diff --git a/backend/api/cl2k_maker.py b/backend/api/cl2k_maker.py
index f81ffbe7..db41e89d 100644
--- a/backend/api/cl2k_maker.py
+++ b/backend/api/cl2k_maker.py
@@ -268,14 +268,25 @@ def _save_response(logger: Any, *, done: str, what: str, run) -> JSONResponse:
     )
 
 
-def _crop_tuple(req: Any):
-    """Assemble the (x, y, w, h) fit crop from a request, or None if unset.
+def _framing(req: Any) -> geo.Framing:
+    """Flat request fields -> the renderer's framing bundle.
 
-    Works for any request carrying ``crop_x/y/w/h`` (GenerateRequest, SeasonsRequest).
-    Only used in ``fit`` mode; all four fields must be present for a crop to apply
-    (a partial crop is ignored so the whole backdrop is fitted)."""
-    parts = (req.crop_x, req.crop_y, req.crop_w, req.crop_h)
-    return tuple(parts) if all(p is not None for p in parts) else None
+    ``crop_*`` is poster-models-only, and a PARTIAL crop is ignored (whole backdrop).
+    """
+    parts = (
+        getattr(req, "crop_x", None),
+        getattr(req, "crop_y", None),
+        getattr(req, "crop_w", None),
+        getattr(req, "crop_h", None),
+    )
+    return geo.Framing(
+        focus_x=req.focus_x,
+        fit_mode=req.fit_mode,
+        v_pos=req.v_pos,
+        zoom=req.zoom,
+        mirror=req.mirror,
+        crop=tuple(parts) if all(p is not None for p in parts) else None,
+    )
 
 
 class LogoFetchError(Exception):
@@ -659,12 +670,7 @@ def preview(
             imdb_id=req.imdb_id,
             mask_bytes=mask_bytes,
             apply_ai=req.remove_text,
-            focus_x=req.focus_x,
-            fit_mode=req.fit_mode,
-            crop=_crop_tuple(req),
-            v_pos=req.v_pos,
-            zoom=req.zoom,
-            mirror=req.mirror,
+            framing=_framing(req),
             band_label=req.band_label,
             logo_scale=req.logo_scale,
             logo_y_offset=req.logo_y_offset,
@@ -722,12 +728,7 @@ def generate(
             custom_logo_bytes=_b64_to_bytes(req.logo_b64),
             mask_bytes=mask_bytes,
             apply_ai=req.remove_text,
-            focus_x=req.focus_x,
-            fit_mode=req.fit_mode,
-            crop=_crop_tuple(req),
-            v_pos=req.v_pos,
-            zoom=req.zoom,
-            mirror=req.mirror,
+            framing=_framing(req),
             band_label=req.band_label,
             logo_scale=req.logo_scale,
             logo_y_offset=req.logo_y_offset,
@@ -869,11 +870,7 @@ def square_preview(
         req,
         lambda raw: render_square_art(
             backdrop_bytes=raw,
-            focus_x=req.focus_x,
-            fit_mode=req.fit_mode,
-            v_pos=req.v_pos,
-            zoom=req.zoom,
-            mirror=req.mirror,
+            framing=_framing(req),
         ),
     )
 
@@ -905,11 +902,7 @@ def square_generate(
             imdb_id=req.imdb_id,
             backdrop_path=req.backdrop_path,
             backdrop_bytes=_b64_to_bytes(req.backdrop_b64),
-            focus_x=req.focus_x,
-            fit_mode=req.fit_mode,
-            v_pos=req.v_pos,
-            zoom=req.zoom,
-            mirror=req.mirror,
+            framing=_framing(req),
             season_number=req.season_number,
             save_local=req.save_local,
             upload_gdrive=req.upload_gdrive,
@@ -959,11 +952,7 @@ def background_preview(
             backdrop_bytes=raw,
             width=1920,
             height=1080,
-            focus_x=req.focus_x,
-            fit_mode=req.fit_mode,
-            v_pos=req.v_pos,
-            zoom=req.zoom,
-            mirror=req.mirror,
+            framing=_framing(req),
         ),
     )
 
