Internal or external effects?

[Reactor]

This time, it's not about asking help, it's more like asking an opinion, as it is a question of quality. I shall be using GZDoom for my project, and some of GZDoom's effects are quite enthralling. Majorly, I'm talking about dynamic lights and scaling up/down possibilities, but the topic may as well relate to any image-effects. My question is simple as it is - which is better? Applying glow effect externally (using Paint.NET for instance), and manually resizing the images, or using internal dynamic light and size scaling inside GZDoom? Or perhaps both? Which one gives better quality, which method you'd recommend?
We're talking about unpaletted photorealistic PNGs here of course.

[ramon.dexter]

Hi, personally I like the scaling method. You can have more details in a smaller sprite/voxel. Also, I like the gldefs method of adding lights and stuff, because adding glow in any kind of paint is little bit glitchy. Also, the gldef'ed lights are good to work with - you can have pulsing light, you can have flickering lights... Yes, I know you can make pusling light with a sprite frames, also you can make kind-of flickering effect with the frames. But espeacially the flickering effect is hard to make, since it always cycles the available frames and therfore there always will be a pattern.

[kodi]

Or perhaps both? Which one gives better quality, which method you'd recommend?

Short answer: "it depends."

Seriously though: dynamic lights are fairly demanding en masse so you wouldn't want to light a whole level with them. But exporting the entire level as a 3d model and baking raytraced light into it's textures with a 3D modeling/rendering program of your choice will ramp up the file size to pretty fantastic amounts for a large and/or complex map with high res textures, so you have to choose a middle ground and utilize the various ways of texturing and lighting a level on a case-by-case basis.
Smaller images (in image data size) will of course be less demanding than a larger graphic scaled down in-engine (which doesn't remove pixel data, it increases the pixel density). And then there's the question of weighing performance vs quality, which also depends on what kind of level you're making.