@JakeParkerLIVE Learning to script is the easy part, if somebody is incapable of that, they're going to have a very difficult time doing something meaningful in this field.
@okowa_T I think a product developed in that manner may be the best sort. Not for everyone. It's more specialized. If you're compromising for everyone, you're not making something great.
Verse will be easier than C++, and likely no more difficult than blueprint. Refusing to code logic is such an odd hang-up. Any scripting modality is just friction you engage with by necessity. It's less effort defining any non-trivial logic in code. A meaningful game will contain something non-trivial.
@mattworkman@GregoryRich19 There's a built in animation modifier that takes in a mirror data table and uses it to mirror the anim sequence that the modifier is being applied to. I can't recall the exact name, I think the plugin containing it is enabled by default.
Slowed down, the sway of the shoulders and twisting through the spine look better in v2. I'd reccomend splitting the difference in the arm swing between both, and slowing v2 down considerably. It looks like it's being played in fast forward at its normal speed, it reads as if the character is wasting energy. Additionally, while the spine rotation is good in v2, the lateral sway of the head should be countered, to an extent by distributing a counter rotation through the spine. We naturally stabilize our heads when moving, it's useful to imagine it as a fulcrum point. You could also try adjusting the foot trajectory so that the character zig zags enough to keep the head over the center line. I can definitely see improvement in the second anim, but it's obscured by the rapid play-rate. Hope this helps!
Regardless of the quality of this initial output, be it good or bad, I'm delighted to find any lab or organization focusing on the enhancement of these models' creative writing skills.
A Focus solely aimed towards the progression of scientific and programming capabilities, both valuable pursuits, overlooks many potential applications of these models.
Storytelling is crucial. When crafting a novel and meaningful tale, one signifies a special sort of understanding. The sort that often reflects substantial truth.
My belief is that for large language models to align with the good and true, both these paths must be pursued.
This is a really neat project, and something I've wanted to see done for quite a while. I do most of my modelling in blender, and haven't really messed around with UE5's built in toolset yet, but I've always found hammer's approach to level design, especially the utilization of hotspot UV mapping, really compelling. I'll have to try this out!
@kronok As someone who's spent far too much time making locomotion implementations, I must say, that is some great looking foot placement / directional handling. Is this a procedural system? That transition into a sidestep was really smooth.
According to Ryan, it occurs when curve compression is set to uniform indexable. Im not sure if there are more factors involved or not. I havent attempted to replicate the issue yet, but I believe Ryan ended up extending his AMS plugin to not require uniform indexable curves for distance matching.
Unless I missed something, the native Steering node does not handle distance matching scenarios. So, I made my own. Among other things, it allows the steering to be driven by an anim curve or directly by a float. The experimental node will be in the next update of my AMS plugin.
This is a very minimal state machine implementation. Much of the coverage present in the motion matched system, such as stops, don't seem to be present in the state machine implementation. Motion matching is great, but this isn't a fair comparison of the technologies by any means. That being said, the guide itself seems useful, and is worth reading.
My newest 3.5-minute devlog, which discusses source control and basic game framework, is now available: https://t.co/at8YxFOc7t. Let me know whether this type of video is something you like.