Kula Project

It's been an year since I started this project and almost half an year since I started this post that was supposed to introduce you the concept behind this project. Things stumbled many times, many corrections had to be done and a lot of work to be updated. Lessons were learned, new white hair threads appeared and a patient collaborator was won. This is the story of Kula Project.

Jump to

I’ve always been fascinated by lost worlds, civilizations, kingdoms, cities, species, and things that no longer exist. What did they look like? How were they organized? Were they thinking much differently than us? Is this what awaits us in the future as a society?


Having these sorts of questions in my head, it would seem natural to want to visualize how a now ruined fortified boyar house from the early 19th century looked like in its early days, and what the inhabitant’s lifestyle might have been back then. So, back in December 2021, I decided to go for this want. I started a personal project for the virtual reconstruction of such a building with all its interior objects and furniture. More than that, I wanted to experiment with storytelling modern technologies in online media. 

One specific idea that I wanted to implement for this project is related to the way the user interaction with the 3D reconstruction is experienced. The concept is quite simple, it simulates a real-life ‘doll-house’ approach. Meaning that the user would be able to interact with certain elements of the 3D scene to access new areas. For example, clicking on an exterior wall would determine the wall to move away and let the user take a look inside the room behind that wall. The same with a selection of cabinet doors and other objects.

Another aspect that I wanted to tackle was the short interest span for the public this kind of project has. Once they are revealed, with the exception of those passionate about this field, they lose the return value. In this regard, I had a few ideas but wasn’t sure they could be materialized in a practical way. In short, I wanted the 3D scene with the virtual reconstruction to be … alive. To have its own timeline with day and night cycles and different weather or seasonal events, so that whenever you would visit the world, there would be a chance for a different experience. For this complex approach, I was joined early in the project by Alex from VaultCG.

This project has three main development stages, each with its own components: 

  • Reference research and scene design
  • 3D virtual reconstruction with 3D modeling, texturing, and optimization for online usage
  • Development of the web app with its features.

So far, the first two stages are done while the third is still ongoing, as it started as soon as the 3d modeling was done. Not long ago.

About Cioabă-Chintescu kula

This fortified house, a kula (from the Turkish kule = tower), is a typical fortified building found in southwestern parts of Romania but also in northwestern parts of Bulgaria and neighboring regions of Serbia, built between the late 18th century to mid 19th century. These buildings were built by boyars (rich families or nobles) with the initial purpose of defense, refuge, and signaling against the often brigand raids coming from the south. They were usually built in strategic geographical areas so they could chain the signaling (with the firelight, much like in the Lord of the Rings) to send the warning to distant areas in no time. Many of these types of buildings have been destroyed in time, but some of them are still around.  A very nice recent project, Cule în lumină, documented and promoted all these monuments for the public. Of those still around few of them are restored and well-conserved and can be visited, but others are left to ruin and there is no access to them. The Cioabă-Chintescu kula is one of the latter.

Building brief history

  •    started by Răducan Cioabă 1818
  •    finished by Marin Chintescu in 1823
  •    used as storage between 1951-1964
  •    basic restoration between 1964-1967
  •    functioned as local Ethnographic Museum Slivilești 1967-1989
  •    deserted from 1989
  •    bought by the Local Council from the owners in 2016
  •    emergency roof replacement by Ambulanța pentru Monumente in 2020

Building description

  •   Rectangular planimetry with thick walls (0,80 m – 0,85 m)
  •   Three levels
    •     ground floor: functioned as cellar
    •     first floor: as room and kitchen
    •     second floor: as room 
  •   Stairs within the walls 

Its simple, yet charming, architecture has one of the most representative styles for this type of building. It was chosen for this experiment out of pure personal preference but also due to its advanced state of ruin. 


The documentation stage had two directions: collecting information about the monument itself and about similar monuments that are still in use. For the monument itself, there were researched and collected published documents, books, and other studies, publicly available images, public videos and an on-site visit for current state photos, and brief aerial photogrammetry. Because in time the building has suffered several modifications, the purpose of this research was to find photos and data about the monument’s earliest days. The other important part of the documentation was the study of the two very well-preserved kulas in Măldărăști (Vâlcea county), namely the Duca and Greceanu kulas. The important value of these kulas was the interior. As I couldn’t find any information about the interior design of the Chintescu, these two provided very important data in this regard. 

