KDRK · 20.12 · The Darkroom · film ·
What Stars Leave
The second film. Four of my telescope pictures, built on screen one exposure at a time, ending with a star going out.
A picture of the sky isn't taken. It's gathered. The telescope opens its shutter for two minutes, or five, and what comes back is a grainy frame with a few hundred stars and a hint of the thing you pointed at. Then it does it again, and again, and later you lay the frames on top of each other, matched star for star, and the noise cancels while the faint stuff adds up. Every picture in The Plates was made that way. The finished image hides all of it. This film shows it.
Each object rises out of the grey static with its first raw exposure, and it looks like what it is: red speckle, a shape you can almost make out. The second frame fades in over it and the shape firms up. The third, the fourth. Then the next filter joins in the order I shot it, so the Veil's oxygen arrives as a teal haze and clears into blue-white edges over the next four frames. When the last frame lands, the picture holds for fifteen seconds, and that picture is the one in The Plates. Then it falls back into the static and the next object begins. A small readout in the corner keeps count: the object, the telescope, which filter and which frame, the time the shutter opened, and the light gathered so far.
What stars leave
I picked the objects for the story they tell together. They're all endings, and they all still glow.
The Dumbbell comes first, a star like the Sun that has pushed off its outer layers, photographed from the big telescope at Sierra Remote in late September. Then the Veil, the remains of a star that exploded ten to twenty thousand years ago, from the small wide telescope in Utah on the last night of the free week. Then the Crab, the supernova of 1054, with the pulsar still spinning at its center. The Crab builds in five filters, luminance and the three colors and then hydrogen, which is why its red comes in last.
The fourth is NGC 7331 and the supernova that went off in it this September. That one builds from my October 1 visit, fifteen frames in four filters, and when the full picture is up two small ticks fade in beside a point of light in the galaxy's disk. Then the picture dissolves into the same galaxy sixteen days later, from the frames I took on October 12 and 17. The galaxy holds still, and the point of light is a third as bright. The readout gives the two magnitudes, 12.87 and 14.04, and the music, which has been building since the Dumbbell, lands its one timpani stroke as the star goes dim and the film ends.
Why I like it
Everything on screen is one of my exposures, aligned the way I align them for the pictures, and the only thing changing from frame to frame is how much light has been collected. That took some care. If you stretch each partial stack on its own, the picture jumps around as the noise changes. Instead every stage is finished with the settings of the full stack, the same black point, white point, and color balance, so what you watch is signal accumulating and nothing else. The only motion is a slow push in on each picture, about five percent over the whole shot.
The readout comes from the same files. The times are the shutter times from the frame headers, in UTC. The light-gathered counter climbs through each dissolve and stops at the real total: forty minutes for the Veil, thirty for the first visit to NGC 7331. Across the four objects, the film shows about two hours of telescope time in seven minutes, and nothing in it is faster than that.
I had help. Claude, Anthropic's AI, wrote the code that aligns the frames and renders the reveal, from the same tools we use for the stacking and the photometry. The Veil was the test. When I saw it rise out of the static with the counter running I asked them to build the rest this way, but slower, and in themes. This is the first theme.
The music
Three pieces from the library, one after the other: a low breathing drone, then the ether journey, then the original ether ship track, the one that was written for a voyage out of the galaxy before any film existed. That last track permits itself a single timpani stroke, which was meant to be the moment of breaking into open space. Here it lands on a dying star. Same stroke, different arrival. The whole score is cut so that the end of the music is the end of the picture.
Molly Rose's rule still applies. Leave the headphones on through the static.
The next ones
Two more films are planned in the same treatment. Where Stars Begin goes the other way, with the Elephant's Trunk, the Eagle and its Pillars of Creation, and the Sculptor Galaxy, which is making stars as fast as any galaxy nearby. Those premium sets came with six filters each, so they'll build in more layers than anything here. Island Universes is galaxies only, traveling outward by distance from Andromeda's core to Stephan's Quintet at 290 million light-years. Saturn might get a short piece of its own, forty-three quick frames sharpening out of the shimmer of the atmosphere, if the result is worth watching. The supernova series continues through the end of October, and if it fades the way it should, this film will get a longer ending.
Addendum: how it was made
For anyone who wants the numbers. Four objects and fifty-one calibrated exposures from iTelescope: forty-one built in one at a time, plus the ten frames of the second supernova visit, shown as a single stage. The Dumbbell: four hydrogen-alpha and four oxygen III frames of 120 seconds, T24 at Sierra Remote, the night of September 29 (September 30 UTC). The Veil: four hydrogen-alpha and four oxygen III frames of 300 seconds, T14 in Utah, October 1 UTC. The Crab: two frames each of luminance, red, green, and blue at 120 seconds and two hydrogen-alpha at 300, T24, October 15. NGC 7331: six luminance and three each of red, green, and blue at 120 seconds on October 1, then a second visit of four luminance and two red on October 12 and two green and two blue on October 17, which is shown as a single final stage.
Frames are aligned by matching star positions, an affine fit with a wide translation vote so the second NGC 7331 visit, taken on the other side of the meridian and rotated 180 degrees, lands on the first to within a quarter of a pixel. Each stage is the mean of the frames so far, finished with the full stack's statistics so the stretch never changes. The hydrogen-oxygen objects use the same two-color recipe as the published pictures, the Crab and NGC 7331 the LRGB recipe with hydrogen blended into red. Each picture pushes in five percent over its shot, stars and gas together. A faster star layer for depth was tried and dropped, because it doubled the bright stars.
Timing: 1.6 seconds of static before each object, a fade up of six seconds, five seconds of hold per frame, three-second dissolves, fifteen seconds on the finished picture and on the second NGC 7331 visit, then a fade down and a second of static. That puts the Dumbbell and the Veil at 81 seconds each, the Crab at 97, and NGC 7331 at two minutes thirty-seven, for seven minutes eighteen with the title and the end card. The readout is set in Oxanium capitals at the lower left, from the FITS headers: object, telescope and site, filter with frame k of n and exposure, DATE-OBS, and a light-gathered counter that interpolates through each dissolve. The supernova ticks are two thin ember lines, in the same orange as the site. The three score pieces are three minutes each, crossfaded over twelve seconds, and the last seven minutes or so of the result are taken so the timpani falls at the end of the picture. The film is 1920 by 1080 at 24 frames per second, with the house grade from A Field: the grey static between shots, titles fading in white over it, and no breathing zoom.