Document soundbank findings

NOTES.md

@@ -14,6 +14,8 @@
   Packaging: `python tools/build_app.py` -> dist/UnheardCity (self-contained, licenses); packaged exe passes
   `--smoke-test` with PATH = C:\Windows only. Real e2e through the editor code path (tools/e2e_editor.py): PASS
   (music heard in game ncc 1.0, reverts byte-identical, speech line apply/revert byte-identical).
+- NEXT: editing of the ~15,800 in-memory sounds in soundbanks (loose `SFX\*.bnk` footsteps first, then banks
+  embedded in `.upk` packages); not listed or editable yet, survey in `docs/formats/banks.md`.
 - NEXT (optional): verify >2-channel channel order in game; speech heard in-game (needs gameplay).
 - OPEN: channel order for >2 channels unverified.
 
@@ -80,3 +82,7 @@
   User chose no new software (no Inno Setup).
 - 2026-10-09: app/assets is committed (user: public artwork and generated content, not Warner Bros. pictures).
   build_app.py regenerates it only when missing or with --art; after changing make_assets.py run it by hand.
+- 2026-10-09: SFX coverage check (read-only): the catalog covers every WAD sound, but ~15,800 in-memory WEMs in
+  soundbanks are missing (docs/formats/banks.md). Packages were unpacked with Gildor's decompress.exe from
+  `%USERPROFILE%\Desktop\ArkhamCityModding\Audio\decompress\` (outside the repo). User: document now, implement
+  editing later.

README.md

@@ -4,7 +4,9 @@ Replace any voice line, music track or ambience in **Batman: Arkham City (PC, St
 by name, from a file or a recording, with one-click revert. No Wwise, no game patch, no external programs.
 
 - **Unheard City** (`app/`, [user guide](docs/user-guide.md)): the desktop editor for Windows and Linux. Browse and
-  search all ~11,500 sounds, filter by category and character, listen, replace or record, apply and revert.
+  search the ~11,500 sounds of the game's audio archives, filter by category and character, listen, replace or
+  record, apply and revert. Most short effects (footsteps, foley, weapons, UI) live inside soundbanks and are
+  not listed yet ([docs/formats/banks.md](docs/formats/banks.md)).
 - **`ackit`** (`ackit/`): the Python library and command-line tool behind it.
 
 How it works in one paragraph: the game's Wwise 2011.1 decoder only knows the standard Vorbis codebooks, so ackit

REPORT.md

@@ -21,6 +21,9 @@
   Contains a per-user installer (`Install.cmd`): install, upgrade and uninstall (Settings > Apps) verified.
 
 ## What is not verified / limitations
+- **Coverage**: only sounds in the WAD archives are listed and editable. About 15,800 short effects (footsteps,
+  foley, weapons, UI, grunts) are stored inside soundbanks (loose `.bnk` and banks embedded in `.upk`
+  packages) and are not covered yet (`docs/formats/banks.md`).
 - Speech lines being *heard* in the game needs gameplay; verified indirectly (same decoder/WAD path as the music
   test, byte-level apply/revert, boot test with modified speech WAD).
 - Speech/effects replacements must fit the original's size (index lives in compressed `.upk` packages).
@@ -29,5 +32,7 @@
 - Character labels outside the user's lists are inferred from names (marked as such), not verified speakers.
 
 ## Next steps
-1. Extend the curated speaker lists (`data/catalog/source/<lang>/*.txt`) and rebuild with `tools/build_catalog.py`.
+1. Support the in-memory sounds in soundbanks: the loose footstep banks first, then the banks embedded in
-2. Optional: rewrite the BKPK indexes inside `.upk` packages (LZO) to lift the size limit for speech.
+   `.upk` packages (`docs/formats/banks.md`).
+2. Extend the curated speaker lists (`data/catalog/source/<lang>/*.txt`) and rebuild with `tools/build_catalog.py`.
+3. Optional: rewrite the BKPK indexes inside `.upk` packages (LZO) to lift the size limit for speech.

ackit/allocsize.py

@@ -1,6 +1,6 @@
 """Model of the game's codec-setup bump-allocator usage (dwDecodeAllocSize).
 
-Reversed from the dumped BatmanAC.exe (see docs/decoder_re.md):
+Reversed from the dumped BatmanAC.exe (see docs/reverse-engineering/decoder.md):
 WwiseVorbisParseSetupHeader (runtime 0x010B4D10) and its callees are a Tremor-lowmem
 derivative; every pool allocation is rounded up to 4 bytes.
 """

ackit/builder.py

@@ -1,7 +1,7 @@
 """Build an Arkham City compatible WEM from a standard Ogg Vorbis stream.
 
