Best Compression Methods for 'Specific' Games. Q&A

[L33THAK0R]

Originally Posted by KaktoR

For your lz4 sample you have to use lz4f with level 2 and pick the right lz4 library

Compressed 1 file, 32,882,867 => 158,560,939 bytes. Ratio 482.20%
Compression time: cpu 0.08 sec/real 1.18 sec = 7%. Speed 27.85 mB/s
All OK

You have to look which level produces the closest recompressed size to the original size



However in a quick test there are still some streams left behind.

Ah didnt realise I had to specify lz4f, many thanks, also what tool are you using to get that console output if you dont mind me asking?

[shazzla]

Originally Posted by L33THAK0R

Ah didnt realise I had to specify lz4f, many thanks, also what tool are you using to get that console output if you dont mind me asking?

Hi Kaktor and everybody !

What is this tool in the screenshot ? And what parameter was used?
Im out nearly 2 years from repacking!
Btw anyone can share me an lz4 library collection for xtool?
Many thanks!

[KaktoR]

Originally Posted by shazzla

Hi Kaktor and everybody !

What is this tool in the screenshot ? And what parameter was used?
Im out nearly 2 years from repacking!
Btw anyone can share me an lz4 library collection for xtool?
Many thanks!

xtool --verbose mode

Attached file contains some lz4 libraries

[KaktoR]

--verbose mode from xtool

[Edison007]

lzo_sample.rar

The old function LZO1X-1 was used. You can use these libraries. But not all streams are successfully recompressed, because lzo-detect in xtool is't ideal.

[L33THAK0R]

Originally Posted by Edison007

The old function LZO1X-1 was used. You can use these libraries. But not all streams are successfully recompressed, because lzo-detect in xtool is't ideal.

Just ran a quick test, this works wonderfully, thanks so much! I've had this sitting on the back-burner for months, it's very satisfying to have it finally sorted!

[L33THAK0R]

Does anyone know if theres a way to bulk apply/insert a hexadecimal pattern into a given chunk of data if given a start and end offset? I've got some projects I'm working on that require a few hundred blanking/0-filling operations and its awfully slow to do so with HxD's GUI. I've gone through the SFK documentation a half dozen times and alas to no avail. My current solution is just to have a massive blank file and drawn on the given offsets from it but this is obviously not very practical and is a pain having to increase its size should I be attempting to patch out a chunk from a larger archive in the future.

[KaktoR]

Try with xtool's erase/replace command

[dixen]

Pathfinder: Kingmaker (GOG Version) [New repack]

Bundle Folder (excluded manifest files)
XTool v0.65 (unity+lz4 plugin v1.75)+srep+lolz
21 gb > 50 gb > 25 gb > 7.2 gb

Sound files (*.wem & *.bnk)
srep+lzma2
3.2 gb > 2.6 gb

*.assets *.resS & extensionless files
srep+lolz
13.4 gb > 2.8 gb

Rest files
400 mb > 65 mb

Repack's size - 12.7 gb
Install time - 29 min. (R5 3600)

[hwang4]

Which is best compression method ?

[Masquerade]

Originally Posted by hwang4

Which is best compression method ?

Probably zlib used on game PAK file so use xtool+srep+lolz, srep on mp4 videos. Haven't downloaded this game so cannot confirm if zlib is used or not. Check with FModel

[hwang4]

tried this method without success ! it doesn't contains zlibs!

[Masquerade]

Originally Posted by hwang4

tried this method without success ! it doesn't contains zlibs!

... so check with FModel like I said, it will tell you what algo is used.

[hwang4]

Masquerade , checked with fmod , this game uses "oodles"
but need some help , i need to know which version of oo2core.dll , i must use?
I tried (4,5,6,7,8,9) without good ratio!
thx!

[L33THAK0R]

Are there any other LZ4 libraries outside of the ones in the GitHub releases that can yield any half decent results/are there any reliable means to pick up dropped LZ4 streams? In a previous post here I was inquiring on some samples including an LZ4 one, and whilst the level 2 detection processes a number of streams, the total is still fairly low (~65/700) with the final result only being a ~110% inflation. I know not all streams are created equal and some are larger/smaller than others but I'm curious if processing the other dropped streams could result in a more significant result.