Prompting Astra. Part 1.
I am adding another camera angle to the scene, on top of the double-shadow build. This is the side camera: a 24mm wide shot showing the grave The Villain is about to pilfer.
Journal entry ↓Carny Island Basic Grave Rob Test Alpha 1.1, including the new side-camera angle.
“The mind, once stretched by a new idea, never returns to its original dimensions.” — After Oliver Wendell Holmes Sr.
A mate of mine often references this quote. And I think about it often these days, as one frontier model pushes beyond the next and things I once only dreamt about doing have now become frustrating bugs to fix and complaints about token burn.
There is so much great stuff I have achieved using these tools. I have advanced so far over the past few months. It would be easy to look at all the problems and complain, but making mistakes is how we have got this far.
Having gone through a basic grave dig from top to bottom - from dirt to coffin lining - the basic premise works. And the Lab approach in Unity - one sandboxed, gated project at a time - works. To a point. I'm getting stuff done. Stuff still goes left but not as badly and I can see when and where it is going left.
My approach to this game is to build ALL the systems and engines independently. Test each one. Play each one. Iterate. Then bring in a Unity integration engineer to compile them. Then let Carny rip and tune it. I know I have enough fun built in. I'll just turn elements off to balance that fun.
That's the approach. It will take longer, but the end result will be a mad, fun first-release game achieved for about A$500,000.
So back to work. Adding a 24mm side camera angle to the dig. And the Villain on the tools. And shadows. I have drawn the art of Carny very bright so when I bring in the shaders and the weather, the art still cuts through but the game becomes darker. Feels heavier. Eyes peering at you from shadows. Shadows are key. And the side angle is critical as it will show the substrate being dug out and the surrounding yard. And threats incoming. Where you pile that substrate will also play a part - as you need to hide behind it. (Much later)
This will be the most ambitious Lab build to date. It will also prove to be the most expensive in terms of token burn—but more on that later.
This is the prompt I gave Astra6, along with six storyboards and a Photoshop file containing the animation layers.
Prompt
There are three layers to dig through at the surface: grass, hard dirt and then soil. As The Villain rakes away the grass, it appears in a pile in the background. As The Villain rakes away the hard dirt beneath it, it appears on the pile, on top of the grass. As The Villain shovels the yellow sandy soil, it is added to the pile. The Villain then gets into the grave. I achieved this effect by sandwiching The Villain between foreground and background layers of the grave. The SCRITCH continues to shovel away in the grave, and this substrate is also added to the pile.
As The Villain digs down, he starts to disturb the paranormal. First, small ectoplasmic ghosts appear and swirl around. As he digs deeper, larger green ghosts materialise. When he cracks the lid, Vinnie appears.
This is the first time you have seen ghosts.
The side camera is visible and used only until The Villain hits the coffin lid.
The idea is to cut to this side angle after a few patches of grass have been raked away. Hold on the side angle while a few more patches are raked away, then cut back to the top angle. Cutting back and forth reveals more and more of the substrate piling up.
The artwork is in the Vinnie Grave Rob 16:9 folder, in a file called Side Grave Dig Camera.
There is a lantern on the left-hand side with a flickering flame. You will need to use the shader on it so the scenes match.
The first result from Astra was surprisingly good. Unfortunately, I had clicked on a 40% discount offer on tokens and ticked “Top Up automatically”. Bad move. I'm on a Pro account and pay A$300 a month. I'd estimated that $15 would get me through the day until my account renewed. I estimated wrong and blew through $400.
Tokens are a terrible budgeting unit for a human. You think in tasks: build this system; debug this mechanic; create this agent. The billing system measures computational consumption that you can't reliably know before the task runs.
Coding makes the problem worse.
OpenAI needs a pre-flight meter: “This task will cost you $6.”
I sit here a few days later with the mind expanded. The side angle works and I have just enough tokens left (5%) to bring you this video. Not enough to update the playable scene. That's still played out in broad daylight with only one camera angle. The next Lab will introduce 3D substrate to the 3D Gravebox.


