fix: shrink content-aware output in the right direction

README.md

@@ -113,21 +113,21 @@ previews, and chain in either order without extra dependencies.
 `collapse` lives under **SHAPE & MOTION**. The default `seam` mode behaves like
 Photoshop content-aware scaling: each frame starts at its original composition,
 then its content-aware shrink grows linearly from the first frame to the clip's
-end. Low-energy seams are removed from a reduced analysis frame, and the carved
-coordinate map is stretched back over the original canvas, so output dimensions
-never change. High-gradient regions resist shrinking. Analysis is capped at 192
-pixels on the longest side and updates at most five times per second; the
-displacement field is smoothed between updates. `mode=elastic` keeps the earlier
-seeded, spatially nonuniform elastic warp available as an alternative.
+end. Low-energy seams are removed from a reduced analysis frame. The renderer
+inverts that seam-removal displacement so the content shrinks, then maps it to
+the original canvas dimensions. High-gradient regions resist shrinking.
+Analysis is capped at 192 pixels on the longest side and updates at most five
+times per second; the displacement field is smoothed between updates.
+`mode=elastic` keeps the earlier seeded, spatially nonuniform elastic warp.
 
 | option | range / values | default |
 | --- | --- | --- |
-| `intensity` | 0-100%, overall scale strength (elastic warp strength in elastic mode) | 65% |
-| `horizontal` | 0-100%, share of the maximum width shrink / horizontal warp | 65% |
-| `vertical` | 0-100%, share of the maximum height shrink / vertical warp | 45% |
+| `intensity` | 0-100%, overall scale strength (elastic warp strength in elastic mode) | 100% |
+| `horizontal` | 0-100%, share of the maximum width shrink / horizontal warp | 100% |
+| `vertical` | 0-100%, share of the maximum height shrink / vertical warp | 100% |
 | `preserve` | 0-100%, content/edge protection | 75% |
 | `mode` | seam, elastic | seam |
-| `seam_strength` | 0-100%, maximum content-aware shrink by clip end | 60% |
+| `seam_strength` | 0-100%, maximum content-aware shrink by clip end | 100% |
 | `motion` | 0-100%, elastic motion influence | 50% |
 | `smoothness` | 0-100%, spatial field smoothness | 70% |
 | `speed` | 0-4 elastic cycles per second | 0.35 |
@@ -161,18 +161,18 @@ The report measures processor frame time and peak process RSS at 640x360 and
 end-to-end playback guarantee. Seam mode also reports carving-update time.
 
 Recorded on this Linux x86_64 host with one OpenCV CPU thread, 120 frames per
-case, Python 3.14, NumPy 2.5.4 and OpenCV 5.0 (raw data in
+case, Python 3.13.5, NumPy 2.2.4 and OpenCV 4.12.0 (raw data in
 `benchmarks/collapse.json`):
 
 | mode | frame size | mean ms/frame | p95 ms/frame | peak process RSS |
 | --- | --- | --- | --- | --- |
-| elastic | 640x360 | 4.55 | 7.02 | 72.8 MiB |
-| elastic | 1280x720 | 11.57 | 12.30 | 119.1 MiB |
-| seam | 640x360 | 3.43 | 6.37 | 68.1 MiB |
-| seam | 1280x720 | 9.72 | 13.20 | 119.1 MiB |
+| elastic | 640x360 | 4.23 | 4.57 | 63.3 MiB |
+| elastic | 1280x720 | 9.79 | 10.60 | 110.1 MiB |
+| seam | 640x360 | 16.59 | 89.64 | 62.6 MiB |
+| seam | 1280x720 | 25.31 | 123.05 | 109.0 MiB |
 
 Seam averages include frames reusing a cached map; carving-update frames alone
-averaged 6.68 ms at 360p and 12.87 ms at 720p. Persistent state was about
+averaged 89.37 ms at 360p and 109.63 ms at 720p. Persistent state was about
 0.38 MiB for elastic and 0.53 MiB for seam mode in these cases. These are
 measurements of this fixture on this host, not a promise for every input.
 

