# Quarter Circle > A fun character maker: type a real person's harmless quirks (a friend, a relative, a colleague, yourself), one per line, pick a fighting style and a tone, and get a made-up fighting-game character - a name card, five stats, one special move per quirk and a super, each with a joystick input in numpad notation and frame data. A second run balance-checks the move list and patches any move that is too safe, too quick, too strong or too slow. It adds nothing about the person the quirks do not say, never mocks looks, health or identity, and names no real game, character or studio. URL: https://quarter-circle.skillsafe.ai/ API tutorial: https://quarter-circle.skillsafe.ai/api.html ## Free, in the browser - Quirk reading: each quirk is read by its words as a grab (hugs, carries, lifts), a projectile (throws, texts, shouts, sends), a rush (runs, late, sprints, fast), a comeback (technically, corrects, sighs) or a wild card, and reserved a unique input and button: 236 (quarter-circle forward), 214, 623, 41236, 63214, [4]6 (hold back, then forward), [2]8, 22, 63214789 (full circle); the super uses 236236 or 214214; P is punch, K is kick. - The select card: a portrait tile, the name, the style, and stat bars (power, speed, defense, range, combo; 1 to 5 each, 15 in all; draft values by style until a forge replaces them). - The move list: a joystick glyph per input and a training-mode frame meter per move (one cell per frame at 60 a second: startup, active, recovery). The draft's frame numbers are the page's placeholders, not measurements - all frame data here is invented for a made-up character. - Editing and flags: edit any frame number; the page flags a move that is too safe (on block +3 or more, not a grab), too quick (startup 4 or less with 100 damage or more), too strong (200 damage, a super 400) or too slow (startup 30 or 70 frames in all; a super at startup 20), and names the call each move earns (nerf, buff, keep) and the verdict the rule gives. - Exports: a card image (PNG, 1080 x 1350: the select card, the move list and the last verdict - shared straight to an app where the browser can), plain text for a group chat (one line per move), Markdown, a frame-data CSV (with flags, the page's call and any patch), JSON, print; a roster of kept fighters (in the browser, or in your account when signed in). ## Metered (gpt-terra, two tasks) - `forge`: a title; the fighter's name (yours, or a nickname from the quirks), epithet, style, bio, intro and win lines; stats that add up to 15 and fit the style; one special per quirk (up to four) copying the page's reserved input, then a super; every move with startup, active, recovery, on-block advantage, damage, what it does and its tell, all inside fair ranges (special: startup 5-28, active 1-8, recovery 8-40, on block -20 to +2, damage 60-180, 65 frames at most; super: startup 5-15, active 1-10, recovery 20-60, on block -40 to -10, damage 250-380; a grab is 0 on block); and a weakness. - `tune`: per move, the call from the page's flags (keep, nerf, buff), a patch (keep copies the numbers; a nerf or buff changes as few as needed to clear every flag), why, the tell and the counterplay that matches the patched on-block value; a game plan as and against the fighter. Verdict by a fixed rule, checked in order: broken (two or more nerfs), needs_patch (any nerf or buff), tournament_ready. - The page re-checks every reply: the name and style, the stat total and style rule, the number of specials and the super, each special's quirk and input, the fair ranges and flags, the verdict and every call against the rule, keep patches unchanged and nerf or buff patches clearing every flag, and no he/she, internal ids, real game names or stray numbers in the prose. ## Sources - Built on the SkillSafe skills @thedivergentai/godot-combat-system (https://skillsafe.ai/skill/@thedivergentai/godot-combat-system): combat systems - damage data, special moves and combos, startup, active and recovery frames, invincibility frames, combat state machines - for the forge; and @mengto/design-action-combat (https://skillsafe.ai/skill/@mengto/design-action-combat): readable action combat - attack timing, guard windows, telegraphs and fair counterplay - for the tune.