The clusters could describe a character and stat a monster; now they can show you both. Two new text-to-image endpoints render the picture an agent's been building toward.

  • character-portrait ($0.08, rollforge) — a portrait from structured inputs: race, class, gender, a free-text description, and an art style. Pairs with character-gen — roll the character, then render its face.
  • creature-image ($0.08, bestiary) — art of any creature, named from myth, folklore, a novel, a film, a game, or your own words. Pairs with the bestiary lookups — read the lore or stat block, then picture the monster.

Two real renders, straight from the endpoints — the input is right there in the caption:

creature-image — {
creature-image — { "creature": "Wendigo", "style": "fantasy-art" }
character-portrait — {
character-portrait — { "race": "Elf", "class": "Ranger", "description": "green cloak, longbow, scar over one eye" }

Both take an art style — fantasy-art (default), realistic, anime, oil-painting, ink-sketch, watercolor, pixel-art, or comic-book — and return a ready-to-use PNG URL.

With these, a single agent can run the whole arc: generate a character and its portrait, build an encounter, pull a creature's lore, stat it, picture it, and resolve the fight — every step a per-call USDC settlement. rollforge is now 26 endpoints and bestiary is 6.