pixelkiln 0.43.0 → 0.45.0

package/docs/MANIFEST.md

@@ -59,6 +59,8 @@ not merely a label edit.
 | `mirror` | string | Another asset of the same style flipped left to right, made locally at no cost. See [Mirrors](#mirrors). |
 | `proportions` | preset or object | `character` styles: this base's proportions, over the style's. |
 | `reference` | string or object | `character` styles, bases: the character's own south-facing sprite (or `{ "south": ..., "east": ... }`), which PixelLab rotates instead of drawing from the prompt. See [Characters](#characters). |
+| `concept` | string | `character` styles, `pro` bases: a manifest-relative concept image (up to 1024px) the design is seeded from. |
+| `styleCharacter` | string | `character` styles, `pro` bases: this base's style anchor, over the style's. |
 | `providerInputs` | JSON scalar/sequence map, `{}` | Named per-asset inputs consumed by the active provider. ComfyUI accepts scalars and, for `frames`, one ordered 2–64 value sequence. Image bindings upload PNG/JPEG inputs. |
 | `styles` | string array, `[]` | Restrict the asset to named styles; empty means every style. |
 | `promptByStyle` | string map, `{}` | Replace only the asset prompt for a named style. |
@@ -105,6 +107,8 @@ constant across the set.
 | `proportions` | | `character` only, `standard` humanoid bases. A preset (`default`, `chibi`, `cartoon`, `stylized`, `realistic_male`, `realistic_female`, `heroic`) or multipliers `{ headSize, armsLength, legsLength, shoulderWidth, hipWidth }`, each 0.5 to 2. An asset may override it. |
 | `textGuidanceScale` | `8` | `character` only. How closely a `standard` base and a template loop follow their text, 1 to 20. |
 | `isometric` | `false` | `character` only. Draw `standard` bases and every loop in isometric view. |
+| `enhancePrompt` | `false` | `character` only, `v3` bases. Let PixelLab expand the prompt into a fuller one before drawing. |
+| `styleCharacter` | | `character` only, `pro` bases. The asset id of a generated 8-direction character in this style whose look bases drawn from text or a concept follow. An asset may override it. |
 | `mount` | object | Stable-cell sheet placement; documented below. |
 | `quality` | object | Optional native-grid, final-palette, and human-approval contract; documented below. |
 | `tags` | string array, `[]` | Tags inherited by every generated provider object in the style. |
@@ -524,6 +528,8 @@ credentials, cost semantics, recovery, and the paid live-test boundary.
 | `mirror` | string | Another asset of this style, flipped left to right. No prompt, no provider, no cost. See [Mirrors](#mirrors). |
 | `proportions` | preset or object | This base's body proportions, over the style's. `standard` humanoid bases only. |
 | `reference` | string or object | The character's own sprite(s) for PixelLab to rotate. Bases only. |
+| `concept` | string | A concept image a `pro` base is designed from. Bases only. |
+| `styleCharacter` | string | The character in this style a `pro` base takes its look from, over the style's. |
 
 ## Characters
 
@@ -584,6 +590,8 @@ override it, and only the mannequin template has proportions to set.
 is followed, and applies to template loops as well. `isometric` draws the
 base and every loop in isometric view. The v3 and pro engines take none of
 the three for a base; a v3 style may still set the last two for its loops.
+A `v3` base takes `enhancePrompt` instead, which has PixelLab expand the
+prompt into a fuller one before drawing.
 
 A base can also start from your own sprite. `reference` names a
 manifest-relative PNG or JPEG of the character facing south, and PixelLab
@@ -605,11 +613,27 @@ is read again at submit time and refused if it changed since `plan`.
 States, animations, and mirrors take their look from their parent and
 cannot carry a reference.
 
+The pro engine has two more ways in. `concept` on a base names a concept
+image (a painting, a photo, a sketch, up to 1024px) that seeds the design
+in place of the prompt alone (PixelLab's `create_from_concept`); the
+prompt still guides it. `styleCharacter`, on the style or a base, names a
+generated 8-direction character in the same style whose look the base
+follows (`style_character_id`); its south sprite becomes the style image
+unless a style image is also given. The anchor is a dependency like a
+loop's character: the base is `blocked` until the anchor is downloaded and
+current, `gen` runs it in the wave after the anchor lands, and a
+regenerated anchor makes every base drawn in its style `stale`. The anchor
+itself ignores the setting, so a style can name its own first character.
+PixelLab wants the base at least as large as the anchor's visible sprite
+and fails the job fast otherwise. A base rotating its own `reference`
+takes neither: the rotate method has one image slot and it is the
+character.
+
 Style images on a `character` style are a pro-only input: one image of up
-to 168px that anchors the look of a `pro` base drawn from text (PixelLab's
-`create_with_style`). `standard` and `v3` have no such slot and refuse
-them; a pro base with a `reference` refuses them too, since the rotate
-method has one image slot and it is the character.
+to 168px that anchors the look of a `pro` base drawn from text or a
+concept (`create_with_style` and `create_from_concept` both take it).
+`standard` and `v3` have no such slot and refuse them; a pro base with a
+`reference` refuses them too.
 
