fix(editor): a crop of a video with wide pixels keeps what its frame showed

The browser reports such a video by its shown size and ffmpeg's crop counts stored pixels, so a
frame over the middle half of a 320 stored, 640 shown video kept the whole width. A crop now carries
the size it was measured on and is scaled to the file's own pixels. A spec from before it did is cut
as it always was. Also drops a stale note in downloads/quota.py: edit jobs are held to the cap.

backend/app/downloads/edit.py

@@ -60,6 +60,31 @@ class EditAborted(Exception):
 _MAX_SEGMENTS = 100
 
 
+def _crop_frame(crop: dict) -> dict | None:
+    """The picture size a crop was measured on, or None when it has none it could be scaled by (a
+    spec from before it was sent, which `crop_filter` cuts in the file's own pixels)."""
+    try:
+        w, h = int(crop["frame_w"]), int(crop["frame_h"])
+    # silent-ok: no frame is a crop's older form, and it is still the crop that was asked for.
+    except (KeyError, TypeError, ValueError):
+        return None
+    return {"frame_w": w, "frame_h": h} if w > 0 and h > 0 else None
+
+
+def crop_filter(crop: dict) -> str:
+    """The ffmpeg `crop` of an edit. Its numbers count the picture as the browser reports it, which
+    is its SHOWN size: a video with wide pixels (320 stored, 640 shown) would be cut off to one side
+    in ffmpeg's stored ones, so they are scaled to `iw` / `ih` by the size they were measured on. A
+    rotated file needs nothing more: both the browser and ffmpeg's filters see it upright."""
+    frame_w, frame_h = crop.get("frame_w"), crop.get("frame_h")
+    if not frame_w or not frame_h:
+        return f"crop={crop['w']}:{crop['h']}:{crop['x']}:{crop['y']}"
+    return (
+        f"crop=iw*{crop['w']}/{frame_w}:ih*{crop['h']}/{frame_h}"
+        f":iw*{crop['x']}/{frame_w}:ih*{crop['y']}/{frame_h}"
+    )
+
+
 def normalize_edit_spec(spec: dict | None) -> dict:
     """Canonicalize a raw edit spec: clamp/round trim, coerce crop to ints, drop empties.
 
@@ -81,7 +106,8 @@ def normalize_edit_spec(spec: dict | None) -> dict:
             )
             c = None
         if c and c["w"] > 0 and c["h"] > 0 and c["x"] >= 0 and c["y"] >= 0:
-            out["crop"] = c
+            frame = _crop_frame(crop)
+            out["crop"] = {**c, **frame} if frame else c
 
     # Multi-segment JOIN (cut-list → one concatenated file). Takes precedence over `trim`.
     segs = spec.get("segments")
@@ -268,7 +294,7 @@ def build_edit_cmd(src: Path, dest: Path, spec: dict, out_ext: str) -> list[str]
     if reencode:
         # Input -ss + re-encode is frame-accurate (decodes from the prior keyframe, discards up to
         # the exact cut). A crop adds the filter; an accurate trim only evens the frame.
-        cropf = f"crop={crop['w']}:{crop['h']}:{crop['x']}:{crop['y']}," if crop else ""
+        cropf = f"{crop_filter(crop)}," if crop else ""
         cmd += ["-vf", cropf + EVEN_FRAME, *X264_ARGS, *AAC_ARGS]
     else:
         cmd += ["-c", "copy", "-avoid_negative_ts", "make_zero"]
@@ -313,7 +339,7 @@ def build_concat_plan(src: Path, dest: Path, spec: dict, out_ext: str, staging:
     ]
 
     if needs_reencode(spec):
-        cropf = f",crop={crop['w']}:{crop['h']}:{crop['x']}:{crop['y']}" if crop else ""
+        cropf = f",{crop_filter(crop)}" if crop else ""
         parts, labels = [], []
         for i, s in enumerate(segs):
             st, en = s["start_s"], s["end_s"]