Midjourney reports Medical launch video used Sony DSLR, DaVinci, and browser-based three.js

• Midjourney's launch thread linked to a technical dive on its new scanner, while DavidSHolz said he personally filmed and edited the promo video and built the realtime 3D visualization that appears in it.
• In a follow-up, DavidSHolz's later post said it had been about a decade since he last shot and edited a video like this, which makes the launch feel less like a polished agency handoff and more like a founder-made production.
• The visualization stack was browser-native: DavidSHolz's web stack reply said it was built with three.js and WebGL, and another DavidSHolz reply added Claude and Codex to the toolchain.
• The camera and edit setup were comparatively plain. According to DavidSHolz's gear reply, he used a Sony DSLR and DaVinci, a tool he said he had never used before.

You can jump from Midjourney's original post to Holz's production note, then to his browser-stack breakdown and his gear list. The surprising part is how little of this sounds like a traditional launch shoot: a Sony DSLR, first-time DaVinci use, and browser visualizers that Holz said were originally headed toward an interactive web version.

Holz said he shot and edited the Midjourney Medical launch video himself, then added that it had been around 10 years since he last made a video like this. That detail shifts the story from a generic product trailer to a creator workflow case study built inside the company.

A separate reply from DavidSHolz on the creative goal says he wanted "gutsy" detail, not an Apple-style commercial. That matches the final piece's technical-art direction tone.

Holz said the 3D visualization was realtime, browser-based, and built with three.js and WebGL. In the same thread, his first reply says he implemented the visualization himself.

That makes the video more interesting than a motion-design wrapper around product shots. The core visual language came out of working web graphics, not a pre-render-only pipeline.

The production stack breaks cleanly into two parts:

• Web visuals: three.js, WebGL, Claude, Codex.
• Capture and edit: Sony DSLR, DaVinci.

Holz also said in a follow-up about coding tools that modern coding tools made him feel like "a technical art director with no limits." For AI-native creators, that is the sharpest line in the whole thread.

Holz said he originally wanted to make every part of the stack interactive on the website, then cut that scope after burning out and pushed for a video instead. He also said he might return to a web version later.

That leaves one more useful production note in the open: the shipped video was not the first concept. It was the reduced-scope version of a larger browser experience.