ROADMAP.md

@@ -380,10 +380,11 @@ Shared rules for the stateful effects:
   - Test linear progress against clip duration, fixed output size, content preservation, repeatability, bounded state and real CLI/menu renders.
   _Done: `collapse` defaults to a content-aware seam scale that begins at the
   source composition and grows linearly to its target over probed clip duration.
-  The low-resolution energy map removes seams from low-detail regions, then the
-  source-coordinate map is stretched back to the original frame dimensions.
-  Analysis is capped at 192px; shrink is limited to 50% per axis and refreshed
-  at <=5Hz with temporal smoothing. The seeded elastic warp remains selectable.
+  The low-resolution energy map removes seams from low-detail regions, and the
+  renderer inverts that displacement so salient content shrinks instead of
+  growing. Each output is mapped to the original frame dimensions. Analysis is
+  capped at 192px; shrink is limited to 50% per axis and refreshed at <=5Hz with
+  temporal smoothing. The seeded elastic warp remains selectable.
   Duration flows through the shared stateful adapter. Tests cover exact start,
   linear midpoint field, end-frame deformation, output shape/audio, resets,
   preservation, optional dependencies and TUI/CLI integration. Optional headless

benchmarks/collapse.json

@@ -1,5 +1,5 @@
 {
-  "recorded_utc": "2026-10-23T22:29:41.163465+00:00",
+  "recorded_utc": "2026-10-25T14:16:27.765814+00:00",
   "platform": "Linux-7.1.13+parrot7-amd64-x86_64-with-glibc2.41",
   "logical_cpus": 22,
   "scope": "synthetic moving RGB; process() only; fresh process per mode/size; not end-to-end render FPS",
@@ -10,34 +10,34 @@
       "height": 360,
       "frames": 120,
       "config": {
-        "intensity": 65,
-        "horizontal": 65,
-        "vertical": 45,
+        "intensity": 100,
+        "horizontal": 100,
+        "vertical": 100,
         "preserve": 75,
         "mode": "elastic",
-        "seam_strength": 60,
+        "seam_strength": 100,
         "motion": 50,
         "smoothness": 70,
         "speed": 0.35,
         "coherence": 80,
         "seed": 17
       },
-      "mean_ms": 4.550746899894875,
-      "p50_ms": 4.366609497083118,
-      "p95_ms": 7.016759949146943,
-      "max_ms": 11.28179499937687,
-      "processor_fps": 219.74414793824297,
+      "mean_ms": 4.230598408328963,
+      "p50_ms": 4.365750500255672,
+      "p95_ms": 4.567256800373798,
+      "max_ms": 5.016768999666965,
+      "processor_fps": 236.37318021754476,
       "seam_updates": 0,
       "seam_update_mean_ms": null,
-      "peak_rss_mib": 72.765625,
+      "peak_rss_mib": 63.31640625,
       "persistent_state_mib": 0.375762939453125,
       "work_shape": [
         108,
         192
       ],
-      "python": "3.14.0",
-      "numpy": "2.5.4",
-      "opencv": "5.0.0",
+      "python": "3.13.5",
+      "numpy": "2.2.4",
+      "opencv": "4.12.0",
       "opencv_threads": 1
     },
     {
@@ -46,34 +46,34 @@
       "height": 720,
       "frames": 120,
       "config": {
-        "intensity": 65,
-        "horizontal": 65,
-        "vertical": 45,
+        "intensity": 100,
+        "horizontal": 100,
+        "vertical": 100,
         "preserve": 75,
         "mode": "elastic",
-        "seam_strength": 60,
+        "seam_strength": 100,
         "motion": 50,
         "smoothness": 70,
         "speed": 0.35,
         "coherence": 80,
         "seed": 17
       },
-      "mean_ms": 11.565306316727705,
-      "p50_ms": 11.400867499105516,
-      "p95_ms": 12.299060149962315,
-      "max_ms": 18.616312998346984,
-      "processor_fps": 86.46550057680967,
+      "mean_ms": 9.785860641697278,
+      "p50_ms": 9.651802000007592,
+      "p95_ms": 10.604852450251201,
+      "max_ms": 14.228853000531672,
+      "processor_fps": 102.18825268561746,
       "seam_updates": 0,
       "seam_update_mean_ms": null,
-      "peak_rss_mib": 119.12890625,
+      "peak_rss_mib": 110.08984375,
       "persistent_state_mib": 0.375762939453125,
       "work_shape": [
         108,
         192
       ],
-      "python": "3.14.0",
-      "numpy": "2.5.4",
-      "opencv": "5.0.0",
+      "python": "3.13.5",
+      "numpy": "2.2.4",
+      "opencv": "4.12.0",
       "opencv_threads": 1
     },
     {
@@ -82,34 +82,34 @@
       "height": 360,
       "frames": 120,
       "config": {
-        "intensity": 65,
-        "horizontal": 65,
-        "vertical": 45,
+        "intensity": 100,
+        "horizontal": 100,
+        "vertical": 100,
         "preserve": 75,
         "mode": "seam",
-        "seam_strength": 60,
+        "seam_strength": 100,
         "motion": 50,
         "smoothness": 70,
         "speed": 0.35,
         "coherence": 80,
         "seed": 17
       },
-      "mean_ms": 3.429766949830082,
-      "p50_ms": 2.686837002329412,
-      "p95_ms": 6.3660287009042795,
-      "max_ms": 10.528957998758415,
-      "processor_fps": 291.56499978797166,
+      "mean_ms": 16.585925233357557,
+      "p50_ms": 2.023097500568838,
+      "p95_ms": 89.63626609997846,
+      "max_ms": 92.01955000025919,
+      "processor_fps": 60.29208415752432,
       "seam_updates": 20,
-      "seam_update_mean_ms": 6.684701200538257,
-      "peak_rss_mib": 68.12109375,
+      "seam_update_mean_ms": 89.37203475011302,
+      "peak_rss_mib": 62.61328125,
       "persistent_state_mib": 0.533966064453125,
       "work_shape": [
         108,
         192
       ],
-      "python": "3.14.0",
-      "numpy": "2.5.4",
-      "opencv": "5.0.0",
+      "python": "3.13.5",
+      "numpy": "2.2.4",
+      "opencv": "4.12.0",
       "opencv_threads": 1
     },
     {
@@ -118,35 +118,35 @@
       "height": 720,
       "frames": 120,
       "config": {
-        "intensity": 65,
-        "horizontal": 65,
-        "vertical": 45,
+        "intensity": 100,
+        "horizontal": 100,
+        "vertical": 100,
         "preserve": 75,
         "mode": "seam",
-        "seam_strength": 60,
+        "seam_strength": 100,
         "motion": 50,
         "smoothness": 70,
         "speed": 0.35,
         "coherence": 80,
         "seed": 17
       },
-      "mean_ms": 9.723339424726873,
-      "p50_ms": 9.247966498151072,
-      "p95_ms": 13.197447299717169,
-      "max_ms": 15.983984998456435,
-      "processor_fps": 102.84532466869939,
+      "mean_ms": 25.305225183334795,
+      "p50_ms": 9.182545500152628,
+      "p95_ms": 123.04726984962144,
+      "max_ms": 130.75589300024149,
+      "processor_fps": 39.51753018418377,
       "seam_updates": 20,
-      "seam_update_mean_ms": 12.868869349949819,
-      "peak_rss_mib": 119.125,
+      "seam_update_mean_ms": 109.627773550028,
+      "peak_rss_mib": 109.03125,
       "persistent_state_mib": 0.533966064453125,
       "work_shape": [
         108,
         192
       ],
-      "python": "3.14.0",
-      "numpy": "2.5.4",
-      "opencv": "5.0.0",
+      "python": "3.13.5",
+      "numpy": "2.2.4",
+      "opencv": "4.12.0",
       "opencv_threads": 1
     }
   ]
-}
+}