I must mention that I did not work with an art historian or any specialist that studied this monument’s history, as it would have been normal. But I had a brief and very informative discussion over the phone with Carina Tataranu, a volunteer with Ambulanța pentru Monumente, who knew a lot of details about this monument. She confirmed much data I already had but also offered other new details that reshaped a little bit the modeling of the interior. 

Documents and publications

A great help during the documentation stage was the new online repository for cultural publications in Romania, the Digital Library of Cultural Publications ProEuropeana,  which digitized and made available entire journal collections with cultural or archaeological themes. It can be accessed freely at this link. Other important sources were also found online.

  • Radu Crețeanu, Culele și casele întărite de pe valea Motrului, 1958, Monumente și Muzee – Buletinul Comisiei Științifice a Muzeelor și Monumentelor Istorice și Artistice,  nr. I, p.93, Editura Academiei R.P.R.
  •  Elisabeta Ancuța-Rușinaru, Despre două cule din județul Dolj, 1979, Revista Monumentelor Istorice, nr. 2, p 77. 
  • Șerban Bonciocat, Corina Popa, Cristian Bracacescu, 2014, Cule. Case fortificate între fală și ruină, Editura Igloo Media
  • Cule in lumină project at culeinlumina.ro 

Public media sources

Of great value were all the vintage photos that could have been found (see the above gallery). The older, the better. Until I managed to get in-person on-site, a very valuable resource especially for the interior was all the online videos and interviews that were filmed there throughout the time. Also different photos with the interior on social media or other online news outlets reporting on-site were very valuable. Since I never managed to get inside (it is quite dangerous and also restricted access) these images proved crucial for interior design. Again, this being a personal project I did not have real weight or leverage in persuading anyone to give me access inside.

On-site documentation

In truth, I managed to get on-site while most of the virtual reconstruction was almost finished. After getting back some major changes with the interior, especially the first floor, had to be done. I had two floor plans where each indicated a different position of the door entering the kitchen. Using my drone I could photograph through the windows the real location of that door. 

I used the opportunity to make a brief aerial photogrammetry scan. I used a DJI Mavic 2 PRO with its 20 MPx sensor. I did not use GCP markers, so the precision and scale were based on Mavic’s GPS coordinates for each photo. Which is not stellar by itself. A total 271 of RAW photos were recorded for the photogrammetric reconstruction purpose. There were some problems with a tree that was too close to the building so I had to drive the drone underneath branches and make all sorts of creative shots to get as much of that section of the building facade. Images were processed with Adobe Lightroom and then processed with Metashape. A 30 million polygons mesh resulted in Ultra High-quality setting. The result is okay for 3D modeling purposes but not at my own quality standard for monument 3D documentation.

Screen capture with the camera positions estimated in the photogrammetric 3D reconstruction (Agisoft Metashape)
Photogrammetry result: Collage of three style renderings: color texture, white texture and mesh surface texture

3D modeling

The documentation was enough for a precise reconstruction of the building and its interior space. In terms of virtual reconstruction transparency the building structure, walls, windows, doors, and floor plans were modeled in accord with reality. The interior design followed the information from the documents: the ground floor was used for storage, much like a cellar; the first floor was used as a kitchen while the third was used as a bedroom. Access between the floors was made on an in-wall staircase. On the first floor, under the stairs, there was a secret panic room, for hiding in case of robberies. 

The modeling was done with Blender and it had several re-iterations and important changes throughout the project, as new data was revealed. This aspect lead to a time-consuming updating of the UV mapping and baking for all the affected meshes and sometimes it caused development and process blockage, as some changes were quite big. A lesson learned throughout this project, regarding 3D modeling, is that planning and sticking to the plan is crucial. The documentation and the sketch of the final product must be as definitive as possible. Any change during the development might lead to unending edits and updates. 

building structure

According to our project plan, the building structure was modeled from many modules that could be easily taken apart and treated individually. So, for every level, the walls for each side of the building were individual objects and, more than that, interior and exterior faces were separated in order to apply different texture and mapping properties for exterior and interior lighting in the web app. The same with the floors and ceilings between the levels.