-Field rules were derived from all 2467 WEMs of the ITA WAD and are validated by
+Field rules were derived from the WEMs of the ITA WAD and are validated by
-rebuilding original files byte-for-byte (scripts/validate_builder.py).
+rebuilding original files byte-for-byte (`ackit verify`, tests/test_realgame.py).
 """
 import struct
 

ackit/catalog.py

@@ -1,4 +1,5 @@
-"""Index of every sound in the game, with names, categories and characters.
+"""Index of every sound in the game's WAD archives, with names, categories and characters
+(sounds stored inside soundbanks are not indexed yet: docs/formats/banks.md).
 
 Sources, in order:
   1. data/catalog/<ARCHIVE>.csv  -- shipped catalog (instant) when the WAD fingerprint matches;

ackit/codebooks.py

@@ -1,7 +1,7 @@
 """Wwise packed codebook library (ww2ogg 'packed_codebooks.bin' format).
 
 The game's built-in table (runtime 0x0158E748, 598 entries) is byte-identical to
-ww2ogg's standard packed_codebooks.bin; see docs/wem_format.md.
+ww2ogg's standard packed_codebooks.bin; see docs/formats/wem.md.
 """
 import os
 import struct

ackit/pack.py

@@ -9,7 +9,7 @@ BKPK header (u32 LE):
   0x2C end of table (0x48 + 12N)   0x38 embedded bank size   0x3C embedded bank offset
   table: N x {u32 media id, u32 size, u32 offset-in-WAD-entry}, sorted by id; each media
   starts at the next 0x800-aligned offset after the previous one.
-See docs/wad_format.md.
+See docs/formats/wad.md.
 """
 import struct
 

ackit/verify/debugger.py

@@ -31,7 +31,7 @@ def find_cdb():
             return p
     raise FileNotFoundError("x86 cdb.exe not found (install WinDbg or set ACKIT_CDB)")
 
-# Runtime addresses (image base 0x400000; ASLR off for this exe). See docs/decoder_re.md.
+# Runtime addresses (image base 0x400000; ASLR off for this exe). See docs/reverse-engineering/decoder.md.
 ADDR = dict(
     get_or_create_setup=0x010B3400,   # __thiscall(cache, AkSrcInfo*, ...) ; info+0x9c hash
     parse_setup=0x010B4D10,           # WwiseVorbisParseSetupHeader(info, channels, reader, pool) -> 0 ok

ackit/wad.py

@@ -11,7 +11,7 @@ Layout (little endian):
   0x1C  N x {u32 hash, u32 size, u32 offset}   sorted by hash as *signed* int32
 Entries are stored in table order, each at a 0x1000-aligned offset; 'size' is the
 entry payload size padded to 0x800. Padding bytes are zero; the file ends 0x1000-aligned.
-See docs/wad_format.md.
+See docs/formats/wad.md.
 """
 import struct
 from dataclasses import dataclass

ackit/wwvorbis.py

@@ -2,7 +2,7 @@
 
 Wwise stores a stripped setup header (codebooks referenced by 10-bit IDs into the
 built-in library, narrowed fields) and "modified" audio packets (no packet-type bit,
-no prev/next window flags). See docs/wem_format.md.
+no prev/next window flags). See docs/formats/wem.md.
 """
 from .bits import BitReader, BitWriter, ilog
 from .codebooks import CodebookLibrary, wwise_book_to_vorbis, skip_vorbis_book, book_bits_key

data/catalog/README.md

@@ -2,7 +2,7 @@
 
 | file | content |
 |------|---------|
-| `<ARCHIVE>.csv` | every sound of one audio archive: `ITA`, `INT`, `FRA`, `DEU`, `ESN` (speech, one per language) and `SFX` (music, ambience, effects) |
+| `<ARCHIVE>.csv` | every sound of one audio archive: `ITA`, `INT`, `FRA`, `DEU`, `ESN` (speech, one per language) and `SFX` (music, ambience, effects of the SFX WAD; sounds stored inside soundbanks are not listed, see `docs/formats/banks.md`) |
 | `characters.csv` | `name,character`: verified speaker per sound name (language independent) |
 | `index.json` | fingerprint of the game WAD each CSV was built from (`ackit/game.py: wad_fingerprint`) and counts |
 | `source/<lang>/*.txt` | **the curated original lists** (one file per character, one wemscan-style file name per line, e.g. `Speech_TAPE_JOKER_1_4.wem`). `Files.txt` = all files, `Stats.txt` = counts, `Checkpoint.txt` = progress marker. Never edited by tools. |