effex.py

@@ -1918,11 +1918,11 @@ def build_temporal_sediment(ctx: EffectContext) -> Plan:
 
 COLLAPSE_PARAMS = (
     Param("intensity", "intensity", "how strongly the displacement is applied",
-          "percent", 65, minimum=0, maximum=100, step=5),
+          "percent", 100, minimum=0, maximum=100, step=5),
     Param("horizontal", "horizontal scale", "horizontal deformation amount",
-          "percent", 65, minimum=0, maximum=100, step=5),
+          "percent", 100, minimum=0, maximum=100, step=5),
     Param("vertical", "vertical scale", "vertical deformation amount",
-          "percent", 45, minimum=0, maximum=100, step=5),
+          "percent", 100, minimum=0, maximum=100, step=5),
     Param("preserve", "content preservation", "protect strong edges from collapse",
           "percent", 75, minimum=0, maximum=100, step=5),
     Param("mode", "mode", "progressive content-aware scaling or elastic warping",
@@ -1932,7 +1932,7 @@ def build_temporal_sediment(ctx: EffectContext) -> Plan:
               ("elastic", "elastic warp", "smooth content-weighted CPU remapping"),
           )),
     Param("seam_strength", "shrink amount", "maximum content-aware shrink by the end of the clip",
-          "percent", 60, minimum=0, maximum=100, step=5),
+          "percent", 100, minimum=0, maximum=100, step=5),
     Param("motion", "motion influence", "amplify elastic deformation in moving regions",
           "percent", 50, minimum=0, maximum=100, step=5),
     Param("smoothness", "field smoothness", "larger, smoother deformation regions",

