Your expectations

[kimkulling]

What are your expectations for the Open Asset-Importer-Lib?

Feel free to brainstorm!

[kcrif]

Improved performance for Asset::FindUniqueID, see #3444, Should be a simple fix

[kimkulling]

Good point!

[kimkulling]

What the ... . This looks terribly slow.

[johnb003]

Pick formats that matter most, USD, FBX, glb/gltf mostly, and extend core functionality for these formats with a matrix of feature falloff on other formats, but don't let the other formats limit productivity where it matters most.

The problem is that when the library is first considered for a project it's great that it has a wide range of supported formats, and it seems like people have gotten it working, but then you quickly run into limitations, and fixing this library, if the key value is wide support, feels daunting because so many of the formats aren't that important.

There are a number of basic issues for me that are pushing me to consider direct SDK alternatives for the formats I care most about.

• I don't like that the skeleton is not identifiable in the scene without skinned meshes.
• I saw instances when preserve pivots is used, it inserted extra transforms, but they weren't added to the scene, so you're left with trying to parse the strings to understand where in the scene those channels are related.
• Lack of tangents limits usefulness of curves that aren't baked.
• Pre/post infinity options aren't as complete as in the dcc tools (for example cycle w/ offset, I think?)
• Attribute animation is a big one -- I need general float anims, and the lib assumes all the animation is just for transforms.
• blendshapes would be nice to load, and I'm talking about sections of mesh being sculpted in static poses, where you'd drive those activations with animated weights (not vertex animation).

In general the way the skeleton is determined by gathering what are joints out of skin clusters or guessing based on what's animated, and then crawling the hierarchy, feels like a lot of unnecessary processing that makes the library less easy to use.

I guess my expectation is:
Don't assume the project maturity is good enough to just polish bugs and performance and forget about features. There's still a lot improvements needed for the main architectural design of it and key features that will make or break the success in the long run.