I just did a quick test of the AI tools for the viewer and it went pretty well. Codex definitely worked more efficiently with the updated integration. I’m excited to get into building something larger to see how much less burden it takes upon my limits.
Here is the prompt I gave it and the video it made. This is pretty much how I’ve always talk to it so far. I don’t feel anything will have to change there. I’m usually much more specific. It just will now be able to find those things without making them first.
That’s a game changer… thanks Dale!
All in all, it took about 10 minutes for a 900 frame 15 second video. The time is definitely less important to me than the rate of its workload. I also asked it for a brief review of how this video was different from previous ones it made. This particular chat in Codex made all three of the
“Dial P47-M For Mayhem!” An Aces High III Trilogy! films.
The only two thing I didn’t like was it did not quite catch the outside view. I wanted at the hanger explosion, but it had the right idea so then it’s just a matter of me dialing in and picking the people or spots I would’ve wanted to see it from them, exactly. The other thing was it, swapped the frames of the rockets, and the bombs dropping having the rockets drop after when they were actually first. This could also be partly my fault not completely describing the timeline when I told it it wasn’t important later on.
Still all in all pretty freaking cool!
My Prompt to CIN3: (this is my chat outside of Codex that I use for set up so as not to burn through context and tokens with the Codex side) I did all of this from my phone and just talked.
There have been some updates to the film viewer. VIDCDX6 has updated you in the communications. I think the new AI tools might help reduce our token usage and make things less manual for the workers. I want to do a test of the new capabilities so I’m gonna describe a scene to you as we would normally and I wanna see how it is built.
There is Ah film titled “Regs” in that film. I’m flying a P 51 and I drop my ordinance on a vehicle hanger on airfield 99 so I wanna see a 15 second dynamic shot of me diving on the hanger and taking out the hanger. I also want to see as I am egressing after dropping it and I want to see the bombs from all kinds of angles and the rockets and especially as I’m pulling away. I wanna see a front facing downward view as I pull away and up and we see the bombs drop taking out the hanger?
No need for any cards, captions, or music. I just want to test a scene I want dynamic shots in and around the plane, cockpit views, outside views, sweeping views, always matching the speed of the moment, cinematic and intense. It doesn’t have to be all in real time either. there can be dynamic shots of the ordinance dropping from different angles at different spot, or players or the hanger itself throughout the same timeframe. Perhaps some snap backs all around when the hanger explodes from different enemy of viewpoints or friendly viewpoints close by. unless you have any questions, please set this up in communications and I’ll tell it to get to work.
CIN3’s Response…VIDCDX6 — Check current communications and execute CIN3’s new WO018 Regs cinematic test through completion. Return the finished 15-second scene for review.
And the video…Aces High III Film Viewer AI Assistant | Test 1.0 | 15 Seconds | 16:9 Widescreen 1060pCodex Quick Review…* Camera placement became more precise: position tracks supplied actual aircraft locations instead of repeated manual navigation.
* Finding action was faster: one event search identified the rockets, bombs, and A99 hangar-destruction message.
* Framing checks were nearly instant: the viewer tested 495 camera/aircraft projections, reporting under its displayed timing precision.
* Capturing became simpler: direct export produced 1080p60 video with game audio, removing OBS setup and recording synchronization.
* Rendering was measurable: 16 full-HD exports, including four framing revisions, took about 122 seconds total. The initial weapon-camera test took another 15 seconds.
* Creative judgment remained the hard part: projections cannot judge drama, darkness, walls blocking the hangar, or whether an explosion reads clearly. Visual review still mattered.
* Token savings weren’t measured: structured commands reduced manual interaction, but I have no reliable before-and-after token or total-work-speed comparison.
I also asked it to suggest some additions that might be further helpful…* Useful future additions: target/building coordinates and damage-state queries would make hangar framing easier. Projection currently checks aircraft, not buildings or falling weapons.
* Another improvement: exact frame-count export and an automated moving-preview contact sheet would reduce endpoint trimming and repetitive QA work.
* Multi-angle batch rendering: submit several camera plans and output ranges in one job, with progress and per-shot results. I had to load and export each angle separately.
* Audio metering built into export: report peak level, average loudness, and silent sections alongside the video result. I used a separate tool to check audio afterward.
* Camera-path safety checks: flag cameras passing below terrain, through buildings, or behind obstructing walls before rendering. Aircraft projection helped with framing, but I still had to inspect pictures and reposition obstructed cameras.