@@ -995,11 +984,7 @@ def background_generate(
             imdb_id=req.imdb_id,
             backdrop_path=req.backdrop_path,
             backdrop_bytes=_b64_to_bytes(req.backdrop_b64),
-            focus_x=req.focus_x,
-            fit_mode=req.fit_mode,
-            v_pos=req.v_pos,
-            zoom=req.zoom,
-            mirror=req.mirror,
+            framing=_framing(req),
             resolution=req.resolution,
             season_number=req.season_number,
             save_local=req.save_local,
@@ -1182,12 +1167,7 @@ def psd_export(
             logo_y_offset=req.logo_y_offset,
             logo_flip_bytes=_b64_to_bytes(req.logo_flip_b64),
             logo_erase_bytes=_b64_to_bytes(req.logo_erase_b64),
-            focus_x=req.focus_x,
-            fit_mode=req.fit_mode,
-            crop=_crop_tuple(req),
-            v_pos=req.v_pos,
-            zoom=req.zoom,
-            mirror=req.mirror,
+            framing=_framing(req),
             whiten=req.whiten,
             flat_white=req.flat_white,
             logo_3d=req.logo_3d,
@@ -1382,12 +1362,7 @@ def _progress(entry: Dict[str, Any]) -> None:
                 backdrop_bytes=backdrop_bytes,
                 logo_path=req.logo_path,
                 custom_logo_bytes=logo_bytes,
-                fit_mode=req.fit_mode,
-                focus_x=req.focus_x,
-                crop=_crop_tuple(req),
-                v_pos=req.v_pos,
-                zoom=req.zoom,
-                mirror=req.mirror,
+                framing=_framing(req),
                 logo_scale=req.logo_scale,
                 logo_y_offset=req.logo_y_offset,
                 whiten=req.whiten,
diff --git a/backend/modules/cl2k_maker.py b/backend/modules/cl2k_maker.py
index 613e0d82..2f06e811 100644
--- a/backend/modules/cl2k_maker.py
+++ b/backend/modules/cl2k_maker.py
@@ -3,6 +3,7 @@
 import os
 import shutil
 import tempfile
+from dataclasses import replace
 from typing import Any, Callable, Dict, List, Optional, Tuple
 
 from backend.util.cl2k import color
@@ -197,12 +198,7 @@ def _resolve_and_render(
     mask_bytes: Optional[bytes] = None,
     backdrop_bytes: Optional[bytes] = None,
     apply_ai: bool = False,
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    crop: Optional[Tuple[float, float, float, float]] = None,
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
     band_label: str = "",
     logo_scale: float = 1.0,
     logo_y_offset: int = 0,
@@ -265,9 +261,9 @@ def _resolve_and_render(
     # extend). The fill happens here, before the gradient/logo, so the AI only sees
     # the backdrop; the resulting canvas is already 1000×1500, so it renders as a
     # straight cover (identity).
-    if fit_mode == "extend":
+    if framing.fit_mode == "extend":
         canvas_bytes, extend_mask = renderer.fit_extend_canvas(
-            backdrop_bytes, crop, zoom=zoom, v_pos=v_pos
+            backdrop_bytes, framing.crop, zoom=framing.zoom, v_pos=framing.v_pos
         )
         # AI runs only on a real generate (allow_ai_extend) AND when the provider
         # is fully configured — unavailable_reason(), not is_enabled(): a provider
@@ -291,7 +287,7 @@ def _resolve_and_render(
             # CANVAS_W x CANVAS_H with zoom/v_pos baked in by fit_extend_canvas.
             # Re-applying them re-scales the finished image and, for v_pos > 0,
             # crops rows off the top and fades a black band over the fill.
-            fit_mode, crop, zoom, v_pos = "cover", None, 1.0, 0.0
+            framing = replace(framing, fit_mode="cover", crop=None, zoom=1.0, v_pos=0.0)
         else:
             if extend_mask is not None and logger:
                 reason = (
@@ -305,7 +301,7 @@ def _resolve_and_render(
                 logger.info(
                     f"cl2k: extend — {reason}; using the free edge-extend fit instead"
                 )
-            fit_mode = "fit"
+            framing = replace(framing, fit_mode="fit")
 