stateful_effects.py

@@ -293,8 +293,8 @@ class ElasticCollapse:
     MAX_WORK_SIDE = 192
     MAX_STATE_BYTES = 4 * 1024 * 1024
     DEFAULTS = {
-        "intensity": 65, "horizontal": 65, "vertical": 45, "preserve": 75,
-        "mode": "seam", "seam_strength": 60, "motion": 50,
+        "intensity": 100, "horizontal": 100, "vertical": 100, "preserve": 75,
+        "mode": "seam", "seam_strength": 100, "motion": 50,
         "smoothness": 70, "speed": 0.35, "coherence": 80, "seed": 0,
     }
 
@@ -383,7 +383,7 @@ def displacement(self) -> Any:
         """A copy of the applied normalized (dx, dy) field for diagnostics."""
         if self._field is None:
             return None
-        amount = (self._progress if self._config["mode"] == "seam"
+        amount = (-self._progress if self._config["mode"] == "seam"
                   else self._config["intensity"] / 100.0)
         return self._field * amount
 
@@ -501,7 +501,8 @@ def process(self, frame, frame_index: int, timestamp: Optional[float],
                 or not (self._config["horizontal"] or self._config["vertical"])):
             return frame.copy()
         full = cv.resize(self._field, (width, height), interpolation=cv.INTER_LINEAR)
-        amount = (self._progress if self._config["mode"] == "seam"
+        # Seam removal's resize map expands the retained content; invert it to shrink.
+        amount = (-self._progress if self._config["mode"] == "seam"
                   else intensity)
         map_x = full[:, :, 0] * (amount * max(0, width - 1))
         map_y = full[:, :, 1] * (amount * max(0, height - 1))

tests/collapse_regression.py

@@ -171,7 +171,11 @@ def test_collapse_is_registered_with_all_controls(self):
         effect = effex.EFFECT_BY_KEY["collapse"]
         self.assertEqual({p.key for p in effect.params}, set(stateful_effects.ElasticCollapse.DEFAULTS))
         self.assertIs(effect.frame_processor, stateful_effects.ElasticCollapse)
-        self.assertEqual(effex.default_opts(effect)["mode"], "seam")
+        defaults = effex.default_opts(effect)
+        self.assertEqual(defaults["mode"], "seam")
+        self.assertEqual([defaults[key] for key in
+                          ("intensity", "horizontal", "vertical", "seam_strength")],
+                         [100, 100, 100, 100])
 
     def test_seam_mode_is_distinct_and_zero_strength_disables_it(self):
         frame = self.frame()
@@ -205,6 +209,18 @@ def test_content_aware_scale_grows_linearly_over_clip_duration(self):
         self.assertEqual(final.shape, frame.shape)
         self.assertGreater(float(np.abs(final.astype(float) - frame).mean()), 4)
 
+    def test_content_aware_scale_shrinks_the_salient_subject(self):
+        frame = np.zeros((64, 96, 3), dtype=np.uint8)
+        frame[20:44, 30:66] = 255
+        processor = self.processor(
+            mode="seam", intensity=100, horizontal=100, vertical=0,
+            preserve=100, seam_strength=100, coherence=0, _clip_duration=1,
+        )
+        out = processor.process(frame, 30, 1.0, {})
+        ys, xs = np.where(out.mean(axis=2) > 128)
+        self.assertEqual(out.shape, frame.shape)
+        self.assertLess(int(np.ptp(xs)) + 1, 36)
+
     def test_seam_updates_are_rate_limited_and_deterministic(self):
         frame = self.frame()
         a, b = self.processor(mode="seam"), self.processor(mode="seam")

tests/menu_smoke.py

@@ -19,6 +19,7 @@
 UP = b"\x1bOA"
 DOWN = b"\x1bOB"
 RIGHT = b"\x1bOC"
+LEFT = b"\x1bOD"
 ENTER = b"\r"
 
 
@@ -860,8 +861,8 @@ def scenario_collapse(c: Checker):
                     press_and_wait(d, UP)
                 c.check("collapse elastic: legacy warp mode can be selected",
                         "elastic warp" in last_frame(d))
-            press_and_wait(d, RIGHT)
-            c.check(f"collapse {mode}: intensity is editable", "70%" in last_frame(d))
+            press_and_wait(d, LEFT)
+            c.check(f"collapse {mode}: intensity is editable", "95%" in last_frame(d))
             if mode == "seam":
                 for _ in range(4):
                     press_and_wait(d, DOWN)