Suggestion: 3 Simple Post-Process Effects, which will dramatically increase the quality of engine, at minimum performance cost

Hello, I would leave a suggestion 3 simple post-process effect to implement the engine, but that would dramatically increase the quality of it.

1º Gaussian Blur: This simple effect, connected with a control variable ranging from 0 to 1, could create interesting effects for Cutscenes, and if used in a low value (something like 0.1 or 0:05) dramatically enhances the effect of Antalising at a minimal performance cost.

2º Motion Blur: Another simples effect, which also already present in Misushi, could increase and the quality of the movement of the characters and the camera, all with a minimum cost of performance.

3º Split RGB: This effect is present in Cryengine, Unreal 4 natively, and can be placed in the Unity 5 through plugins. When used sparingly, it dráticamente increases the graphic quality, also at minimal cost processing. For help simulate the real color distortion on an older camera.

AA issues are perfectly pwned by SMAA.
Color filters would be good addition.
p.s. 90%+ people hate motion blur.

About motion blur I agree in part. Exaggerated values really do not like, but a low amount of motion blur and assists in realism, and does not irritate the players.

Hello Rodrigo and thanks for the suggestions.

  1. We already have some kind of Gaussian Blur available in Paradox (from editor: click on scene root node -> add a new render effect to the renderers list and choose Gaussian Blur as effect type). Does it seems enough for your need?

  2. For Motion Blur, it is something that we definitely want in the engine for the mid/long term but it is not in our priority list for the moment as it requires a non-trivial amount of work

  3. What are you exactly meaning by “RGB Split” effect? is it adding some pixels offsets between color channels (like in https://www.youtube.com/watch?v=maxGoDOmpPU) or Color filters?

1 Like

Not exactly, it is a fixed effect and the three color channels are separated as a picture taked by an old can, or an old movie. The extent of this distortion is regulated by a single parameter, how many higher is that value, more the colors as separated.
Following two samples:


Ok thank for the clarification :]

1 Like

I would like to see a motion blur applied only to objects in the scene and not to the camera. Current tech usually calculates a velocity map as a post process and uses it to apply motion blur to a scene. The problem with this is that the velocity map detects changes that the camera makes when moving.

This is why people don’t like motion blur. I’ve been trying to think of a way to apply the motion blur as a post process without detecting the camera movement but it’s not easy. There was a paper from the early 2000’s(I think) that had velocity calculated for each moving object and then rendered trailing meshes behind at different alpha levels to fake a nice motion blur. Something like that might be nice, but it would be slower.

Here in this book is a powerful technique that lets you create a motion blur effect with a good quality and great performance.

http://www.amazon.com/dp/0763778885/?tag=terathon-20

Unfortunately I still do not have technical knowledge to convert the example of HLSL to GLSL, and implement in the engine. :sweat: