New Add-on: Image To Mesh

Image To Mesh Conversion

Quick Video Overview:

Pretty self-explanatory (but do check the tool-tips…) ;>


  • Added Color 2 Material Mode

      • Assigns a material by the most common colors, separated with a threshold value.

      • Works with or without Alpha/RGB-as-alpha:

        (Worst case example - such high detail images at 1k+ will take minutes. Smoothing is not recommended.)

      • Adjust material outlines with smoothing value.

        (Tip: Automatic, will not work for all cases: Disable, and use manual smooth-tool for tricky cases.)

      • Please Note: c2m mode is intended for distinct color-separation: “Busy” images will not work:

  • Color reduction threshold option added for Vertex Color Mode.
  • Custom panel location fixed. You can now assign the kit to a custom N-panel location. (in Add-on prefs)
  • Add-on category changed from “Modeling” to “Import-Export”


  • Add Border Expand feature (dilation). Also works in rgb mode. IMAGE
  • Added support for using the same image file on all 3 boolean slots (Front, Right & Top)
  • Fix rounding error (bottom left mesh faces were offset issue, noticable on low res) IMAGE
  • Fix “reverse” tolerance in rbg mode (behaves more like the alpha slider now)
  • Fixed some object invalidation issues when switching between to & from boolean geo ops


  • Batch Process bug fixed (invalid screw flip property when redoing batch)


  • Custom Work Resolution
  • Vertex Color Mode (or, “retro pixel gfx mode”) - uses vertex colors instead of texture(s) for each face converted from the image pixels.
  • Video demo


  • Fixed Console Text Output Total count error


  • Fixed update notification link : should go directly to gumroad instead of 404 the next update ;>


  • Added Batch Processing operator for PLANE and SCREW modes. Uses last used settings: Load 1 image with regular method to fine tune settings before batching an entire folder.
    There is a chance I make bool mode possible to batch: ~0.01% chance atm.


  • Better Dissolve (and just one now) Worst Case-ish example : 1K @ ~1 minute with 12C Threadripper (2017) No multithreading jfyi.
  • User RGB instead of Alpha option (in add-on prefs) Note: The opacity tolerance is kind of reversed for RGB; use lower values.


  • Developed & Tested on Manjaro Linux with the latest official Blender release
  • I do not support alpha/beta versions of Blender.
  • I have no way of testing on Mac (& rarely on windows)

Download here


impressive !!!
zbrush’s shadowbox for blender, using pictures instead of mask.


Thanks for sharing.


I did not know about shadowbox, but that seems to be more for ideation?

i2m was created mostly for situations where you (a concept artist or FX artist) needs a picture turned into 3d in your scene, quickly. i2m can save you time if you don’t need perfect results, or save setup-time as a head-start if you need do more accurate manual modeling.

But, I suppose you could actually use i2m in a similar way to shadowbox, if you edit the image layers and just update & re-run :slight_smile: (I’d recommend not using the dissolve modes for quicker updates in lower working res for “near” realtime updates in such a case)

1 Like

Incredible, thank you! Do you think there’s a way to use a white background instead of an alpha channel?
Anyway, incredible tool, thanks again!


Hey, thanks!
Sure, that could be an interesting method: I’d probably add an option in add-on prefs where one can pick an RGB value or Alpha. (So any color can be used) Next version ;> (RGB will be slightly slower but shouldn’t matter much)

New Dissolve method also coming in next version. A bit better :wink:
Edit: Done, v1.001 :wink:


Wow, nice one!

I had a much more limited version of something similar for Softimage XSI.
Thanks for this, it will useful for sure over here :slight_smile:



The good thing. Thanks for sharing

1 Like

Insane! Now the only thing missing is the RGB color selector on-screen in the menu since different images could have different backgrounds. But holy guacamole, I love this. Thanks.

thank you Kjell
I try to upload an alpha channel PNG (below) but there is an error message that reads: “image not found (blend file) Cancelled”

That image does not have any alpha.

p.s. If it did, that would be a worst case scenario: If you find a version with an alpha-channel, expect long process times & very high polycount. jfyi.

thanks, it had an alpha channel when I opened it in GIMP but I went back and created a new alpha channel and now it seems to work just fine

awesome and very usefull - thanks a lot !

is there any chance of using the alpha channel form procedural maps as input ?

cheers . oliver

No, not unless it’s baked down to an image file.

would it be possible with this type of geometry ?

kei2m is probably not the solution for this situation.

thx . keep up your brilliant work !

1 Like

Is there anyway you can include functionality that would instead of just a png file allow it to operate on PNG sequences

Should be doable, I’ll look into it.

1 Like

Version 1.005 is out, with new features: