diff --git a/.gitignore b/.gitignore
index f5fd260..83b39c3 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,4 +1,6 @@
 frames/
+# local extracted takes for comparing source cadences
+/scratch/
 *.task
 *.take
 
diff --git a/README.md b/README.md
index 58f8a7c..1b2e149 100644
--- a/README.md
+++ b/README.md
@@ -42,7 +42,7 @@ wasm, which is fetched from a CDN on first use.
 For real footage:
 
 ```sh
-./extract.sh /path/to/clip.mov 12   # -> frames/*.png, audio.wav, manifest.json
+./extract.sh /path/to/clip.mov      # -> frames/*.png, audio.wav, manifest.json
 ```
 
 then **Load frames**. MediaPipe's wasm is fetched from jsdelivr on first use;
@@ -52,16 +52,16 @@ Frames are pre-extracted rather than decoded in the page because browser video
 seeking is approximate and `requestVideoFrameCallback` only delivers frames at
 playback speed — neither gives a deterministic per-frame pass.
 
-`manifest.json` records the true extraction rate. The page reads it rather than
-assuming, because a guessed fps desynchronises audio from picture — and sync is
-the one thing this view exists to show.
+`manifest.json` records the source rate. The extractor keeps every source frame;
+the page reads that rate because a guessed fps desynchronises audio from picture.
+Choosing a lower picture rate happens after analysis.
 
-**Exposure** decides how often the picture gets a new drawing: rip at 24 and
-render `on 2s` for 12, `on 3s` for 8. The dense track and the audio are
-untouched, so it is a dropdown rather than a re-rip, and the export emits keys
-only on the grid instead of the same pose twice. Everything rides the same grid
-— mouth, eyes, teeth, plate — because a head cutting on the odd frames while the
-mouth cuts on the even ones reads as two performances laid over each other.
+**Picture fps** decides how often the finished roto gets a new pose. Analyze all
+source frames, then sample those frozen poses at 12, 24 or the source rate while
+keeping the original duration and audio. **Exposure** can hold a drawing across
+more than one picture slot. The tracing editor chooses source frames for cel
+references separately. Shared timing is the useful default for mouth, eyes,
+teeth and plate so their changes read as one performance.
 
 **Audio is the playback clock**: `frame = floor(audio.currentTime * fps)`. A slow
 render loop therefore drops frames instead of drifting, and ½x / ¼x work by
@@ -322,13 +322,13 @@ the tool a person made by hand; everything else regenerates. They are not in the
 ## Two kinds of sparseness
 
 Sparseness has two unrelated causes, and conflating them was the original design
-error here. **Aesthetic** sparseness is set by the extraction rate — pick 12fps and
-you have already chosen your timing. **Labour** sparseness is a human drawing
-each one, and it binds only on the plate.
+error here. **Aesthetic** sparseness is chosen from the full analyzed source
+track at rendering time. **Labour** sparseness is a human drawing each cel and
+selecting which source frames to use as tracing references.
 
-Aesthetic sparseness is the **exposure** control, not the extraction rate —
-making it a render-time grid means auditioning 12 against 24 costs a dropdown
-instead of a re-rip and a full re-detection.
+Aesthetic sparseness is the **picture fps** control, with exposure available for
+longer holds. Both happen after analysis, so auditioning 12 against 24 needs no
+re-extraction or re-detection.
 
 So the mouth keeps **every** frame: it is traced, and therefore free. In limited
 animation lip sync is routinely the densest element, on 1s, while heads hold on
diff --git a/docs/animation-model.md b/docs/animation-model.md
index f9c7e8a..1988fdf 100644
--- a/docs/animation-model.md
+++ b/docs/animation-model.md
@@ -333,11 +333,15 @@ Every node may map the frame it is evaluated at:
 ```clojure
 :time {:mode :inherit}                        ; the default, and almost always right
 :time {:mode :map :expose 2 :offset -1 :rate 1.0 :loop? false}
+:time {:mode :map :source-fps 30 :sample-fps 12} ; root: lower picture cadence
 ```
 
 Three features that look unrelated are this one mechanism:
 
 - **exposure** is `⌊f/n⌋·n`,
+- **picture fps** quantises source time to a chosen picture grid, then reads the
+  latest source pose at or before that time; source analysis and audio keep their
+  original cadence,
 - **mouth lead** is `f + k`,
 - **a symbol instance's timing** is `(f - at)·rate + in`, with optional looping.
 
@@ -463,8 +467,11 @@ world transform of the node it rides:
 
 Composed with image-pixels-to-local — **both axes divided by `imgH`**, never by
 their own dimension — the photo is registered with the shapes by construction,
-and an unregistered underlay is merely decorative. Which frame it shows, the
-current one or the held plate frame, is a UI choice and not a stored one.
+and an unregistered underlay is merely decorative. The tracing editor chooses
+which source frame to show under a cel. That reference choice is independent of
+the finished picture fps and does not change the dense analysis track. A cel can
+therefore use any useful source frame as its drawing reference, even when that
+frame is not one of the displayed picture poses.
 
 A photo that has to sit *between* two drawn layers is the case that would make it
 a `:bitmap` node with an op of its own. Nothing wants that yet: a reference is
@@ -527,6 +534,7 @@ Proof that it covers what exists, not just what is wanted:
 | painted background cel | node per layer, `[:geom :pts]` **framed**, `[:style :color]` framed |
 | `mouth lead` | `:time {:offset k}` on performance nodes only |
 | `exposure` | `:time {:expose n}` on the clip root, inherited |
+| picture fps | `:time {:source-fps s :sample-fps p}` on the clip root, applied after analysis |
 | hand correction | an `:over` layer, `:offset` or `:replace` |
 
 The brow row is the one worth looking at twice. `docs/design.md` argues at length
diff --git a/docs/architecture.md b/docs/architecture.md
index 175c736..18c4ab3 100644
--- a/docs/architecture.md
+++ b/docs/architecture.md
@@ -49,6 +49,11 @@ overrides and kept-frame sets are all in clip-frame space, so a clip slides on
 the timeline without a single stored number changing. Exposure and lead are
 transforms *within* clip space:
 
+Detection retains every source frame. A chosen picture fps samples the frozen
+roto in clip time; it changes neither source-frame count nor the audio clock.
+The set of source frames an artist uses as cel tracing references is another
+selection, independent of the picture fps.
+
 ```clojure
 (defn pose-frame [clip cf]
   (-> cf (expose (:exposure clip)) (shift (:lead clip) (count-frames clip))))
@@ -73,7 +78,7 @@ project
 │   ├── scene                node tree
 │   ├── channels             node+property -> keyframe stream, in cf
 │   ├── overrides            (node, property, cf) -> value, applied last
-│   └── cels                 painted vector layers, per exposure slot
+│   └── cels                 painted vector layers, keyed by chosen clip frames
 └── sequence[]               clip placements: {clip-id, at, in, out}
 ```
 
@@ -81,6 +86,10 @@ project
 it a name is most of the work: `state` in `app.js` is a clip with its analysis
 inlined and its palette global.
 
+Cel keys select where drawings begin and how long they hold. The source frames
+shown beneath a cel while tracing are chosen independently, and picture fps
+only controls which analyzed pose the finished roto displays at a given time.
+
 ### Two things called "track"
 
 `docs/design.md` says "dense track" for the landmark stream. A timeline also
diff --git a/docs/port-plan.md b/docs/port-plan.md
index 8dd0f7e..70763f4 100644
--- a/docs/port-plan.md
+++ b/docs/port-plan.md
@@ -4,8 +4,10 @@ Self-contained. You should not need any prior conversation to execute this.
 