     # Only run AI removal when explicitly requested (a brushed mask, or the
     # apply_ai flag for OpenAI's maskless mode) — never on every auto-render.
@@ -388,12 +384,7 @@ def _resolve_and_render(
         flat_white=flat_white,
         logo_3d=logo_3d,
         invert=invert,
-        focus_x=focus_x,
-        fit_mode=fit_mode,
-        crop=crop,
-        v_pos=v_pos,
-        zoom=zoom,
-        mirror=mirror,
+        framing=framing,
         band_label=band_label,
         place_logo=place_logo,
         text_logo_stroke=cfg.text_logo_stroke,
@@ -432,12 +423,7 @@ def generate_for_item(
     mask_bytes: Optional[bytes] = None,
     backdrop_bytes: Optional[bytes] = None,
     apply_ai: bool = False,
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    crop: Optional[Tuple[float, float, float, float]] = None,
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
     band_label: str = "",
     logo_scale: float = 1.0,
     logo_y_offset: int = 0,
@@ -505,12 +491,7 @@ def generate_for_item(
         mask_bytes=mask_bytes,
         backdrop_bytes=backdrop_bytes,
         apply_ai=apply_ai,
-        focus_x=focus_x,
-        fit_mode=fit_mode,
-        crop=crop,
-        v_pos=v_pos,
-        zoom=zoom,
-        mirror=mirror,
+        framing=framing,
         band_label=band_label,
         logo_scale=logo_scale,
         logo_y_offset=logo_y_offset,
@@ -560,11 +541,7 @@ def generate_square_art(
     imdb_id: Optional[str] = None,
     backdrop_path: Optional[str] = None,
     backdrop_bytes: Optional[bytes] = None,
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
     season_number: Optional[int] = None,
     save_local: bool = True,
     upload_gdrive: Optional[bool] = None,
@@ -601,11 +578,7 @@ def generate_square_art(
         backdrop_bytes = image_fetch.download(backdrop_path)
     blob = renderer.render_square_art(
         backdrop_bytes=backdrop_bytes,
-        focus_x=focus_x,
-        fit_mode=fit_mode,
-        v_pos=v_pos,
-        zoom=zoom,
-        mirror=mirror,
+        framing=framing,
     )
     return _persist_poster(
         db,
@@ -645,11 +618,7 @@ def generate_background_art(
     imdb_id: Optional[str] = None,
     backdrop_path: Optional[str] = None,
     backdrop_bytes: Optional[bytes] = None,
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
     resolution: str = "1080p",
     season_number: Optional[int] = None,
     save_local: bool = True,
@@ -692,11 +661,7 @@ def generate_background_art(
         backdrop_bytes=backdrop_bytes,
         width=width,
         height=height,
-        focus_x=focus_x,
-        fit_mode=fit_mode,
-        v_pos=v_pos,
-        zoom=zoom,
-        mirror=mirror,
+        framing=framing,
     )
     return _persist_poster(
         db,
@@ -1447,12 +1412,7 @@ def generate_seasons(
     year: Optional[int] = None,
     tvdb_id: Optional[int] = None,
     imdb_id: Optional[str] = None,
-    fit_mode: str = "cover",
-    focus_x: float = 0.5,
-    crop: Optional[Tuple[float, float, float, float]] = None,
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
     logo_scale: float = 1.0,
     logo_y_offset: int = 0,
     whiten: Optional[bool] = None,  # None = module config (whiten_logo)
@@ -1503,12 +1463,7 @@ def generate_seasons(
                 backdrop_bytes=backdrop_bytes,
                 logo_path=logo_path,
                 custom_logo_bytes=custom_logo_bytes,
-                fit_mode=fit_mode,
-                focus_x=focus_x,
-                crop=crop,
-                v_pos=v_pos,
-                zoom=zoom,
-                mirror=mirror,
+                framing=framing,
                 logo_scale=logo_scale,
                 logo_y_offset=logo_y_offset,
                 whiten=whiten,
@@ -1557,12 +1512,7 @@ def psd_for_item(
     logo_y_offset: int = 0,
     logo_flip_bytes: Optional[bytes] = None,  # B/W touch-up regions (mask PNG)
     logo_erase_bytes: Optional[bytes] = None,  # erase regions (mask PNG, white=erase)
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    crop: Optional[Tuple[float, float, float, float]] = None,
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
     whiten: Optional[bool] = None,  # None = module config (whiten_logo)
     flat_white: bool = False,  # paint the logo a flat pure-white silhouette
     logo_3d: bool = False,  # extruded art -> flat-white lit face; wins over flat_white
@@ -1605,12 +1555,7 @@ def psd_for_item(
         backdrop_bytes = image_fetch.download(backdrop_path)
     framed = renderer.frame_backdrop(
         backdrop_bytes=backdrop_bytes,
-        focus_x=focus_x,
-        fit_mode=fit_mode,
-        crop=crop,
-        v_pos=v_pos,
-        zoom=zoom,
-        mirror=mirror,
+        framing=framing,
     )
     logo_bytes = custom_logo_bytes
     if logo_bytes is None and logo_path:
diff --git a/backend/util/cl2k/geometry.py b/backend/util/cl2k/geometry.py
index ba368513..9e202a86 100644
--- a/backend/util/cl2k/geometry.py
+++ b/backend/util/cl2k/geometry.py
@@ -13,8 +13,9 @@
 
 from __future__ import annotations
 
+from dataclasses import dataclass
 from pathlib import Path
-from typing import Optional
+from typing import Optional, Tuple
 
 # ----- canvas ----------------------------------------------------------------
 CANVAS_W = 1000
@@ -126,6 +127,23 @@ def auto_logo_size(
 # source only, positive pans down and may edge-extend into the gradient zone.
 V_POS_MIN, V_POS_MAX = -1.0, 1.0
 
+
+@dataclass(frozen=True)
+class Framing:
+    """How a backdrop is fitted to the canvas; one bundle for every art path.
+
+    The API wire format stays FLAT — build one of these at the endpoint boundary.
+    """
+
+    focus_x: float = 0.5  # 0..1 horizontal focal point; 0.5 = centre
+    fit_mode: str = "cover"  # cover (crop to fill) | fit (contain) | extend (AI)
+    v_pos: float = 0.0  # -1..1, 0 = centred (fit/extend read it 0..1 top-anchored)
+    zoom: float = 1.0  # scale relative to the cover/fit baseline
+    mirror: bool = False  # flip the artwork horizontally, at the end of framing
+    # fit/extend only: isolates the subject region before the fit. render_framed_art
+    # (square + background art) has no crop stage and ignores this.
+    crop: Optional[Tuple[float, float, float, float]] = None
+
 # ----- logo whitening (CL2K two-tone) -----------------------------------------
 # Real CL2K logos are black & white, not flat white silhouettes: coloured/bright
 # fills go pure white while the artwork's dark keylines and interior accents stay
diff --git a/backend/util/cl2k/renderer.py b/backend/util/cl2k/renderer.py
index 374c1bc8..64c2ac3e 100644
--- a/backend/util/cl2k/renderer.py
+++ b/backend/util/cl2k/renderer.py
@@ -943,14 +943,12 @@ def render_framed_art(
     backdrop_bytes: bytes,
     width: int,
     height: int,
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
 ) -> bytes:
     """Render plain framed artwork at ``width``×``height`` — no gradient/logo/label.
 
+    Framing knobs below are :class:`geometry.Framing` fields.
+
     ``fit_mode`` ``"cover"`` fills the canvas (cropping the overflowing edges);
     ``"fit"`` contains the whole image on black (letterbox). ``zoom`` (0.5–3.0)
     scales from that baseline — raise it in ``fit`` to punch in from contain toward
@@ -960,7 +958,9 @@ def render_framed_art(
     gradient here to hide an extended band, so ``v_pos`` is source-bounded both
     ways. ``mirror`` flips the finished frame horizontally. Encoded at CL2K quality.
     """
-    zoom = max(geo.ZOOM_MIN, min(float(zoom or 1.0), geo.ZOOM_MAX))
+    focus_x, fit_mode = framing.focus_x, framing.fit_mode
+    v_pos, mirror = framing.v_pos, framing.mirror
+    zoom = max(geo.ZOOM_MIN, min(float(framing.zoom or 1.0), geo.ZOOM_MAX))
     with Image(blob=backdrop_bytes) as img:
         base = (
             min(width / img.width, height / img.height)
@@ -987,35 +987,18 @@ def render_square_art(
     *,
     backdrop_bytes: bytes,
     size: int = 1000,
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
 ) -> bytes:
     """Render square (1:1) art from a backdrop/poster — just the framed artwork."""
     return render_framed_art(
         backdrop_bytes=backdrop_bytes,
         width=size,
         height=size,
-        focus_x=focus_x,
-        fit_mode=fit_mode,
-        v_pos=v_pos,
-        zoom=zoom,
-        mirror=mirror,
+        framing=framing,
     )
 
 
-def _framed_inset_base(
-    backdrop_bytes: bytes,
-    *,
-    focus_x: float,
-    fit_mode: str,
-    crop: Optional[Tuple[float, float, float, float]],
-    v_pos: float,
-    zoom: float,
-    mirror: bool,
-) -> Image:
+def _framed_inset_base(backdrop_bytes: bytes, framing: geo.Framing) -> Image:
     """Frame the backdrop FULL-BLEED and return a full CANVAS image.
 
     The template's stroke is Style=Inside on a full-canvas layer, so it paints
@@ -1032,34 +1015,29 @@ def _framed_inset_base(
     render_cl2k and frame_backdrop both go through here, so they stay
     pixel-identical (the PSD POSTER-layer parity the exporter relies on).
 
-    ``mirror`` flips the framed artwork horizontally. It lands HERE, at the end
-    of framing, rather than on the source bytes: the AI text-removal mask and the
-    extend outpaint (modules.cl2k_maker) are brushed/built in source space, and
-    the logo/label the caller composites next must never come out backwards.
+    ``mirror`` flips the artwork HERE, at the end of framing — never on the source
+    bytes: the AI mask and extend outpaint are built in source space.
     """
     base = Image(
         width=geo.CANVAS_W, height=geo.CANVAS_H, background=Color(geo.BORDER_COLOR)
     )
     with Image(blob=backdrop_bytes) as art:
-        if fit_mode == "fit":
-            _fit_resize(art, geo.CANVAS_W, geo.CANVAS_H, crop, v_pos, zoom)
+        if framing.fit_mode == "fit":
+            _fit_resize(
+                art, geo.CANVAS_W, geo.CANVAS_H, framing.crop, framing.v_pos, framing.zoom
+            )
         else:
-            _cover_resize(art, geo.CANVAS_W, geo.CANVAS_H, focus_x, v_pos, zoom)
+            _cover_resize(
+                art, geo.CANVAS_W, geo.CANVAS_H, framing.focus_x, framing.v_pos, framing.zoom
+            )
         base.composite(art, left=0, top=0)
-    if mirror:
+    if framing.mirror:
         base.flop()
     return base
 
 
 def frame_backdrop(
-    *,
-    backdrop_bytes: bytes,
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    crop: Optional[Tuple[float, float, float, float]] = None,
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    *, backdrop_bytes: bytes, framing: geo.Framing = geo.Framing()
 ) -> bytes:
     """Frame a backdrop to the 2:3 canvas exactly as :func:`render_cl2k` would
     and return PNG bytes.
@@ -1069,15 +1047,7 @@ def frame_backdrop(
     (edge-extend fills, seam blending) live only in this module and must not be
     re-implemented elsewhere.
     """
-    with _framed_inset_base(
-        backdrop_bytes,
-        focus_x=focus_x,
-        fit_mode=fit_mode,
-        crop=crop,
-        v_pos=v_pos,
-        zoom=zoom,
-        mirror=mirror,
-    ) as base:
+    with _framed_inset_base(backdrop_bytes, framing) as base:
         base.format = "png"
         return base.make_blob()
 