 A state is a text edit of an existing character, `state.of`, applied to
 every direction at once: a pose, an outfit, a held object. The prompt is
@@ -634,6 +658,30 @@ the loop and accepts or rejects it whole. One asset per direction; declare
 another asset for another direction, or a [mirror](#mirrors) of this one
 for the direction that faces the other way.
 
+A v3 loop can start and end where you say. `startFrame` is a
+manifest-relative image of the pose to begin from instead of the
+character's rotation; `endFrame` is a pose to reach, and with it the loop
+interpolates from the start frame to that image (both up to 256px, and the
+end frame the same size as the start frame or, without one, as the
+character's rotation, which `plan` checks once the parent is on disk; the
+frames can still come back on a taller canvas when the motion needs it, as
+a 92px crouch did at 92×104). `subject` replaces the character's own description for this loop when it
+would mislead the model (a state that took the armour off), and
+`enhancePrompt` lets PixelLab expand the action into a fuller motion
+description first. A template loop takes none of those; it takes
+`outline`, `shading`, and `detail` overrides instead, over the character's
+own. The style's `palette` goes to every loop as a colour reference, as it
+does to a `standard` base, and `enforcePalette` still snaps the frames
+afterwards. Pose images hash into the loop's identity and are read again
+at submit time, like a base's `reference`.
+
+```json
+"mira.bow": {
+  "prompt": "bowing deeply from the waist",
+  "animation": { "of": "mira", "direction": "south", "frames": 6, "endFrame": "poses/mira-bowed.png" }
+}
+```
+
 States and animations depend on their parent the way a revision does. The
 parent must be downloaded and current before the child is actionable;
 `plan` reports the child as `blocked` and names the parent until then, and

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pixelkiln",
-  "version": "0.43.0",
+  "version": "0.45.0",
   "description": "Manifest-driven pixel-art generation, review, recovery, and packaging with deterministic provenance.",
   "type": "module",
   "sideEffects": false,

package/schema/manifest.schema.json

@@ -206,6 +206,13 @@
                   "isometric": {
                     "type": "boolean"
                   },
+                  "enhancePrompt": {
+                    "type": "boolean"
+                  },
+                  "styleCharacter": {
+                    "type": "string",
+                    "minLength": 1
+                  },
                   "seed": {
                     "type": "integer"
                   },
@@ -499,6 +506,13 @@
                   "isometric": {
                     "type": "boolean"
                   },
+                  "enhancePrompt": {
+                    "type": "boolean"
+                  },
+                  "styleCharacter": {
+                    "type": "string",
+                    "minLength": 1
+                  },
                   "seed": {
                     "type": "integer"
                   },
@@ -788,6 +802,33 @@
                       "v3",
                       "pro"
                     ]
+                  },
+                  "startFrame": {
+                    "type": "string",
+                    "minLength": 1
+                  },
+                  "endFrame": {
+                    "type": "string",
+                    "minLength": 1
+                  },
+                  "subject": {
+                    "type": "string",
+                    "minLength": 1
+                  },
+                  "outline": {
+                    "type": "string",
+                    "minLength": 1
+                  },
+                  "shading": {
+                    "type": "string",
+                    "minLength": 1
+                  },
+                  "detail": {
+                    "type": "string",
+                    "minLength": 1
+                  },
+                  "enhancePrompt": {
+                    "type": "boolean"
                   }
                 },
                 "required": [
@@ -869,6 +910,14 @@
                   }
                 ]
               },
+              "concept": {
+                "type": "string",
+                "minLength": 1
+              },
+              "styleCharacter": {
+                "type": "string",
+                "minLength": 1
+              },
               "mirror": {
                 "type": "string",
                 "minLength": 1

package/schema/recipe.schema.json

@@ -281,6 +281,13 @@
             "isometric": {
               "type": "boolean"
             },
+            "enhancePrompt": {
+              "type": "boolean"
+            },
+            "styleCharacter": {
+              "type": "string",
+              "minLength": 1
+            },
             "seed": {
               "type": "integer"
             },

package/skills/pixelkiln/references/pixellab.md

@@ -27,7 +27,15 @@ providers, or before any PixelLab account operation.
 A base can start from the author's own south-facing sprite (`reference` on
 the asset; standard wants it at the style's size, v3 up to 256px, pro up
 to 168px), and a standard humanoid base takes `proportions` (`chibi`,
-`heroic`, or multipliers) on the style or the asset.
+`heroic`, or multipliers) on the style or the asset. A pro base can be
+designed from a `concept` image (up to 1024px) and follow the look of
+another generated character in the style (`styleCharacter`), which makes
+that character a dependency the way a loop's parent is.
+
+A v3 loop can start from a pose image (`startFrame`) or interpolate to one
+(`endFrame`), take a `subject` when the character's own description would
+mislead, and ask PixelLab to `enhancePrompt`; a template loop takes
+`outline`, `shading`, and `detail` overrides instead.
 
 A loop costs per direction, and a sprite facing east is the sprite facing
 west flipped. Declare the west loop and make the east one `{ "mirror":