 **Implementation status (2026-09-27):** steps 0–6 are in the CLJS frontend.
 Step 6 reads extracted footage from the manifest, detects landmarks with local
-MediaPipe assets, and runs the same freeze path as the synthetic take. Step 7 is
-next: eyes, brows and the pixel-derived mouth interior.
+MediaPipe assets at full source cadence, and runs the same freeze path as the
+synthetic take. The scene time map can sample the frozen roto at a lower picture
+fps without changing source analysis, duration or audio. Step 7 is next: eyes,
+brows and the pixel-derived mouth interior.
 
 ## What arthur is
 
@@ -302,6 +304,10 @@ MediaPipe interop behind one namespace; real frames, real audio, real fps from
 the manifest. **Vendor the wasm** rather than fetching from jsdelivr — it is
 currently the only thing in the tool that silently requires a network.
 
+Decode every source frame for analysis. A lower output picture fps is a time map
+over frozen channels, not a reduced detection track. Selecting source frames to
+trace into cels is independent again and remains outside this port's paint scope.
+
 **Done:** real footage plays back as a rotoscoped mouth.
 
 ### 7 — the rest of measure
diff --git a/extract.sh b/extract.sh
index cdbc9c6..0b276b3 100755
--- a/extract.sh
+++ b/extract.sh
@@ -7,32 +7,77 @@
 # pass. A PNG sequence is exact, instantly seekable, and reproducible.
 #
 # Audio comes out alongside because the page uses it as the PLAYBACK CLOCK -
-# frame = floor(audio.currentTime * fps) - so picture and sound cannot drift
-# apart no matter how long the shot is or how slow the render loop runs.
+# frame = floor(audio.currentTime * fps). Detection sees every decoded source
+# frame; a lower drawing rate is a later playback choice, never an extraction
+# choice. This script accepts CFR footage because frame-index timing needs a
+# single rate. VFR needs per-frame timestamps in the manifest first.
 set -euo pipefail
 
-src="${1:?usage: ./extract.sh CLIP [FPS] [OUTDIR]}"
-fps="${2:-12}"
-out="${3:-frames}"
-
-rm -rf "$out"; mkdir -p "$out"
-ffmpeg -hide_banner -loglevel warning -i "$src" -vf "fps=$fps" "$out/%04d.png"
-count=$(ls -1 "$out" | wc -l)
-
-# Mono is enough for judging sync and halves the file. Absent audio is not fatal.
-if ffprobe -v error -select_streams a:0 -show_entries stream=codec_type \
-     -of csv=p=0 "$src" 2>/dev/null | grep -q audio; then
-  ffmpeg -hide_banner -loglevel warning -y -i "$src" -vn -ac 1 -ar 44100 audio.wav
-  audio='"audio.wav"'
-  echo "audio -> audio.wav"
-else
-  audio='null'
-  echo "no audio stream"
+src="${1:?usage: ./extract.sh CLIP [BUNDLE_DIR]}"
+bundle="${2:-.}"
+if [[ "$bundle" =~ ^[0-9]+([.][0-9]+)?$ ]]; then
+  echo "The FPS argument was removed: extraction always keeps the source rate. Use a directory as argument 2." >&2
+  exit 2
+fi
+if [[ "$bundle" = /* || "$bundle" = *..* ]]; then
+  echo "BUNDLE_DIR must be a relative directory inside this repo" >&2
+  exit 2
 fi
 
-# The page must know the true extraction rate: if it guessed, audio and picture
-# would drift. Source of truth lives here, next to the frames it describes.
-printf '{"fps":%s,"frames":%s,"dir":"%s","audio":%s,"source":"%s"}\n' \
-  "$fps" "$count" "$out" "$audio" "$(basename "$src")" > manifest.json
+probe=$(ffprobe -v error -select_streams v:0 \
+  -show_entries stream=r_frame_rate,avg_frame_rate,nb_frames \
+  -of json "$src")
+fps=$(python3 -c '
+import json, sys
+from fractions import Fraction
+streams = json.load(sys.stdin).get("streams", [])
+if not streams:
+    raise SystemExit("no video stream in source")
+s = streams[0]
+nominal = Fraction(s["r_frame_rate"])
+average = Fraction(s["avg_frame_rate"])
+if nominal <= 0 or average <= 0 or abs(float(nominal / average) - 1) > 0.001:
+    raise SystemExit("variable-frame-rate source needs timestamp-aware playback; refusing to guess its fps")
+print(float(average))
+' <<< "$probe")
 