docs/README.md

@@ -7,6 +7,7 @@
 | [categories.md](categories.md) | group/category scheme and character fields |
 | [building.md](building.md) | building the editor and its installer (Windows, Linux) |
 | [formats/wad.md](formats/wad.md) | WWAD archives, grouped entries, BKPK stream packs |
+| [formats/banks.md](formats/banks.md) | soundbanks and the in-memory sounds not covered yet (footsteps, foley, effects) |
 | [formats/wem.md](formats/wem.md) | WEM layout the game accepts, setup stripping, hash and decode-alloc formulas |
 | [reverse-engineering/decoder.md](reverse-engineering/decoder.md) | BatmanAC.exe decoder functions and addresses |
 | [reverse-engineering/findings.md](reverse-engineering/findings.md) | root cause, findings, in-game experiment log (incl. failed hypotheses) |

docs/formats/banks.md

@@ -0,0 +1,65 @@
+# Soundbanks and in-memory sounds
+
+Most short sound effects (footsteps, foley, combat, weapons, objects, UI, grunts) are **not in the WADs**: they
+are WEMs stored inside Wwise soundbanks, either loose in `SFX\*.bnk` or embedded in the UE3 packages (`.upk`).
+The catalog and the editor only cover the WADs (`wad.md`), so none of these sounds are listed or editable yet.
+Survey of the installed game (buildid 3718300) on 2026-10-09, read-only.
+
+## Inventory
+
+| where | banks | unique in-memory WEMs | in the catalog |
+|-------|-------|-----------------------|----------------|
+| loose `SFX\*.bnk` (Batman/Catwoman footsteps, 32 surfaces each, + `TST_Prep_Media`) | 65 | 1,117 | 0 |
+| banks embedded in `.upk` packages (in 1,432 of the 2,650 packages) | 3,340 | 15,641 (~276 MB) | 0 |
+
+The two sets share 119 media IDs: about **15,800 unique sounds** in total. The WADs, by comparison, are fully
+covered: the BKPK tables in the packages reference 2,543 WAD entries, all of them in the catalog.
+
+Embedded WEMs by bank name prefix: `Speech` 7,872, `CHR` (character foley, combat, bosses) 3,896, `PSO` 781,
+`WPN` 664, `NPC` 475, `OBJ` 439, `SEQ` 265, `GUI` 208, `PSA` 203, `MTN` 196, `PX` 195, `DVC` 100, `GEN` 80,
+`PHY` 73, `AMB` 65, `SBJ` 50, others < 30. The `Speech_` banks (e.g. `Speech_EMOTE_Batman`, `..._GRUNT`) are
+probably efforts/grunts rather than dialogue (there are no per-language packages); not verified by listening.
+All WEMs seen are Vorbis (format tag 0xFFFF), the same codec as the WADs: the footsteps are mono (4 stereo),
+the embedded ones mono or stereo (~22 % stereo in a 2,533-WEM sample).
+
+## Names
+- The community hash list names none of these media IDs; only the bank names are known.
+- `CookedPCConsole\CookedBanks.txt` lists all 3,345 cooked bank names. Bank ID = FNV-1 32 of the lowercased
+  name (the same hash Wwise uses for every name). 3,338 of them are embedded in packages, 4 are loose `.bnk`
+  files (`Init`, `Global`, `Music`, `Music_PDLC`); `AMB_OW_A6`, `AMB_OW_C1`, `AMB_OW_E1`,
+  `Speech_EMOTE_Thug_Muffled_Stop` and `Speech_GFxUI_ArkhamCityStories` were found nowhere.
+- Loose footstep banks are named in their STID chunk (`CHR_BatmanFootsteps_Concrete_Media`, ...).
+
+## Bank layout (version 56, Wwise 2011.1)
+Chunks `BKHD` {u32 version 56, u32 bank id, ...}, `DIDX` N x {u32 media id, u32 offset, u32 size} (offsets
+relative to the `DATA` payload, WEMs 16-byte aligned), `DATA`, `HIRC`, `STID` {u32 type, u32 count,
+count x {u32 bank id, u8 len, name}}. `Init` (`50C63A23.bnk`) has `STMG`/`ENVS` instead of media.
+`701C1363.bnk` holds only `BKHD`.
+
+HIRC Sound object (type 2) payload: u32 object id, u32 plugin id (`0x00040001` = Vorbis), u32 stream type
+(0 = in memory, 1 = streamed, 2 = prefetch), u32 media id, u32 bank id of the media; for in-memory sources
+then u32 offset, u32 size, where offset = `DATA` payload position (from the start of the bank) + DIDX offset
+and size = DIDX size (checked on every in-memory source of the footstep banks and of `BmGame.upk`).
+A few sources use media of another bank. Every streamed media ID found is in a BKPK table, i.e. a WAD sound
+(`wad.md`).
+
+## Banks inside packages
+- The packages are fully LZO-compressed (package version 805, licensee 101). For the survey they were unpacked
+  with Gildor's `decompress.exe` (gildor.org; used from the user's old modding folder, not part of this
+  repository) and searched for `BKHD` (banks) and `BKPK` (WAD indexes). `UnrealX.upk` (512 bytes) is not a
+  readable package (wrong tag, probably encrypted).
+- Bank blobs in packages: 41,798 for 3,340 bank IDs. Many banks are copied into several packages (e.g.
+  `CHR_CatwomanFoleySuit` 31 times, `CHR_BatmanFoleyBelt` 26). 302 bank IDs occur as several blobs with
+  *different* media subsets (often one WEM each, several per package), so a bank's media set is the union
+  of all its blobs.
+- Largest: `CHR_RobinFoleyCape` 300 WEMs, `CHR_BatmanFoleyCape` 282, `CHR_Clayface` 263,
+  `CHR_NightwingFoley` 242, `CHR_CatwomanFoleySuit` 225, `CHR_MrFreeze` 176, `CHR_Combat` 166.
+
+## Consequences for editing (not implemented)
+- **Loose banks**: a replacement means rebuilding `DATA`/`DIDX` and the in-memory offsets/sizes in the HIRC
+  Sound objects of that bank (and in any other bank whose sources use its media). Whether the game accepts a
+  bank that changed size is untested.
+- **Embedded banks**: the package must be rewritten (decompress, patch every blob of the bank, recompress
+  LZO, fix the package's chunk table and export offsets/sizes). A sound must be replaced in **every** package
+  that carries a copy, otherwise the game may load the original from another package.
+- Either way: unverified in game so far; the same WEM rules apply (`wem.md`).

