Along with long-term Knightmare fan Andrew, we're exploring how a new series of Knightmare might look with AI artwork. In part 2, we think about the technical challenges and what leaps might be needed to attract young viewers of today.
In the first part, we looked at a generated AI impression of what a dungeon could look like if reimagined as a Tolkien-esque fantasy kingdom in a fine-art style.
But devil's advocate does kick in at some stage. How might some of these scenes work? And is generative AI really the answer, or any part of it at all?
Technology was always at the forefront of Knightmare, and the show pushed it to its limits over its eight-year run. But there has always been tension between the technological frontiers available and what works well for television viewers.
Even by the mid-1990s, Knightmare creator Tim Child found himself frustrated by the limits of technology and wanting to move towards a full virtual reality framework.
This vision was realised with a pilot for Knightmare VR in the early 2000s, but it proved less satisfying to watch.

Once the player as avatar could see their surroundings, the basis of what made Knightmare successful was gone. The peril, the panic, the mistakes, inexplicably sidestepping left into a chasm - where was that going to come from?
The formula that made Knightmare successful cannot simply yield to technology. But could technology play its part in making the original Knightmare better?
Prop crops
As shown in Part 1, the design suggestions are plentiful. Generative technology understands the properties of certain types of rooms: libraries and studies, caverns, throne rooms, dungeons.

But one problem surfaces immediately. Part of what makes some of these designs atmospheric is fixtures and fittings that require precision engineering.
We recall, Knightmare worked to exacting grids so that blue props such as staircases or wells that were featured within the artwork could be exactly aligned.

No amount of technical wizardry will circumvent the challenge of real people engaging with large props that form an integral part of any artwork.
It would still be prudent to reduce prop complexity where possible. The fewer challenges you add to manufactured environments, the more expansive you can be with them.
Knightmare did this in its later series by replacing well-ways and staircases with other video or CGI transitions. You might see a staircase in a room, but any interaction with it was footage shown through the Eye Shield.

Knightmare always was - and would remain - an expensive enterprise. Technology would still need to promote working smarter, not harder.
Game engine power?
One area where Knightmare was intentionally sparse was immersive perspectives.
Most of the time, we saw the default room view - aligned to the grid - with some secondary angles where a non-playing character was involved.
Very occasionally, we saw reverse views. Some of the hand painted locations, Banburn and Death Valley, had extra artwork designed for this purpose.

Otherwise, this was achieved via video (Giles, Series 4), CGI (Daniel, Series 8), or by avoiding bluescreen altogether.
The Geek Week episode includes an aerial shot of the 'Slice Me Dice Me' challenge in a throwback to the VR pilot where that was introduced.

To feel like Knightmare had made the leap technologically, this variation would need to increase.
Could AI conjure enough backdrops at pace to cover multiple perspectival shots required for a scene? Or could its machine learning capabilities help to generate detailed topologies for room designs, providing the specs to recreate them consistently from any perspective?
Another option might be to increase the capacity of the bluescreen to more of a box, which in turn would increase the opportunity for perspective shots.
The challenge here involves the lighting and cameras, and whether these could be minimised enough to enclose the player more completely.
Alternatively, could we harness the power of a game engine to create 360-degree surroundings on the fly? (Last year, producer Ben Spiteri contemplated Knightmare in the Unreal Engine.)




Rooms with narrow ledges have always been an exciting and popular challenge. Imagine watching a dungeoneer cross a ridge from any number of angles, with the view of the room engineered as if it had been rendered entirely. Not unlike Knightmare VR, but with original dungeoneer gameplay.
And not all of this would have to be prepared in advance for contestants. Simply capturing the movement from multiple angles enables a post-production experiment for a better viewing experience.
The animator's time
That's a world of static scenes. You could create a stock of varied and stunning artwork.
But how then to bring it to life in the same way that the Travelling Matte Company did in the original series?
AI would be fine for light-touch animation, such as flaming torches and igniting bombs. It may also be able to influence a chamber based on a common phenomenon, such as cracking ice.



However, generative technology is probably a little short of being ready to animate complex sequences convincingly, such as a creature being able to move or speak reactively.
That is just one area where human input would be well served. Rasterising might a huge opportunity that technology can offer.
Could an interpretation of a Catacombite or a generated monster be converted to a vector and animated?




Knightmare dealt with many of its threats economically. In the early series, some of the pursuits were pre-filmed and recycled.
In the Corridor of the Catacombs, the Behemoth always targets the right exit even when a team chooses the other.
The goblin pursuits through Banburn and Death Valley were also recycled, with the latter recreated terribly in the international versions.
Were Knightmare to remain on the precipice of technological advances, it would benefit enormously from expertise in animation and game development to render scenes in as close to real time as possible.
It may feel like a reinvention, starting from the beginning again, and iterating – much as the original series did.
But bringing those possibilities together, a future series could be irresistible.


