This is a compressed file, but the actual data offset is unknown. The file only seems to record the length of the filename and the filename itself; the rest is unidentifiable. The internal file appears to be a model format used in the RW engine.fmb.7z

Edited March 28Mar 28 by KuWuniss6

According to online sources, the compression algorithm used for this file is LZO1X. The minilzo library can be used for decompression, but its feasibility cannot be verified.

 

1 hour ago, KuWuniss6 said:

According to online sources, the compression algorithm used for this file is LZO1X. The minilzo library can be used for decompression, but its feasibility cannot be verified.

 

It does seem to be lzo1x from manual tests.  Looks like about 115 files in that fmb archive, but it also seems incomplete and cut off at the end.

Please add the game name and platform to your post title.

11 hours ago, DKDave said:

It does seem to be lzo1x from manual tests.  Looks like about 115 files in that fmb archive, but it also seems incomplete and cut off at the end.
从手动测试来看，它似乎确实为 lzo1x。看起来好像在 fmb 档案中大约有 115 个文件，但似乎也不完整，而且最后部分文件也被截断了。

Please add the game name and platform to your post title.
请在帖子标题中添加游戏名称和平台。

Thanks, I’ll add more details right away.

It appears that the TBL file is compressed using zlib. After decompression, a pure index file can be obtained. Additionally, at the beginning of both the FMB file and the decompressed TBL file, the first few bytes share identical values. The decompressed TBL file records certain information from the FMB, including: the offset of the file data, whether compression is used, the size of the uncompressed data, the size of the compressed data, and an 8-byte unknown field.

On 3/27/2026 at 11:34 PM, DKDave said:

It does seem to be lzo1x from manual tests.  Looks like about 115 files in that fmb archive, but it also seems incomplete and cut off at the end.
手动测试显示确实是 LZO1x。 看起来 FMB 档案里大约有 115 个文件，但看起来也不完整，最后被截断了。

Please add the game name and platform to your post title.
请在帖子标题中添加游戏名称和平台。

Sir, the file has been fully parsed now. I need your help.

First, when we extract the TBL file, we get the correct index. The first 4 bytes of the extracted TBL represent the total file count inside the FMB — that’s 0x73. 0x25CB is the compressed size of the first file (calculated starting from offset 0x1C in the FMB). 01 00 00 00 indicates whether the file is compressed. A4 7D 00 00 is the decompressed size of the first file. A2 25 00 00 points to the compressed data of the first file, excluding the name index and name length. 98 3A 38 01 A8 E3 3F 00 are 8 unknown bytes for the first file. 0D 00 00 00 is the length of the first file’s name. 63 6C 6F 30 31 5F 30 31 2E 64 66 66 00 is the file name itself. That marks the end of the first file entry. We follow the same pattern for the second file and so on.

As previously mentioned, the texture blocks are uncompressed and can be easily extracted; they are in DDS and PNG formats. For convenience, I ignored the texture names and exported the raw texture data directly. As for the models, I used a brute-force method to extract the data, which remains in the LZO1X compressed state.