Make three.js bloom effect similar to Babylon.js

[cx20]

I added a bloom effect to the three.js sample. However, the display result is slightly different from Babylon.js. If possible, I would like to make adjustments so that the display results are the same.

Library	result
Babylon.js + EmissiveStrengthTest
three.js + EmissiveStrengthTest.gltf

Edit 2022-09-23 I fixed it because the link destination was localhost.

[emackey]

@elalish Any thoughts here? It looks like ThreeJS is blooming the clamped white color, not the original emission color. Is there a way to do otherwise?

[elalish]

Interesting; I haven't played with three's bloom shader yet, but I would hazard a guess that perhaps tone mapping is being applied before the bloom pass instead of after? That might also explain why the right two cubes look almost the same. Perhaps it just isn't using a floating-point framebuffer for the intermediate draw, so all the values are getting clamped? @mrdoob in case he has any pointers.

[cx20]

I used the following example as a reference for Bloom's implementation.

https://threejs.org/examples/webgl_postprocessing_unreal_bloom.html https://github.com/mrdoob/three.js/blob/master/examples/webgl_postprocessing_unreal_bloom.html

However, I didn't know how to use toneMapping correctly, so now I use THREE.NoToneMapping. https://github.com/cx20/gltf-test/blob/a171346766c8cd168e62bb53f7101ebf9fa592f7/examples/threejs/index.js#L101-L106

[cx20]

I tried changing the toneMapping to use THREE.ReinhardToneMapping.

const params = {
    exposure: 1,
    bloomStrength: 1.0,
    bloomThreshold: 0.6,
    bloomRadius: 0.5
};

renderer.toneMapping = THREE.ReinhardToneMapping; 

Library	result
before
after

[elalish]

Ah, that makes sense; without tone mapping you lose all this HDR info to clamping. One note: ACESFilmicToneMapping has become a bit of an industry standard, so you might want to go with that.

Using tone mapping is quite important for PBR rendering generally, not just for bloom. Take a look at my article.

[cx20]

@elalish Thanks for all the advice. I'm experimenting with ACESFilmicToneMapping now. I am trying different parameters but Bloom's emission color seems to be white. Any advice?

ToneMapping	result
NoToneMapping bloomStrength: 0.5, bloomThreshold: 0.9, bloomRadius: 0.5
ReinhardToneMapping bloomStrength: 1.0, bloomThreshold: 0.6, bloomRadius: 0.5
ACESFilmicToneMapping bloomStrength: 0.6, bloomThreshold: 0.9, bloomRadius: 0.0

[cx20]

I have created the following sample to facilitate comparison of the effects of Tonemapping and Bloom. three.js Bloom Test https://jsfiddle.net/cx20/x1g2j07y/

ToneMapping	Bloom Off	Bloom On
No
Reinhard
ACES Filmic

[cx20]

I have tried the same combination of tone mapping and Bloom in Babylon.js as well. Babylon.js Bloom Test https://jsfiddle.net/cx20/ky0tcgr6/

ToneMapping	Bloom Off	Bloom On
None
Standard
ACES

[elalish]

I'd say Babylon looks more like I'd expect here; this looks like enough info to file a bug on three.js, especially if you include a link to your code. ACES pulls bright colors toward white, but the fact that the whole bloom looks white makes me think it doesn't have the right shape of exponential fall-off in intensity.

[cx20]

When I upgraded three.js from r152 to r153, the effect of Bloom became dazzling. It might be a usage issue, but I haven't found a solution yet.

model	three.js r152	three.js r153
EmissiveStrengthTest