1;
• party members in and out (-30) use 2;
• member talk (-31) uses 3;
• presents (-32) use 4;
• Fidchell (-40) uses 19.

tools/sound_tables.py now writes the tables into piney_data::sound::inf (EvVoice::field), and Driver::field_voice_request plays them. Because the field UI's Party menu already sends ccEvVoiceRequest(-30, ...), the greetings as members join and leave now speak.

The skill words

• Queueing. ccWordsPlay (0x0017e290) queues (character, skill) pairs, up to four.
• Playing. skillVoicePlay (0x0017e350) plays the first from the sound task, in a field or dungeon. It takes the character's file (spcVoiceData) and its rows (voiceData: voiceKiteTbl to voiceHerubaTbl), and finds the row by a per-character rule on the skill id and on bit 0 of the skill's type.
• In the port. These are Driver::words_play, skill_voice_play and skill_row. Audio::words_play and set_game_area give the runtime its handle, through Event::SkillWords and Event::GameArea. The town and the fields now report their game.area.
• What is left to wire. The fights raising SkillWords is the combat agent's.

Checked

tools/sound_ee.py voice-fixture runs the game's ccEvVoiceRequest, ccWordsPlay and skillVoicePlay in eemu, with GCMN.PRG loaded for the new parts, and records what reaches SEWORDS.

• The field's voices: 2,226 cases, 0 mismatches. Every row of the 20 groups in both languages, with Parody Mode on and off; seven groups with no case; three frames mixing a group with an event voice or a stop.
• The skill words: 6,274 cases. Each of the 18 characters' 304 ids in English, every seventh in Japanese, the gates (an event running, a caster of neither type, id 304), characters 18 and 19, the queue full, and a word left queued.
  • 5,236 of them match line for line.
  • The other 1,038 are outside their tables. The row rule puts them outside a character's table: ids below the base, or past the end. The game reads the memory next to the table there, and the port plays nothing. A character casts only its own skills, so the rows it really reaches are inside.
• The generated tables. tools/sound_tables.py --check finds inf.rs current.
• The rest. The workspace's tests, clippy, fmt and the docs check pass.