@@ -1099,12 +1069,7 @@ def render_cl2k(
     logo_3d: bool = False,  # extruded art -> flat-white lit face
     invert: bool = False,  # plate logo -> clearlogo (white->transparent, black->white)
     font_path: Optional[str] = None,
-    focus_x: float = 0.5,
-    fit_mode: str = "cover",
-    crop: Optional[Tuple[float, float, float, float]] = None,
-    v_pos: float = 0.0,
-    zoom: float = 1.0,
-    mirror: bool = False,
+    framing: geo.Framing = geo.Framing(),
     band_label: str = "",
     place_logo: bool = True,
     text_logo_stroke: int = 0,
@@ -1119,7 +1084,7 @@ def render_cl2k(
     ``COMPLETE LIMITED SERIES`` or ``SPECIALS``), overriding the automatic
     COLLECTION / season label. Long strings use the tighter PSD tracking.
 
-    ``fit_mode`` controls how the backdrop fills the 2:3 canvas:
+    ``framing`` controls how the backdrop fills the 2:3 canvas:
 
     - ``"cover"`` (default): scale up and crop to fill; ``focus_x`` (0..1) and
       ``v_pos`` (-1..1) choose which part is kept (0.5/0 = centre). Best when the
@@ -1139,15 +1104,7 @@ def render_cl2k(
     label_font = font_path or geo.resolve_font(bold=False)
     title_font = font_path or geo.resolve_font(bold=True)
 
-    with _framed_inset_base(
-        backdrop_bytes,
-        focus_x=focus_x,
-        fit_mode=fit_mode,
-        crop=crop,
-        v_pos=v_pos,
-        zoom=zoom,
-        mirror=mirror,
-    ) as base:
+    with _framed_inset_base(backdrop_bytes, framing) as base:
         with Image(filename=str(geo.GRADIENT_PNG)) as grad:
             base.composite(grad, left=0, top=0)
 
@@ -1383,7 +1340,7 @@ def main() -> None:
         season_text=args.season_text,
         logo_max_width=args.width,
         whiten=not args.no_whiten,
-        mirror=args.mirror,
+        framing=geo.Framing(mirror=args.mirror),
         font_path=args.font,
     )
     with open(args.out, "wb") as fh:
diff --git a/tests/test_cl2k_maker.py b/tests/test_cl2k_maker.py
index 53051328..1a3b6395 100644
--- a/tests/test_cl2k_maker.py
+++ b/tests/test_cl2k_maker.py
@@ -17,6 +17,7 @@
 from wand.image import Image  # noqa: E402
 
 import backend.modules.cl2k_maker as maker  # noqa: E402
+from backend.util.cl2k import geometry as geo  # noqa: E402
 
 
 def _logo_png(width, height):
@@ -162,12 +163,15 @@ def test_a_successful_ai_extend_is_not_reframed_again(env, monkeypatch):
         tmdb_id=221230,
         backdrop_bytes=_backdrop_png(),
         logo_path="/iMuXNXkPq9OCKw5jljoIxkv8IRV.png",
-        fit_mode="extend",
-        v_pos=0.4,
-        zoom=1.5,
+        framing=geo.Framing(fit_mode="extend", v_pos=0.4, zoom=1.5, mirror=True),
     )
 
-    assert seen["fit_mode"] == "cover"
-    assert seen["crop"] is None
-    assert seen["v_pos"] == 0.0, "v_pos was already baked into the extended canvas"
-    assert seen["zoom"] == 1.0, "zoom was already baked into the extended canvas"
+    framing = seen["framing"]
+    assert framing.fit_mode == "cover"
+    assert framing.crop is None
+    assert framing.v_pos == 0.0, "v_pos was already baked into the extended canvas"
+    assert framing.zoom == 1.0, "zoom was already baked into the extended canvas"
+    # The reset is scoped to what fit_extend_canvas baked in. mirror is applied
+    # later, at the end of framing, so clearing it here would silently un-flip
+    # every extended poster.
+    assert framing.mirror is True
diff --git a/tests/test_cl2k_mirror.py b/tests/test_cl2k_mirror.py
index 183f546e..97aca248 100644
--- a/tests/test_cl2k_mirror.py
+++ b/tests/test_cl2k_mirror.py
@@ -14,6 +14,7 @@
 
 import base64
 import types
+from dataclasses import replace
 
 import pytest
 
@@ -66,6 +67,12 @@ def _strip(blob, top, height):
         return img.make_blob("RGB")
 
