RESO container · Tongari Boushi Archive

Used by both .ofs (raw) and .lzs (post-LZ11) files. Reference parser: src/archive/reso.py.

Header (16 bytes)

Offset	Size	Type	Field	Value
`0x00`	4	`char[4]`	`magic`	`"RESO"` = `52 45 53 4F`
`0x04`	4	`u32 LE`	`file_count`	Number of TOC entries
`0x08`	4	`char[4]`	`version`	Typically `"1.21"` = `31 2E 32 31`
`0x0C`	4	`u32 LE`	`variant`	`0` (type-tagged, 16-B stride) or `1` (simple, 8-B stride)

TOC entry — variant 0 (16 bytes)

Used by configuration-style containers like global.lzs, title_pack.ofs, sprite atlases.

Offset	Size	Type	Field
`+0x00`	4	`u32 LE`	`unk` (often 0; in `tad` entries holds metadata)
`+0x04`	4	`char[4]`	`type_tag` (reversed ASCII — see below)
`+0x08`	4	`u32 LE`	`length` (payload bytes; 0 = empty slot)
`+0x0C`	4	`u32 LE`	`data_offset` (absolute offset within the RESO file)

TOC entry — variant 1 (8 bytes)

Used by message files (data/message/**/*.ofs), talkstart/*.ofs, and most plain data containers.

Offset	Size	Type	Field
`+0x00`	4	`u32 LE`	`length`
`+0x04`	4	`u32 LE`	`data_offset`

Type tag values (variant 0)

The 4-byte type_tag is reversed ASCII. Reverse to recover the Nitro / Konami magic identifier.

Stored bytes	Reversed	Magic	Description	Count in corpus
`rgcn`	`NCGR`	`RGCN`	Tile graphics	4,337
`rlcn`	`NCLR`	`RLCN`	Palette	2,600
`recn`	`NCER`	`RECN`	Cell / OAM definition	1,553
`rcsn`	`NSCR`	`RCSN`	BG screen / tilemap	1,128
`rnan`	`NANR`	`RNAN`	Animation	225
`rtfn`	`NFTR`	`RTFN`	Bitmap font	1
`ncbr`	`NCBR`	`RBCN`	Linear/bitmap NCGR variant	1
`szl`	`lzs`	—	Nested LZ11-compressed sub-RESO	34,464
`sfo`	`ofs`	—	Nested RESO (uncompressed)	333
`tad`	`dat`	—	Konami widget table (see Konami extensions)	694
`nib`	`bin`	—	Per-container binary blob (variable schema)	251
`tbsn`	`nsbt`	`BTX0`	3D texture pack (NSBTX)	8,773
`mbsn`	`nsbm`	`BMD0`	3D model (NSBMD)	297
`cbsn`	`nsbc`	`BCA0`	3D animation (NSBCA)	35
`ymd`	`dmy`	—	Dummy / placeholder script blob	15,712

Payload region

Payloads start at 0x10 + file_count × stride (no special separator).
Re-emit aligns each payload to a 16-byte boundary (pad = (16 − len % 16) % 16; pad bytes are 0x00).
Empty TOC slots have length=0 and data_offset=0.
Multiple TOC entries may share one (length, data_offset) — see shared payloads.

Annotated hex dump — header + first 4 TOC entries

data/message/msg00/00/msg00000.ofs (variant 1, 100 entries):

offs  bytes                                              meaning
----  -------------------------------------------------- --------------------------------
0000  52 45 53 4F                                        magic = "RESO"
0004  64 00 00 00                                        file_count = 100
0008  31 2E 32 31                                        version = "1.21"
000C  01 00 00 00                                        variant = 1 (8-B stride TOC)

0010  08 00 00 00  30 03 00 00                           entry[0]: len=8, off=0x0330
0018  40 01 00 00  38 03 00 00                           entry[1]: len=320, off=0x0338
0020  EA 00 00 00  78 04 00 00                           entry[2]: len=234, off=0x0478
0028  20 01 00 00  62 05 00 00                           entry[3]: len=288, off=0x0562
... TOC continues for 96 more entries (100*8 = 800 bytes)
0330  ...                                                 payload region begins

Variant-0 example: title_pack.ofs

0000  52 45 53 4F  05 00 00 00  31 2E 32 31  00 00 00 00   header, file_count=5, variant=0
0010  00 00 00 00  6C 7A 73 00  00 F4 00 00  60 00 00 00   entry[0]: unk=0, tag="szl",
                                                            len=0xF400 (62464), off=0x60
0020  00 00 00 00  6C 7A 73 00  C0 5E 01 00  60 F4 00 00   entry[1]: tag="szl", len=89792
0030  ... 3 more 16-B entries ...
0060  11 10 27 00  ... LZ11 stream of inner RESO ...       payload[0] starts at 0x60

Decoder reference

# src/archive/reso.py
def parse(data: bytes) -> list[ResoEntry]:
    h = parse_header(data)
    stride = 0x10 if h.variant == 0 else 0x08
    entries = []
    for i in range(h.file_count):
        off = 0x10 + i * stride
        if h.variant == 0:
            unk          = u32(data, off + 0)
            type_tag     = data[off+4:off+8].decode("ascii")  # reversed
            length       = u32(data, off + 8)
            data_offset  = u32(data, off + 12)
        else:
            length       = u32(data, off + 0)
            data_offset  = u32(data, off + 4)
            type_tag     = ""
        entries.append(ResoEntry(i, type_tag, length, data_offset, unk))
    return entries

Cross-references

LZ11 — outer compression wrapper for .lzs files
Nitro graphics — formats for rgcn/rlcn/recn/rcsn/rnan
Konami extensions — tad/nib/sfo