Workflow Audit & Opportunity Mapping
We analyze your existing post-production pipeline and identify exactly where AI automation delivers the highest ROI.
emd3000 — post-production automation
We design and build intelligent automation systems that replace or dramatically accelerate the repetitive, time-consuming parts of post-production — from VFX compositing and motion graphics to color grading and asset management.
02 — What we do
We analyze your existing post-production pipeline and identify exactly where AI automation delivers the highest ROI.
Bespoke workflow builds tailored to your project — for example automated motion-graphics pipelines.
Seamless embedding of generative models and computer-vision systems into your existing software environment.
03 — How it works Draft copy
04 — Samples
Project_00Track 01
Dijkstra's shortest-path algorithm
Based on the famous algorithm, neuronal growth was visualized for a documentary on intelligence.
1 clipdrag / swipe →
Project_01Track 02
UAV & ground, small to large scale
Small- and large-scale photogrammetry via UAV or on the ground. Operated various drones and cameras, merged the data and reconstructed mostly via RealityCapture and PTGui.





6 clipsdrag / swipe →
Project_02Track 03
BLK2GO × Leica Zurich
Large-scale LiDAR scanning via BLK2GO in cooperation with Leica Zurich. Scans were post-processed to point clouds and polygons in Leica Cyclone, CloudCompare, MeshLab, RealityCapture and others.
Converted to polygons, the scans became the foundation for precise camera reprojections to reconstruct movie sets and virtual walkthroughs. They were merged into a large-scale polygonal model of the city of Zurich, with high detail in specific regions.


3 clipsdrag / swipe →
Project_03Track 04
360 cams, NeRFs, photogrammetry, metadata
Merging tracking data of objects in videos from various sources: 360 cams, photogrammetry, NeRFs, camera metadata and more.
Designing anatomically and physically accurate adaptations of humans, animals and other assets. HQ-HDRI generation via 180° fisheyes and multi-exposures. The Facial Action Coding System was used to redesign facial expressions, and motion sequences and physique were transferred from one “actor” to a digital copy.






18 clipsdrag / swipe →
Project_04Track 05
R&D for final simulations
A collection of R&D work that builds the base for final simulations in Houdini and Blender: meteor impacts, bullet impacts, grenades and more.
11 clipsdrag / swipe →
Project_05Track 06
One million drones over Basel
Boids and other flocking algorithms used to mimic a fish or bird flock of one million AI-controlled drones guarding a building in Basel.
4 clipsdrag / swipe →
Project_06Track 07
V2V · I2V · T2V in ComfyUI
Image and video generation and manipulation workflows — V2V, I2V, T2V and more. A showcase of prototyping visualization workflows with open and closed source models in ComfyUI.
ControlNets for motion and scene transfer with multimodal approaches. Researched sigma curves and mel-band transformers and built automatic pipelines that reprocessed whole feature films locally, as well as audio manipulation.

5 clipsdrag / swipe →
Project_07Track 08
Classic VFX pipelines in VR
Previz and R&D for VR, showing classic VFX pipelines.
5 clipsdrag / swipe →
Project_08Track 09
Reconstructing lost family members
Reconstructing deceased family members and writing budget proposals. Coached a VR-therapy startup in Hamburg (FAMCONS, Christian Lessiak).
Created budget proposals for VR production, designed moods for the “patient view” and developed VR-therapy methods together with the therapist.


2 clipsdrag / swipe →
Project_09Track 10
Reprojection of stabilized 360 backplates
To simulate a POV out of a flying object — a car picked up on a chain hanging from a helicopter — I went through a heavy R&D phase with tests on consumer hardware. The final shot, under my supervision, was then shot with a cinedrone.
3 clipsdrag / swipe →
Project_10Track 11
Markers, survey cams, dome distortion
A very special and complicated task: visual object and scene tracking via markers and various survey cams, merged into a single camera / object track.
Difficulties came from the optical characteristics of the water, from setting up underwater, and from the breath of the actress. Special lens grids were used and the dome distortion of the cinecamera was calculated.


6 clipsdrag / swipe →