Elevation and planimetries used. With color the 3D model wireframe

Exterior landscape and elements were inspired by real life location. A well was introduced and modeled from a similar well found near Duca Kula, but it has no real correspondence on the Chintescu-Cioabă kula location. 

Interior objects

All interior objects and designs are hypothetical but based on analogies with paintings, photographs, and other museum interiors that reconstructed and depicted the life of the early 19th century. For the furniture, a lot of images were used as references with different elements, with an emphasis on noble status conditions at that time. A very important role the photographs within the two existing kulas I mentioned before. 

Everything is modeled from scratch but few certain elements were downloaded online, and here are their credits: 

  • firing weapon on the balcony – “Chassepot 1866 rifle – free game asset” (https://skfb.ly/6UXsH) by Andy woodhead is licensed under Creative Commons Attribution (http://creativecommons.org/licenses/by/4.0/)
  • horse – “Horse” (https://skfb.ly/op9To) by nikita is licensed under Creative Commons Attribution (http://creativecommons.org/licenses/by/4.0/).
  • old lady, grass, rocks – assets from the Kore Formation course (Blender for 3D virtual reconstruction of Cultural Heritage)
  • mushrooms – “Lowpoly Mushrooms” (https://skfb.ly/6yZyQ) by Loïc Norgeot is licensed under Creative Commons Attribution (http://creativecommons.org/licenses/by/4.0/)
  • chicken – “Chicken” (https://skfb.ly/otSvQ) by AIUM2 is licensed under Creative Commons Attribution (http://creativecommons.org/licenses/by/4.0/).
  • cucumber – “Cucumber combo” (https://skfb.ly/osnoO) by Lassi Kaukonen is licensed under Creative Commons Attribution (http://creativecommons.org/licenses/by/4.0/),
  • margarinta flower bush – “Margarita flower bush” (https://skfb.ly/68CKM) by Maf’j Alvarez is licensed under Creative Commons Attribution (http://creativecommons.org/licenses/by/4.0/)

3D Renders

All renderings were generated with Blender and edited with Adobe Photoshop 2022. 

The web application we envisioned is more than a simple online viewer. We wanted to offer the user the real possibility to physically interact with the environment, like opening and closing doors and cabinets, and even sliding away entire walls for a better view of the room layouts. Besides interactivity, we wanted to have a dynamic world, with a day/night cycle and even seasonal changes so that the users can always come back and get the feeling of a “living” world.

Our web app is using WebGL for all our rendering needs, interfaced by Three.js. One of our main goals was to offer a realistic view, so we needed to take a few additional steps in order to reach the level of graphical realism we were comfortable with. The web app features real-time lighting, with most area lights as sources. Indirect diffuse light is baked in Blender and applied via light maps. In order to get the nice-looking soft shadows, we’re baking them in Blender in a special cumulative shadow map. The indirect specular light is handled by local reflection probes, which we implemented ourselves. All this baked information is stored in a high dynamic range, and we’re using a custom-tailored solution for tone mapping and color grading.

Another primary goal of the web app was to have it accessible to as many users as possible. This means it needs to be able to run at an interactive rate even on phones. That’s one of the reasons we chose baking light data, instead of generating it on the fly. Polygon count was also crucial for performance, as well as texture size. Our web app offers its users the option to increase/decrease texture quality in real-time if they feel the need to do so. Special care is being taken in order for the web app to run as efficiently as possible. We’re using GPU-compressed formats for our PBR textures and RGBA packing for our HDR textures. Additionally, we’re statically batching geometries on a same-material basis.

But more about the web app on a future dedicated post, once we advance with the development of the features we want to implement. In short, these features will include:

  • day/night cycle lighting system
  • interactive mesh elements (walls, doors, cabinets etc.)
  • hotspots
  • seasonal changes

That’s it for now. 

If you’ve read this far, thank you for your support. All the interest shown in social media to all the posts regarding this project have pushed us forward. This is the first part of the project presentation, another one will be available when the project is done. 

Kula project is but one of the concepts I want to propose for the online presentation of 3D reconstructions and recreations. Hopefully, this year (2023), I will have more time to test other concepts with different approaches to the WebGL environment.


About Post Author

1 comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.