screengraft 0.39.0 → 0.40.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "screengraft",
-  "version": "0.39.0",
+  "version": "0.40.0",
   "description": "Put a UI screenshot or screen recording onto a photographed device screen with the perspective exactly right \u2014 a homography you confirm by hand, not a generative guess.",
   "keywords": [
     "mockup",

package/scripts/detect.py

@@ -797,12 +797,31 @@ def detect_edges(gray: np.ndarray, click=None, trace=None):
     # Sweep the Canny thresholds off the image's own median rather than fixed
     # numbers, then a few sigmas around it — one exposure doesn't suit both a
     # bright render and a dim photo.
+    #
+    # ... and ALSO off Otsu's threshold, because the median anchor has a hole
+    # the median cannot see: a dark photograph. A black phone on a black
+    # backdrop has a median of 0..4, so every sweep point lands at hi <= 7 and
+    # Canny fires on every pixel of noise -- the edge map is a solid sheet and
+    # no closed quad survives it. Both photographs in the labelled corpus on
+    # which nothing near the screen was EVER proposed (11 Sep 2026) are exactly
+    # this, and edge returned zero candidates on them. Otsu splits the two
+    # modes that are actually there -- 97 and 127 on those two -- and the same
+    # sweep then proposes the screen at 9% and 12% unrefined, which is where
+    # every other photograph's nearest candidate sits. On the other seven the
+    # nearest candidate is unchanged. The saturation channel already anchors
+    # on Otsu for the same reason (one percentile fails when the thing sought
+    # is smaller than the percentile).
     med = float(np.median(blur))
-    for sigma in (0.20, 0.33, 0.50, 0.66):
-        lo = int(max(0, (1.0 - sigma) * med))
-        hi = int(min(255, (1.0 + sigma) * med))
-        if hi <= lo:
+    otsu, _ = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
+    sweep = [((1.0 - s) * med, (1.0 + s) * med) for s in (0.20, 0.33, 0.50, 0.66)]
+    sweep += [(otsu * f / 2.0, otsu * f) for f in (0.5, 1.0, 1.5)]
+    seen = set()
+    for lo_f, hi_f in sweep:
+        lo = int(max(0, lo_f))
+        hi = int(min(255, hi_f))
+        if hi <= lo or (lo, hi) in seen:
             continue
+        seen.add((lo, hi))
         edges = cv2.Canny(blur, lo, hi, L2gradient=True)
         # Close small gaps so a bezel outline broken by a notch or a glare
         # spot still forms one closed contour.

package/skills/inject-screenshot/SKILL.md

@@ -5,7 +5,7 @@ description: Injects a UI screenshot OR a screen recording onto a photographed d
 
 # Inject a screenshot onto a photographed device
 
-**What ships (v0.39):** a local browser UI (`scripts/ui.py`) that walks the designer through the whole job — pick the photo and the screen source, which may be an image **or a video** (recent Desktop/Downloads images, drag-drop, browse, path, or a **Figma frame link**), auto-detect the screen as a starting position — and when detection cannot tell which region is a screen, **Point at screen**: one click inside it and the detector uses that point — or, if this photograph has been fitted before, **the fit it was saved with comes back** as the starting position instead of a detection, recognised by the photo's own pixels so a rename or a drag-drop still match — and every save also writes a **portable `.fit.json` beside the mockup** that can be dropped back onto the page later, which is how a fit survives a re-export, another machine, or someone else's hands — then **match the four edges** (drag an edge's middle to slide it, near an end to pivot; corners still draggable) with canvas navigation that follows the usual conventions — **hold ⌘ and scroll to zoom to the pointer, hold space and drag to pan** — and a rectified strip loupe. The fit and the composite sit **side by side and always have** — the result pane re-renders as you drag, which is how a corner gets judged, so it is the layout rather than a mode you can switch off. Then an on-by-default realism pass that colour-matches the source to the photo's light, **Save** (or **Render**, for a video) into the project folder (`--out-dir`), and a **Send to Claude** button that reaches you through the plugin's own MCP server. The UI is a hand port of the project's Figma design file — dark only.
+**What ships (v0.40):** a local browser UI (`scripts/ui.py`) that walks the designer through the whole job — pick the photo and the screen source, which may be an image **or a video** (recent Desktop/Downloads images, drag-drop, browse, path, or a **Figma frame link**), auto-detect the screen as a starting position — and when detection cannot tell which region is a screen, **Point at screen**: one click inside it and the detector uses that point — or, if this photograph has been fitted before, **the fit it was saved with comes back** as the starting position instead of a detection, recognised by the photo's own pixels so a rename or a drag-drop still match — and every save also writes a **portable `.fit.json` beside the mockup** that can be dropped back onto the page later, which is how a fit survives a re-export, another machine, or someone else's hands — then **match the four edges** (drag an edge's middle to slide it, near an end to pivot; corners still draggable) with canvas navigation that follows the usual conventions — **hold ⌘ and scroll to zoom to the pointer, hold space and drag to pan** — and a rectified strip loupe. The fit and the composite sit **side by side and always have** — the result pane re-renders as you drag, which is how a corner gets judged, so it is the layout rather than a mode you can switch off. Then an on-by-default realism pass that colour-matches the source to the photo's light, **Save** (or **Render**, for a video) into the project folder (`--out-dir`), and a **Send to Claude** button that reaches you through the plugin's own MCP server. The UI is a hand port of the project's Figma design file — dark only.
 
 The geometry is exact (`warp.py`); the detection is advisory (`detect.py`) and the human corrects it. **When a detection is wrong and you want to know why**, ask for the candidate list: `POST /api/detect {"trace": true}` writes `<session>/candidates.json`, or run `python3 scripts/detect.py --photo P --out-corners /tmp/c.json --trace /tmp/t.json` (add `--click X,Y`). Every candidate quad is in there with its score and whether it was accepted, rejected, never reached, or filtered out by the click — which is what separates "the screen was never proposed" from "it was proposed and something else won".