docs/formats/wad.md

@@ -63,8 +63,7 @@ original bytes where nothing moved).
 Replacing a pack media (`ackit` target `EDF036D6/393BC53A`) rewrites the pack table (`.bnk`), the
 WAD entry payload, and the WAD table/offsets. Verified in-game (NOTES.md, E2).
 
-## Regular soundbanks (`SFX\*.bnk`, bank version 56)
+## Regular soundbanks
 
-BKHD / DIDX {id, offset, size} / DATA / HIRC / STID. In-memory WEMs (DATA) are 16-byte aligned.
+The other `SFX\*.bnk` files and the banks embedded in `.upk` packages hold their WEMs in memory (not in a
-`ackit/bnk.py` can read/rebuild them, but most front-end sounds come from banks embedded in UE3
+WAD); see `banks.md`. The toolkit does not read or modify them yet.
-packages (`.upk`), which this toolkit does not modify.

docs/reverse-engineering/findings.md

@@ -68,6 +68,10 @@ Conclusion: approach A works end-to-end in-game. Use >= 30 s probes for the intr
     6 channels work at <= 32 kHz (uncoupled, like Wwise). The encoder falls back to 32 kHz automatically.
 15. Channel order for > 2 channels: the bitstream order equals the array order given to libsndfile; whether the game
     expects WAVEFORMATEX order (FL FR FC LFE BL BR) is unverified (identical-channel probes were used).
+16. (2026-10-09) The WADs hold only streamed audio. About 15,800 more sounds (footsteps, foley, combat, weapons,
+    UI, grunts) are in-memory WEMs inside soundbanks: 65 loose `SFX\*.bnk` (1,117) and 3,340 banks embedded in
+    the `.upk` packages (15,641; many banks copied into several packages). None are in the catalog; the WAD
+    side is complete. Inventory, layout and editing consequences: `docs/formats/banks.md`.
 
 ## Old path -> new path
 `scripts/harness.py` -> `ackit verify --ingame` (ackit/verify/ingame.py); `scripts/validate_builder.py` ->

docs/user-guide.md

@@ -26,6 +26,8 @@ back with one click.
   sound's name (useful hints for the many unlabelled lines); tick *Include characters mentioned in names*
   to filter on them too.
 - Click a column header to sort (length, size, status, ...). Ctrl/Shift-click selects several rows.
+- Most short effects (footsteps, foley, weapons, menu sounds) are stored inside soundbanks and are not listed
+  yet; the list covers voice lines, music, ambience and the effects of the SFX archive.
 - **Language** (top left) switches the speech language. Speech is read from the language the game is set to.
 
 ![Browsing by category](img/browse.png)