Draw 4 different toad personas for moneycanbuy.fun roles in 4 variants each: four DIFFERENT toad characters, one per role, each in 4 different variants so we can pick one, for the How it works page of our site (every role is a wide row with its character on the left). moneycanbuy.fun is a board of paid dares: an AUTHOR names a dare and its reward and locks the reward in a contract, a DOER does the dare and posts proof, NFT JUDGES (holders of identity.md NFTs) vote LEGIT or FAKE when a proof is disputed, PREDICTORS bet WILL DO or WON'T DO on whether the dare gets done.
REPOSITORY: use only brand/mark.png (our toad: STYLE reference, do not paste or trace it), brand/colors.md (palette), toad-head/front.png and toad-head/three-quarter.png (style of a toad figure), explainers/imd/nft-1.png (shape of the judge's badge only). Ignore every other file and folder: they belong to other jobs.
ONE FAMILY, FOUR DIFFERENT TOADS: the four roles are four different toads, as different as four people with four jobs. Each has its own face, body build, age, headwear or hair, expression and shade of green. What they share is only the drawing style of brand/mark.png: the same line weight, thick dark outline #09210B, flat fills with one shadow tone, no gradients, darker spots, cream #FFF0CF eyes, recognizably from the same world as our toad. Greens only from the palette: #82B63E, #599A35, #44842A, each toad with a different main green and spot mix. Never a generic cartoon frog, never the look of a real or famous character.
NO PRICE TAGS: no price tag, no "$" tag, no chain with a tag, no "$" sign anywhere on any persona.
ROLES (who, what we must see):
author: the toad who announces a dare and sets its reward himself. One hand points forward or holds up a blank rolled scroll (the dare, no letters) or a megaphone; the other locks gold coins in a clear transparent box with a closed padlock. Hands and props anatomically right: a hand holding the box is behind its front wall or under it, never drawn on top of the glass. Confident, in charge.
doer: the daredevil who does the dare and films himself. Dynamic action pose, holds a phone turned toward his own face (selfie filming), the screen a blank plain glow. His outfit must not melt into black: light accents (light hood lining, bright sweatband, light sneakers or stripes) so the silhouette reads on #000000. Bold, fearless, younger.
judge: the NFT holder who votes LEGIT or FAKE. Judge's robe, wooden gavel, and a badge on the robe that clearly reads as the identity.md NFT face from explainers/imd/nft-1.png: a tall faceted mask with a pointed top and chin, two angled eye slits and one bright white eye ring in the middle, green outline on one side, violet on the other. Nothing else from that image (no words, bars, rings around it). Calm, serious, older.
predictor: the toad who bets WILL DO or WON'T DO. Holds a blank betting slip (only plain lines) and a real stack of at least 6 gold coins; beside him a small floating dark panel #0D120E with one rising green #3ECF7A line and one falling red-pink #F43F5E line, no axes, no numbers. Sly, excited.
4 VARIANTS PER ROLE: -v1 to -v4 are four genuinely different designs of that role: different pose or camera angle, different character design (face, outfit, headwear, props placement). Not the same drawing recolored or mirrored. All 16 keep the role's must-see props and stay in the one family style.
FRAMING: waist-up or three-quarter figure, centered, at least 8% empty margin on every side, nothing cut off. Each file must read at 96 px: clean silhouette on #000000 and on the dark card #0D120E, the prop recognizable. Colors from brand/colors.md plus neutral tones; gold is money, green #3ECF7A is WILL DO and LEGIT, red-pink #F43F5E is WON'T DO and FAKE.
FORMAT: artifacts//-v1.png to -v4.png, each exactly 1024x1024, PNG RGBA with a TRANSPARENT BACKGROUND: alpha 0 everywhere outside the figure, no checkerboard, no box, no ground shadow, no halo or fringe. Cut the background to true alpha after generation. Any tool is fine: an image model with the repository files as reference, or vector drawing in code; no photo, no 3D.
NO TEXT: no letters, numbers, words, logos, prices, amounts, wallet addresses, QR codes, UI or watermarks anywhere in the 16 persona files: phone screen, scroll, slip, panel, box, coins, robe and badge stay blank.
CONTACT SHEET: artifacts/contact-sheet.png, all 16 files in a 4x4 grid on #000000, one row per role in the order author, doer, judge, predictor, columns v1 to v4; file names as labels only here.
README (required, the job is not done without it): artifacts/README.md, one row per file: tool, model name and version, the exact prompt (or drawing method), reference files used, every rejected and regenerated file and why. State that moneycanbuy.fun may use, change and publish the files without attribution.
REJECTION LIST, check every file, regenerate anything that fails, one line per item in README.md:
- Any letter, number, word, logo, price, address or UI.
- A price tag or a "$" tag or sign on any persona.
- Two personas of different roles with the same face or the same body.
- Broken anatomy of hands or props: a hand on top of the glass, extra or missing fingers, a prop floating out of the hand.
- Variants of one role that differ only by color or mirroring.
- Not in the family style: other line weight, gradients, generic frog, greens outside the palette, photorealism, 3D look, the look of a real or famous character.
- A missing role prop: author without the pointing hand or blank scroll or megaphone and the clear locked coin box; doer without the selfie phone and light accents; judge without robe, gavel and a badge that reads as the NFT face; predictor without the slip, the coin stack of 6 or more and the panel.
- Background not true alpha, checkerboard, halo, ground shadow; not exactly 1024x1024; figure cut off or blurry.
- README.md not delivered.
FILES, exactly these names under artifacts/: author/author-v1.png, author/author-v2.png, author/author-v3.png, author/author-v4.png, doer/doer-v1.png, doer/doer-v2.png, doer/doer-v3.png, doer/doer-v4.png, judge/judge-v1.png, judge/judge-v2.png, judge/judge-v3.png, judge/judge-v4.png, predictor/predictor-v1.png, predictor/predictor-v2.png, predictor/predictor-v3.png, predictor/predictor-v4.png, contact-sheet.png, README.md.
