X3M

Any Idea why it got pulled/deleted? seems strange.

tunepunk

People understand, they just aren't going to buy the product. Making C3 a service is alienating a huge chunk of the userbase, and largely seems to be done as a step to get more money from schools.

cjbruce

If Q3D master is not global aswell that might lead to the issue, since you're destroying the scene the objects are linked to, but they can persist in that scene, while another scene is created that they are not a part of.

cjbruce

I'm not sure what you mean, its unclear. Even if it's in the debugger it might not exist in the same scene, but there are too many factors for me to understand whats going wrong. Send me a Capx that demonstrates the problem.

imothep85

No, it's not at all that simple. Qfx shaders aren't a complete feature, and shadertoy doesn't use some kind of standard format anyway.

I think a subscription model just doesn't fit with the use case of a game engine. You work on a game for potentially *years* at a time without releasing it, and need to maintain your ability to work on the project, save old versions of the engine so you have at least some semblance of version control, and have a static work environment that wont change in any way so that you can always make updates, access source files etc.

Being browser based already opens up construct 3 to all kinds of issues with the browser engine misbehaving and potentially screwing up potentially years of work, but to have to pay just for the privilege of working on your project is really not something a serious developer would consider acceptable. The idea of being locked out from your tools is something anyone using Construct 3 will become aware of, even if they aren't right now. Construct 3 isn't enterprise software, yet you're expecting individual persons to throw money at it as if they're businesses who can afford the fixed cost and write it off as an operational fee.

I cant afford to pay the subscription fee. I might potentially use the product once every few months, it's just not justifiable. I think a lot of users who would have gotten Construct 3 as a 'toy' were it another normal piece of software they could install and open from time to time, wont anymore. 100$ to play around every couple of months is too much, and the subscription model makes you reluctant to work on anything that you might wanna get back to at a later date.

And as someone who STILL uses construct classic, I feel kind of burned that Construct 3 is moving further and further from what made Construct good in the first place. Construct 2, and now 3, still lack core usability features that would improve the engine significantly for serious developers such as storing object references, lists , SOL's, Arrays, in object variables, and building prefabs of objects in a general class based system rather than having to rely on the botched container and family features that always wind up requiring weird work arounds when structural changes have to be made. Instead we're encumbered with the overhead of storing UID's in other objects and having to a search over N objects every-time we want to make an object reference, so even building prefabs, constructors/destructors in events carries a severe overhead. The event engine/SOL system is great, it can be used for rapid prototyping and really quick development, but it needs improvements to actually be usable for performant code in a large game. The same issues that were in construct classic are largely unaddressed even in Construct 3. There needs to be more types than just string and numbers, and more control and object oriented paradigms implemented into the core workflow.

Construct 3 seems to want to be professional development software with a fee that's even steeper than Unity's, yet the features that have been added are largely unimportant to anyone using the tool as a professional developer. dictionary/array editors, the "real equation" expression feature that makes things more unreadable, the "work anywhere on any device" (really who will be able to code with any capability on the subway on a tablet/phone without a mouse or keyboard?), the useless cloud integration, all feel like additions that are meant to make the software appealing to the most novice of users, yet the pricing is set so that only serious professional users will pay it. It kind of feels patronizing to your own user base.

Why wasn't more done to actually appeal to advanced users if that's the audience the pricing targets? If that's not the audience you wanted, a subscription model, especially at such a high cost, is really not the right way to go.

tunepunk

You can't direct it using euler angles, however you can set a direction with an XYZ vector which is essentially the same thing. You can use a dummy Q3D point, or the Q3D model you want it to rotate with it, e.g. if you want to specify euler angles set the raycasters direction to x,y,z=Q3Dpoint.nTx,Q3Dpoint.nTy,Q3Dpoint.nTz. and it's position to the object position. Just make sure its on an objects that isn't a child of other objects.

tunepunk

Nice!

3D is pretty taxing for mobile, but the best way to get increased performance is to avoid bone based animations and use morph targets (these run almost entirely in shaders so they can be quite fast), and to cap the resolution to something lower than the native screen resolution (in Q3D Master) since mobile GPUs lack the performance to keep up.

you need to write a shader that samples an area and redraws the sampled area in a flipped view, it's a bit involved but its the best way to do it. By shaders i mean "effects" in construct on a sprite.

It can be done through custom shaders.

tunepunk

It uses version 71.

I'm pretty sure they made some change to the file format, you'll have to use the old blender exporter in the 71 .zip file until I update.

https://github.com/mrdoob/three.js/releases?after=r72

There should be the relevant exporters in "three.js-master\utils\exporters"

Because of the way three.js is maintained, it makes it very difficult to continually update the version because of massive breaking changes each revision introduces that require large amounts of code to be rewritten.

For TowerClimb we just set the fullscreen resolution to native, and scaled the game using a canvas on a top layer that captures the frame and is stretched using the distort map function on a canvas by altering the UVs to the proper size. This means we can get pixel perfect true scaling for whatever small resolution we want (we use 320x240) and not have to mess around much. Obviously this means you render more of the game underneath everything but it doesn't really impact performance. This seems to be the best way to get non-blurry point sampling and not rely on low fullscreen resolutions which are usually blurry because graphics cards today don't support point sampled fullscreen very well. It is a big work-around if u haven't thought of it from the beginning (you might need to do a lot of refactoring).