Chapter 4

Overlays, Color and Display

The last chapter was about what goes on the screen. This one is about how it gets there: how a modern picture becomes something a 1982 graphics chip can display, what choices you have along the way, and how to stack information on top of whatever is playing.

The Six Display Modes#

The VIC-II can arrange its screen in several ways, and each is a different bargain between resolution and color. c64cast exposes six, chosen with a scene's display setting.

The important thing to understand about the C64 is that its color limits are spatial, not global. All sixteen colors are available at once, but only a few of them may appear within any one 8×8 block of pixels. Every choice below is really a choice about how to spend that budget.

mhires is the default for good reason. Halving the horizontal resolution buys four colors per cell instead of two, and for photographic material that is overwhelmingly the better trade. Use it unless you have a reason not to.

hires keeps every pixel and spends the color budget instead. It suits line art, text, diagrams and anything where a sharp edge matters more than a hue.

petscii builds the picture out of letters, picking a character by how bright each cell is and a color by its hue. The result is unmistakably a Commodore, and because character modes shift far less data than bitmaps, it runs at the machine's full frame rate.

Color, and How It Is Chosen#

Between the source picture and the screen sits a color pipeline, configured once in the [color] section and applied to every scene:

[color]
auto_fit = true
dither = "auto"
dither_strength = 0.5
color_match = "auto"
motion_smoothing = 0.25

auto_fit stretches contrast and saturation to suit the source. Real footage is rarely made for a sixteen-color palette, and letting c64cast fit the material to the palette first makes a large difference. It is on by default.

dither trades spatial noise for apparent color. Without it, a gentle sky gradient becomes visible bands; with it, the eye blends neighboring pixels back into the missing shades. Choices are ordered, blue_noise, floyd-steinberg, atkinson and none.

color_match decides what "the nearest color" means. perceptual measures distance the way human vision does, which usually looks better; rgb is a cruder measure that is occasionally more faithful to a specific palette.

motion_smoothing applies only to mhires. Because that mode picks four colors per cell per frame, a cell whose contents change can flicker as its palette is re-chosen. Smoothing damps that at the cost of slight after-images. Raise it for calm footage; lower it for fast motion.

Most of these are set to "auto" out of the box, which is not a refusal to choose but a decision made per scene: a still slideshow gets floyd-steinberg dithering, while moving video gets blue_noise, because error-diffusion crawls unpleasantly when the picture moves.

Forcing a Palette#

Setting force_palette restricts the picture to a chosen number of colors, or to a specific list of them:

[color]
force_palette = true
force_palette_colors = [
  "black", "blue", "light blue", "white",
]

This is a deliberate stylistic effect rather than a fidelity improvement. Four cold colors make everything look like a monitor from a submarine film, and that is sometimes exactly what you want.

Overlays#

An overlay is a decoration painted on top of a scene after the scene has drawn itself. Overlays attach to a scene, several at a time, and paint in the order you list them:

[[scenes]]
type = "slideshow"
name = "Photographs"
file = "~/Pictures"
duration_s = 120.0

  [[scenes.overlays]]
  type = "clock"
  row = 0
  fg_color = "white"

  [[scenes.overlays]]
  type = "scrolling_text"
  row = 24
  speed_cells_per_s = 6.0
  messages = [
    { text = "WELCOME", color = "yellow" },
    { text = "COFFEE AT ELEVEN", color = "cyan" },
  ]

row is the character row to paint on, counting from 0 at the top to 24 at the bottom, regardless of the scene's display mode. Overlays occupy whole rows, so keeping them at the top and bottom leaves the picture alone.

What Is Available#

Run c64cast --list-overlays for the current catalog. At the time of writing it holds:

Which Overlays Work Where#

Not every overlay works on every display mode, because some paint characters and some paint pixels. To see the whole matrix at once:

c64cast --compat

Two rules cover almost all of it. big_text takes over the entire display while it scrolls, driving the hardware scroll registers directly, so it must be the only overlay on its scene. And the spectrum analyzer comes in two versions because character modes and bitmap modes need genuinely different implementations: use spectrum_petscii on character modes and spectrum_bitmap on mhires.

Sound#

Audio has two possible paths out of the Commodore, and they are not close in quality.

The $D418 DAC works on every machine. It abuses the SID's volume register as a crude digital-to-analog converter, which is how digitized sound was done on the C64 in period. Written the obvious way that gives four bits, and c64cast can still do exactly that. By default it does something better: it parks all three voices as steady sources and writes the whole register each sample — volume, filter mode, and the voice-three switch — which lands on roughly 256 distinct output levels. They are unevenly spaced, so the useful resolution is nearer six or seven bits than eight, but it is a long way past four.

It still sounds rough, because it is still a volume register being wobbled twelve thousand times a second, and that roughness is most of the charm. It is the only path available on a TeensyROM, and the path used for microphone and webcam audio everywhere.

The Ultimate Audio sampler is available on the C64U. It is a proper PCM sampler in the FPGA, entirely off the C64's bus, and it sounds like a normal audio device.

[audio]
enabled = true
backend = "auto"

Left on auto, c64cast uses the sampler when the hardware has one and falls back to the DAC otherwise. Set backend = "dac" to force the rough path deliberately, which for some material is the more appropriate choice.

You now know how the picture is made. The last chapter is about doing larger things with it.

Edit this page on GitHub