HM3 — style directions for options 01 and 03
What these are. These are surface studies rather than logo files. Each render was driven image-to-image from an exact PNG of the vector mark with instructions to change material and light only. A diffusion model still redraws pixels, so treat every panel as a direction to hand-execute, never as artwork to ship.
How to judge. Compare each render against the control strip above it. The tag on each card says whether the letterforms survived: clean means the drawing held, usable means it drifted in ways a hand rebuild fixes, and drifted means the model re-lettered the mark and the panel is only a mood reference.
Palette. Every prompt banned gold, brass, bronze, copper and amber. Where a render still shows warmth, that is the model leaking, and it is called out on the card.
1.1Exponent
Pure type. The 3 set as a true superscript. Nothing here changes the drawing; every panel is the same outline under a different surface.
Control: the vector mark the five renders were drawn from. Compare letterforms against this.
4.1Brushed platinum clean
Machined aluminium with a directional grain, floating off a dark wall. The physical-object read: reception signage, a stamped plate, an awards wall.
Letterforms survived intact, including the superscript. Cool grey throughout with no warm leak. The weakness is generic: raised metal type on a dark wall is the default mockup of every B2B firm alive, so it flatters the mark without saying anything about television.
4.2CRT phosphor clean
Emission on a tube monitor, scanlines and bloom. The most literal claim to broadcast in the set, and the one that dates the company on purpose.
Drawing held, but two problems follow. Bloom fattens the strokes and closes the superscript, which is the exact failure the mark already has at small size, so this treatment amplifies its weakest point. The tube vignette also eats most of the frame, leaving a crop rather than a whole usable image.
4.3Extruded graphite clean
Solid mass under raked light with an architectural shadow. Buys permanence and weight, which is the one thing a forty-year-old shop can claim that a new one cannot.
The most faithful render of the ten: spacing, vertex and superscript all match the vector. The 3 rides marginally low against the control. Depth this deep only works at hero scale; at nav size the extrusion becomes noise.
4.4Smoked glass usable
Dark glass picked out by cool edge light. The coldest and most expensive-looking panel here.
Handsome and completely impractical. The mark exists only as highlights on its own edges, so it has no fill at all: it dies against any ground that is not near-black, cannot be printed, cannot be embroidered, and fails the one-ink test outright. Treat it as a hero animation frame, never as the logo.
4.5Letterpress deboss clean
Pressed into heavy uncoated stock. The print case: cards, folders, the deck that arrives on paper rather than as a link.
Impression is convincing and the stock stayed neutral grey rather than drifting cream. The superscript is already starting to fill at this size, which tells you the real card would need the 3 opened up or set larger than the screen version.
1.3Tally
The chip. A cap-height square with the numeral knocked out, the way a network drops a bug in the corner of frame.
Control: the vector mark the five renders were drawn from. Compare letterforms against this.
4.6Anodised badge usable
Milled letters, bevelled chip, numeral cut clean through to the wall behind. The chip as hardware.
Material direction is right and the knockout reads exactly as intended, with a bevel on the cut edge worth keeping. The type drifted though: the M vertex drops nearly to the baseline where Cabinet 900 stops well short, and the H widened. Rebuild from the vector and keep only the surface.
4.7On-air tally clean
Chip as an illuminated panel on a studio camera body, HM in matte grey beside it. Puts the mark where the business actually happens.
The strongest argument in the set. Drawing held, the knockout reads at a glance, and the lit chip explains the whole logic of the mark without a caption. Two cautions: it only works in a dark studio frame, so it is an art-direction system rather than a logo, and the chip drifted very slightly off square.
4.8Moulded tag usable
Soft-touch dark polymer body with the HM debossed and the chip inlaid in brushed metal. Product-hardware register rather than broadcast.
Useful mainly as a warning. Debossing HM into dark polymer leaves it dark-on-dark with almost no contrast, so the mark half disappears and the inlaid chip carries the whole thing. It also stretched the chip off square and deepened the M vertex. The lesson is that the HM needs tonal separation rather than texture alone.
4.9Screen bug usable
Chip composited into the corner of dark footage at partial opacity. This is the job the mark was drawn for.
The frame came out almost entirely black because the prompt asked for dark grainy footage, so the image is close to useless as a picture. What it does prove is the point that matters: at true bug scale, in the corner, at reduced opacity and through video grain, the chip still reads. None of the other four treatments survive that test.
4.10Embroidery usable
Stitched on black cotton with the numeral left as bare fabric. The merch and one-ink durability test.
Passes the one-ink test, which is what it was for, and leaving the 3 unstitched is the correct construction. But thread bulk closed the gap between the M and the chip so the whole mark reads as one continuous slab. Production needs a wider gap between letters and chip plus a stated minimum size.