Share with us some of the results you generated with RapidCompact!
I’ll do one of our standard examples, to have sth to start with 
This armchair is originally from Turbosquid and has 878K triangles / 58 MB:
One demo version of the same asset, reduced by RapidCompact using UV-preserving simplification, has 220K Triangles (~25%) / 2.4 MB, and I believe it is visually nearly identical to the original:
As you can see, the file size is already in a pretty neat range, which was possible due to the UV preserving simplification that allowed the texture maps being tiled, even in the simplified version, and hence we could keep a lot of detail without requiring too much resolution. Also, 100KB were consumed by the transparent texture of that that “soft shadow” plane, which was somehow modeled into the original asset (and could have been removed):
IIRC, what required the still pretty high resolution, in order to have good-looking output, were all these little strands from the piece of cloth.
Asset size could be further reduced by applying basis compression (using the glTF KTX2 container format) to the texture maps, and I’m not sure if we used draco… but anyway, 2.4MB is a nice size for this kind of asset, guess improvement should go into the geometry reduction (seeing how far we can go down before it breaks), in order to provide a more fluent interaction on devices with weaker GPUs
2 Likes
To start my internship at DGG and to get to know RapidCompact better I tried to optimize the following bread model. The original file has 266K faces and 12.2 MB.
I tried to reach a megabyte target of under 100 KB without major inaccuracies. With a mesh resolution of 3% of the faces, switching of all the textures except the base color map and a ktx2 texture compression the result was 6670 faces and 99.7 KB.
2 Likes
< 100KB is quite impressive (especially since it still looks usable^^)
Thanks for sharing this!
1 Like
Hi there,
would it be possible to see the example in 3D? It is difficult to see the quality with only an image.
Thanks for sharing!!
2 Likes
Couldn’t resist to share another quick one - this one mainly benefits from texture optimization (131.61MB => 5.42MB). Cool thing is, it is actually animated (has rigid transform animations, which are preserved by the optimizer)
Live demo:
https://api.rapidcompact.com/viewer?id=5sWVCHilSP
2 Likes
Here is an example of a model optimized on our cloud platform with a preset to hit the optimal requirements for a specific software usecase.
(cloud preview image of the input)
Input Model 5.13 MB & 22.700 faces (with Adobe Aero Preset) => 1.41 MB & 22.700 faces
This is a great example of how much optimization is possible even without touching the face count.
Adobe Aero ready asset:
3 Likes
Sure!
The first link is for the original:
https://rapidcompact.com/demo/Brot/index.html?modelURL=scene.gltf
and this is the small version:
https://rapidcompact.com/demo/Brot/index.html?modelURL=scene-small.glb
2 Likes
The case of last Model of the Month was quite impressive: using the DropTextures Workflow, it went from 150 MB to 0.85 MB. Apart from texture simplification, the model allowed quite a lot of decimation, as it was hard-surface and had big surfaces.
The 3D Embed
1 Like
Here is another impressive result I got with Rapid Compact.
This original model was 312 MB
Here is the optimised version coming in at 5.3MB . Despite optimising the model it still looks as good as the original.
Link to 3D viewer : RapidCompact 3D Viewer
2 Likes
I wanted to play around with this sci fi helmet in AR but its original version was not optimised for augmented reality. I used the Hololens preset to take it from 32 MB to 1.75 MB. I decided to try my own custom preset and reduce it to 1.68 MB. The optimised version of the helmet maintained the details impressively well.
RapidCompact 3D Viewer
As you can see there isn’t much of a difference 
2 Likes
very nice, thanks for sharing 