Color & Flow is the consulting practice of JD Vandenberg, a global technology leader and innovator with 20+ years building production technology, color pipelines, and creative tools for Netflix, Disney, Marvel, Technicolor, and ARRI.
We tailor each engagement to your content-creation standards and production needs, integrating AI in ways that protect creative intent and improve the workflow. Our strength is translating between filmmakers, post teams, engineers, and emerging AI tools, making complex technology practical, reliable, and ready for production.
Frontier model integration via ComfyUI and Python: tool evaluation, LoRA training, and fine-tuning. Safe adoption, workflow guardrails, and team training.
Level up your color pipeline with open standards. Color grading setup, display/projector calibration, color-critical viewing environments, and LUT/CLF development built on ACES, OTIO, and CLF.
Post-production workflow design, camera matching (ACES/CDLs/3D-LUTs), dailies, VFX delivery, cloud pipeline architecture, and animation/live-action/hybrid workflows.
HDR/SDR home and cinema deliveries (Dolby Vision, HDR10, Dolby ATMOS), QC and delivery validation.
Custom scripting and pipeline automation (Python, MATLAB, Swift).
We are still in the early days of generative AI. We see its greatest value not in replacing creatives, but in giving them more room to create.
Our approach is to apply these tools thoughtfully: reducing friction, accelerating iteration, and shortening the path from idea to delivery, so your workflow supports your vision, rather than forcing your vision to conform to the workflow.
Day for night is a technique where scenes are shot in daylight, then processed to look like night footage. Productions use it to avoid the cost, safety risk, and scheduling problems of real night shoots, especially with child actors, restrictive permits, and stunt or crowd work.
The catch: sunlit shadows, daytime sky brightness, uncontrolled highlights, atmospheric haze, and color all read as daytime, and no existing process convincingly removes them. Modern high-dynamic-range sensors make this harder, not easier.
We shot a series of aligned day and night video pairs to build our own training data, then trained multiple LoRAs on it.
We applied the trained LoRAs to footage the models had never seen, to test how well the results held up.
Use an SDR-to-HDR fine-tuned model to add missing details in shots' highlights.
Trained a LoRA to learn how to colorize black-and-white images.