 
+# What every hop must receive: mirror on, the rest defaulted. Comparing the whole
+# bundle (not just .mirror) is the point of the refactor — a layer that rebuilds
+# the object field by field and drops one now fails here.
+_FRAMING = geo.Framing(mirror=True)
+
+
 def _logger():
     return types.SimpleNamespace(
         info=lambda *a, **k: None,
@@ -85,21 +92,25 @@ def test_framed_art_mirror_swaps_the_sides(size):
     width, height = size
     kw = dict(backdrop_bytes=_two_tone(), width=width, height=height)
     plain_l, plain_r = _sides(render_framed_art(**kw))
-    mirror_l, mirror_r = _sides(render_framed_art(**kw, mirror=True))
+    mirror_l, mirror_r = _sides(render_framed_art(**kw, framing=geo.Framing(mirror=True)))
 
     assert plain_l.red > 0.9 and plain_r.blue > 0.9
     assert mirror_l.blue > 0.9 and mirror_r.red > 0.9
 
 
 def test_square_art_forwards_mirror():
-    left, right = _sides(render_square_art(backdrop_bytes=_two_tone(), mirror=True))
+    left, right = _sides(
+        render_square_art(backdrop_bytes=_two_tone(), framing=geo.Framing(mirror=True))
+    )
     assert left.blue > 0.9 and right.red > 0.9
 
 
 def test_frame_backdrop_mirrors_the_poster_frame():
     # The .psd POSTER layer is this function's output, so an unmirrored frame here
     # would export a document that disagrees with the poster beside it.
-    left, right = _sides(frame_backdrop(backdrop_bytes=_two_tone(), mirror=True))
+    left, right = _sides(
+        frame_backdrop(backdrop_bytes=_two_tone(), framing=geo.Framing(mirror=True))
+    )
     assert left.blue > 0.9 and right.red > 0.9
 
 
@@ -108,14 +119,12 @@ def test_mirror_composes_with_the_framing_it_does_not_replace_it(fit_mode):
     # Off-centre framing, so a flip applied to the SOURCE instead of the framed
     # result would land the crop window on the other side and fail this.
     # frame_backdrop returns PNG, so the comparison is lossless and exact.
-    kw = dict(
-        backdrop_bytes=_two_tone(),
-        fit_mode=fit_mode,
-        focus_x=0.2,
-        v_pos=0.3,
-        zoom=1.6,
+    off = geo.Framing(fit_mode=fit_mode, focus_x=0.2, v_pos=0.3, zoom=1.6)
+    on = replace(off, mirror=True)
+    art = _two_tone()
+    assert _sig(frame_backdrop(backdrop_bytes=art, framing=on)) == _sig(
+        frame_backdrop(backdrop_bytes=art, framing=off), flop=True
     )
-    assert _sig(frame_backdrop(**kw, mirror=True)) == _sig(frame_backdrop(**kw), flop=True)
 
 
 def test_render_cl2k_mirrors_the_artwork_and_never_the_label():
@@ -126,7 +135,8 @@ def test_render_cl2k_mirrors_the_artwork_and_never_the_label():
         title="Mirror Test",
         season_text="Season one",
     )
-    plain, mirrored = render_cl2k(**kw), render_cl2k(**kw, mirror=True)
+    plain = render_cl2k(**kw)
+    mirrored = render_cl2k(**kw, framing=geo.Framing(mirror=True))
 
     left, right = _sides(mirrored)
     assert left.blue > 0.9 and right.red > 0.9  # artwork flipped
@@ -175,7 +185,7 @@ def test_an_endpoint_forwards_mirror_to_its_maker(route, target, model, tasks, m
     # Split, because these endpoints answer a raising stub with a clean 4xx —
     # "never called" would otherwise read as "called without mirror".
     assert seen, f"{route} never reached {target}"
-    assert seen.get("mirror") is True, f"{route} dropped mirror on the way to {target}"
+    assert seen["framing"] == _FRAMING, f"{route} mangled the framing reaching {target}"
 
