VideoLab · Source archive

Videolab: Mara Voss and Sorrow Lake

Originally published 2026-09-21 on Advisory Hour.

Originally published as an Advisory Hour Substack post on 2026-09-21. This first-party site copy preserves the text, localizes public supporting media, and connects the record to its project trail.

Original post

Videolab: Mara Voss and Sorrow Lake image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake image from the original Advisory Hour post.

This builds on ideas from Mara Voss and the Mechanical Monsters, as well as the Golden Egg to focus on sound and dialog design. It’s a story of complications.

Videolab: Mara Voss and Sorrow Lake video 1 from the original Advisory Hour post.

The scene art for this video doesn’t disappoint with so many outstanding scenes. The following gallery is just a small gallery.

Going forward there’s a thought I have to build out an expanded Mara Voss section on my site. This never would have occurred to me prior to doing these videos. There’s a unique power in just taking action. Don’t sit idle, or you’ll have to contend with your own Sorrow Lake.

Proof sites are websites that collect and organize the material of your project as the coding harness does it’s work. It’s like a live documentation snapshot. I mentioned in prior articles that I’m a big fan of proof sites. I used to human in the loop them. I don’t these days. Now the agents perform that action and in combination with local vision AI plus exceptional computer-use skills can accomplish quite a bit.

Videolab: Mara Voss and Sorrow Lake image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake image from the original Advisory Hour post.

Individual shots still get tossed into a contact sheet. These area a way for the model to leverage a vision system that doesn’t support video. Mysteriously, a vision model that’s just image-native can still process these contact sheets.

Videolab: Mara Voss and Sorrow Lake image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake image from the original Advisory Hour post.

There’s still a lot more in the production of these long context videos. They extend the work from John’s World and now I feel like this 5 minute to 10 minute region is largely solved. Soon I’ll be embarking on a far more ambitious hour-long project. What I’ll say is when it comes to local AI having this media capability feels like something most everyone will likely have in the next ten years. Working with the tools, it just seems likely every pixel and every world will be one prompt away. The bigger puzzle is how to navigate that world, and that is a very interesting puzzle because it departs from everything we know today.

Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.
Videolab: Mara Voss and Sorrow Lake gallery image from the original Advisory Hour post.

Connected work