This one started the way these little projects always start: I knew something was not quite right, and I happened to have spare time and spare inference to go and chase it. Cinematic Mode is the corner of EF-Map where I let myself do that. The rest of the EVE Frontier map is a tool, and a tool has to behave. Cinematic Mode, the mode I rebuilt from the ground up a couple of weeks ago, is the one place I get to break things for fun and see what the starfield can actually do.
What follows is six or seven hours of work, most of it spent waiting on an LLM to action prompts, chasing one itch into the next and somehow ending up with the best the mode has ever looked.
Gas between the stars
The thing that had been nagging me was that the space between EVE Frontier's stars was just, well, empty. Real space is not. There is a faint medium out there, thin gas and dust drifting between the stars, and I wanted some sense of that in the map.
We had tried this once before and it went badly. That first attempt dropped what looked like little atmospheric clouds into the scene, and they sat in strange places amongst the stars, as if someone had photoshopped weather into deep space. It looked wrong enough that we pulled it straight back out.
This time we came at it as layers. The LLM produced five of them, and out of the five we kept three. The first, and the one that does most of the work, is a diffuse gas spread through the whole universe. I want to be clear that this is not a background effect like the nebula we already paint behind everything. This gas lives in amongst the stars, in the actual volume you fly through, so as the camera moves the stars pass in front of it and behind it. The second layer is a set of darker passages, dust lanes that swallow a little light, kept deliberately subtle so they read as depth rather than as smudges. The third is a soft glow that pools where the stars cluster most thickly, the combined light of thousands of them rather than bloom on any single one.
All three were tuned with sliders until they looked right, and then the values were locked in. That part was done, and it already looked far better than the empty version. But the trouble with really looking at a scene is that you start noticing everything else that is wrong with it too.
The flicker that would not go away
Here is one that had been quietly bugging me for ages, the sort of thing you put up with because you know it is not right but it never quite climbs to the top of the list. The bloom on distant stars flickered. As you moved the camera, faraway stars would pop their glow on and off, and once you have noticed that you cannot un-notice it. In motion it was genuinely distracting.
The first attempt to fix it failed, and it failed in an almost funny way. It gave bloom to the distant stars and took it away from the close ones, which is exactly backwards. For a while there the whole effect was reversed, which at least proved we had found the right lever.
The second pass got it. Instead of guessing, it went back to the older, known-good values, compared the two side by side, and rebuilt the bloom so that a star contributes the same amount of glow no matter how far away it sits. The flicker that had persisted for so long simply went away. That alone would have made the session worth it.
Making the explosions messier
With spare inference still on the clock, the next thing on the list was the supernova. When one went off, the shockwave expanded as a perfect circle, a clean ring being scaled up on the screen. Space does not do clean rings. So we spent a while turning that into something with real turbulence in it: an uneven shell with variable thickness and ragged, fragmented edges, still clearly a blast wave but no longer a diagram from a geometry class. The central flash, and the difference in character between a supernova and a kilonova, all stayed intact. It just stopped looking manufactured.
Events that light the gas
And this is where it got fun, because fixing the explosions led straight back to an idea we had already failed at once. A while ago we wanted the transient events, the nova, the neutron star, the shooting star and the supernova, to light up the material around them. Back then the wispy clouds were not fit for purpose. There was nothing real out there to catch the light, so it never worked and we dropped it.
But now there is an actual interstellar medium between the stars. So we tried again. When one of these events fires, it now illuminates the gas near it. The mental model is exactly what you would expect from the real thing: a bright event goes off, the dense material around it catches some of that light and briefly glows, and the empty gaps stay dark. It fades as the event fades. A supernova throws the most light and reaches the furthest, a nova is smaller and quicker, the neutron star pulses in time with its own beat, and the shooting star drags a faint smear of light along with its head as it streaks past.
The important part, the part that makes it feel real rather than like a lamp being switched on, is that the light only shows up where there is gas to catch it. The shape you see is the shape of the medium itself, not a glowing ball drawn around the explosion. Dark lanes stay dark even with a supernova going off behind them. I have to say, it is looking a great deal better than it ever did before.
None of this is trying to be accurate, mind, and that is true of the mode in general. When a supernova lights up the gas it does so instantly, across hundreds of light years in every direction, which is not remotely how light behaves in the real universe. But this is cinematic mode, not a physics simulation, and that is a trade I will happily make for something that looks good in motion. The irony is that my own favourite thing in the whole mode is one of its more realistic touches: the stars are coloured by their actual temperature, and the hotter ones bloom brighter for it.
So why do I keep coming back here?
Fair question, and I asked myself the same thing. We recently started tracking how often each panel gets opened, and over the last few days the Tools Library was opened more than four hundred times while Cinematic Mode was opened a little over thirty. It gets used, but it is nobody's main reason for being here, and I know that.
The honest answer is that Cinematic Mode is the only place I am allowed to play. The main starfield is a tool that people rely on, so I cannot go messing with how it renders just because I fancy trying something. Cinematic Mode has no such job. It is where I get to experiment with things that have an immediate visual payoff, and that is exactly the sort of work that keeps me coming back.
It is not free, mind you. On my own machine, at 1440p on an RTX 4070 Super, the mode runs at around 75 frames a second and eats between fifty and sixty percent of the GPU while it does it. This is not a feature for a laptop. The everyday map even has a whole Performance Mode for weaker hardware, and Cinematic Mode is cheerfully the opposite of that. It is a feature for sitting back and letting the EVE Frontier universe drift past.
Where it goes from here
Will any of this make it into the main map one day? Probably not, and that is fine. Is it a decent showcase of what WebGL can do inside a browser? I think so, yes. And if you are the sort of person who likes having something quietly beautiful running on a second screen, you could do a lot worse than this.
Will it change again? Almost certainly. Do I have plans for it beyond the obvious? Absolutely. I will leave that there for now.
What I will say is that none of this is only about making the map look pretty. Every one of these passes is a learning exercise, in the rendering itself and in driving the LLM tooling and picking up the vocabulary that goes with it, and that knowledge carries forward into whatever comes next. There is a point to it. And at the very least, it is turning into rather a nice showcase for the EVE Frontier universe.
If you want to see it for yourself, open EF-Map, drop into Cinematic Mode and let the auto-tour take you for a wander. There is no button for the big events, so you will have to keep half an eye out and wait for a supernova to go off on its own. When one does, watch the gas around it catch the light.
Related Posts
- Refreshing Cinematic Mode: Our Oldest EVE Frontier Feature, Rebuilt - The rebuild this work continues, where the mode dropped its faked effects
- Giving Our EVE Frontier Starfield HDR Stars - The HDR bloom and nebula backdrop these new layers sit on top of
- Making EF-Map Faster: Three.js, WebGL and Performance Mode - Why an effect this heavy stays out of the everyday map
- Cinematic Mode: Immersive Exploration - Where the whole idea started