 
 @pytest.mark.parametrize("route,model", [("square_preview", api.SquareArtRequest),
@@ -189,7 +199,7 @@ def test_an_art_preview_forwards_mirror_to_the_renderer(route, model, monkeypatc
 
     getattr(api, route)(req, db=object(), logger=_logger())
 
-    assert seen.get("mirror") is True
+    assert seen["framing"] == _FRAMING
 
 
 def test_the_seasons_job_carries_mirror_into_every_season(monkeypatch):
@@ -202,7 +212,7 @@ def test_the_seasons_job_carries_mirror_into_every_season(monkeypatch):
 
     api._run_seasons_job(api._new_season_job(1, "T"), object(), _logger(), req)
 
-    assert seen.get("mirror") is True
+    assert seen["framing"] == _FRAMING
 
 
 @pytest.mark.parametrize(
@@ -228,10 +238,10 @@ def test_an_asset_maker_forwards_mirror_to_the_renderer(maker_fn, renderer_fn, m
         title="T",
         tmdb_id=7,
         backdrop_bytes=b"ART",
-        mirror=True,
+        framing=_FRAMING,
     )
 
-    assert seen.get("mirror") is True
+    assert seen["framing"] == _FRAMING
 
 
 def test_resolve_and_render_forwards_mirror_to_the_poster_renderer(monkeypatch):
@@ -257,10 +267,10 @@ def test_resolve_and_render_forwards_mirror_to_the_poster_renderer(monkeypatch):
         tmdb_id=7,
         backdrop_bytes=b"ART",
         custom_logo_bytes=b"LOGO",
-        mirror=True,
+        framing=_FRAMING,
     )
 
-    assert seen.get("mirror") is True
+    assert seen["framing"] == _FRAMING
 
 
 def test_generate_seasons_carries_mirror_into_each_poster(monkeypatch):
@@ -276,10 +286,10 @@ def test_generate_seasons_carries_mirror_into_each_poster(monkeypatch):
         title="T",
         tmdb_id=7,
         seasons=[1],
-        mirror=True,
+        framing=_FRAMING,
     )
 
-    assert seen.get("mirror") is True
+    assert seen["framing"] == _FRAMING
 
 
 @pytest.mark.parametrize(
diff --git a/tests/test_cl2k_renderer.py b/tests/test_cl2k_renderer.py
index d4fe1b3a..73a45925 100644
--- a/tests/test_cl2k_renderer.py
+++ b/tests/test_cl2k_renderer.py
@@ -321,7 +321,10 @@ def test_render_framed_art_dims_and_format():
 def test_render_framed_art_fit_letterboxes_on_black():
     # A square source contained in a 16:9 frame leaves black pillarboxes.
     blob = render_framed_art(
-        backdrop_bytes=_backdrop(w=500, h=500), width=1920, height=1080, fit_mode="fit"
+        backdrop_bytes=_backdrop(w=500, h=500),
+        width=1920,
+        height=1080,
+        framing=geo.Framing(fit_mode="fit"),
     )
     with Image(blob=blob) as img:
         bar = img[10, 540]  # left pillarbox
@@ -374,7 +377,8 @@ def test_psd_poster_layer_matches_renderer_framing():
     from backend.util.cl2k.psd_export import export_psd
 
     framed = frame_backdrop(
-        backdrop_bytes=_backdrop(), fit_mode="fit", v_pos=0.2, zoom=1.4
+        backdrop_bytes=_backdrop(),
+        framing=geo.Framing(fit_mode="fit", v_pos=0.2, zoom=1.4),
     )
     blob = export_psd(backdrop_bytes=framed, kind="movie", title="X")
     psd = PSDImage.open(io.BytesIO(blob))