You’re wrong. Partially. Stand by while I illustrate.

File names are not included in content hashes.

Let’s first see how a torrent file is created, using quick-and-dirty bitmaps I just drew. File sizes will be in whole MiB (1 MiB = 1 pixel width) for convenience.

🚧 UNDER CONSTRUCTION 🚧

So you have a file tree you wanna share.

The client will arrange the files in some order. I think you can create a valid torrent file with any ordering but let’s just assume it’s by path, alphabetical. (Sizes are not to scale)

Now file contents are concatenated.

Then split into 14MiB chunks (in practice 16MiB but I forgot to account for border while drawing lol) and zero-padded (grey) to fill the last chunk.

v2 .torrent files can be aligned (where large files always start at chunk border) but this is not allowed/encouraged in most trackers due to hit-and-miss support by clients.

Now a hash is calculated for each chunk, creating a hash map. I’ll assume you know what hashing is.

🚧 FOR INTERNAL USE 🚧 cuz I'm too lazy to transfer the text to my PC using a more reasonable method

Externally, You can also omit .nfo files etc. but the whole chunk will not be seeded because