04-26-2016, 04:24 AM
Hi there, a guy who's been working on a couple Vita-related things over the last few months here. For one, I've written GXTConvert, which is a GXT-to-PNG converter compatible with Cyber Sleuth and other Vita games using GXT textures: https://github.com/xdanieldzd/GXTConvert - In fact, CS (along with SAO: Hollow Fragment and Senran Kagura: Shinovi Versus) was one of the first games whose GXTs I worked with 
Next up, after having seen some posts by Toastline about CS' .geom models a couple of days ago, I went ahead and took a stab at reversing them to render, and possibly export. Here's my preliminary results (in the spoiler, because tall images!):
Your initial findings about the file structure gave me a small headstart, but more research was needed (and, frankly, still is!) to get to texture coordinates, normals, primitive types, texture IDs, etc... Most important for these are the vertex attribute definitions, which are the block of data you had down, I think, as "???B" in your notes. These define what each element of data inside one vertex is used for, like - hypothetical example - "at offset 0x0 inside the vertex are 3 32-bit floats, which are used as the vertex position", or "at offset 0xC are 2 16-bit floats, which are used as the texture coordinates". And yes, there are 16-bit floating point numbers - I had to guess how to read them, failed multiple times, then ended up finding some code on Stack Overflow or somewhere like that that happened to work.
Anyway, my personal TODO list for this:
I should be able to get my viewer and exporter far enough along so that you can easily render and export static models, but I'm unlikely to be able to figure out and export animations and the like; 3D animation has always been my weakness.
So, yeah, hope that helps somehow already?
Next up, after having seen some posts by Toastline about CS' .geom models a couple of days ago, I went ahead and took a stab at reversing them to render, and possibly export. Here's my preliminary results (in the spoiler, because tall images!):
Your initial findings about the file structure gave me a small headstart, but more research was needed (and, frankly, still is!) to get to texture coordinates, normals, primitive types, texture IDs, etc... Most important for these are the vertex attribute definitions, which are the block of data you had down, I think, as "???B" in your notes. These define what each element of data inside one vertex is used for, like - hypothetical example - "at offset 0x0 inside the vertex are 3 32-bit floats, which are used as the vertex position", or "at offset 0xC are 2 16-bit floats, which are used as the texture coordinates". And yes, there are 16-bit floating point numbers - I had to guess how to read them, failed multiple times, then ended up finding some code on Stack Overflow or somewhere like that that happened to work.
![[+]](images/collapse_collapsed.png "[+]")
Spoiler![[Image: 6RYqzsQ.png]](/image.php?src=http%3A%2F%2Fi.imgur.com%2F6RYqzsQ.png&hash=84605a4423cd6f55a561c90c176a1f42)
![[Image: LkLI6nn.png]](/image.php?src=http%3A%2F%2Fi.imgur.com%2FLkLI6nn.png&hash=1d52dac93b7be644e22308c14896ccbe)
Anyway, my personal TODO list for this:
- Fix and improve the Collada exporting code (current code has been ripped from an earlier project of mine and just crammed into this one...)
- Generally figure out more about the format, ex. how the data blocks work that contain ex. texture IDs
- Generally improve the program's code and post a build for people to test
- After the above, eventually put the source for this and Cobalt (3D rendering helper library) up on GitHub
- Pipe dream time! Skeletal animation! Brrrr... <.<
I should be able to get my viewer and exporter far enough along so that you can easily render and export static models, but I'm unlikely to be able to figure out and export animations and the like; 3D animation has always been my weakness.
So, yeah, hope that helps somehow already?