-echo "$count frames at ${fps}fps -> $out/  (manifest.json written)"
+if [[ "$bundle" = "." ]]; then
+  dir="frames"; audio_path="audio.wav"; manifest_path="manifest.json"
+else
+  dir="${bundle%/}/frames"
+  audio_path="${bundle%/}/audio.wav"
+  manifest_path="${bundle%/}/manifest.json"
+fi
+
+rm -rf "$dir"; mkdir -p "$dir"
+ffmpeg -hide_banner -loglevel warning -i "$src" -fps_mode passthrough "$dir/%04d.png"
+count=$(find "$dir" -maxdepth 1 -name '*.png' -type f | wc -l | tr -d ' ')
+
+expected=$(python3 -c 'import json,sys; print(json.load(sys.stdin)["streams"][0].get("nb_frames", ""))' <<< "$probe")
+if [[ "$expected" =~ ^[0-9]+$ && "$count" != "$expected" ]]; then
+  echo "decoded $count frames but source reports $expected; refusing an inaccurate manifest" >&2
+  exit 1
+fi
+
+# Mono is enough for judging sync and halves the file. A silent clock lets a
+# mute source use the same audio-driven transport.
+if ffprobe -v error -select_streams a:0 -show_entries stream=codec_type \
+     -of csv=p=0 "$src" 2>/dev/null | grep -q audio; then
+  ffmpeg -hide_banner -loglevel warning -y -i "$src" -vn -ac 1 -ar 44100 "$audio_path"
+else
+  duration=$(python3 -c 'import sys; print(int(sys.argv[1]) / float(sys.argv[2]))' "$count" "$fps")
+  ffmpeg -hide_banner -loglevel warning -y -f lavfi -i anullsrc=r=44100:cl=mono \
+    -t "$duration" -c:a pcm_s16le "$audio_path"
+fi
+
+# JSON escaping belongs to a JSON writer, especially for source filenames.
+python3 - "$fps" "$count" "$dir" "$audio_path" "$src" "$manifest_path" <<'PY'
+import json, os, sys
+fps, count, frames, audio, source, path = sys.argv[1:]
+with open(path, "w") as out:
+    json.dump({"fps": float(fps), "frames": int(count), "dir": frames,
+               "audio": audio, "source": os.path.basename(source)}, out)
+    out.write("\n")
+PY
+
+echo "$count source frames at ${fps}fps -> $dir/  ($manifest_path written)"
diff --git a/frontend/README.md b/frontend/README.md
index 9992540..33e6d1d 100644
--- a/frontend/README.md
+++ b/frontend/README.md
@@ -87,16 +87,32 @@ and real footage use `src/arthur/flow/take.cljs` for the measurement order and
 From the repo root, extract a clip, then click **load frames** in the CLJS app:
 
 ```sh
-./extract.sh /path/to/clip.mov 12
+./extract.sh /path/to/clip.mov
 ```
 
-This writes `frames/0001.png` onward, `audio.wav`, and `manifest.json` at the
-repo root. The manifest supplies the exact frame count, fps and audio path.
+This keeps every source frame and writes `frames/0001.png` onward, `audio.wav`,
+and `manifest.json` at the repo root. The manifest supplies the exact frame
+count, source fps and audio path. Variable frame rate sources are rejected until
+the manifest and clock carry per-frame timestamps.
+
+To keep multiple takes or compare with a previous extraction, pass a bundle
+directory and enter its manifest path in the app:
+
+```sh
+./extract.sh /path/to/clip.mov scratch/my-take
+# source manifest field: /scratch/my-take/manifest.json
+```
+
+`scratch/` is ignored by Git. The directory contains its own frames, audio and
+manifest, so extracting it does not replace another take's files.
 Loading detects one face per frame, freezes the measured mouth into channels,
 and adds a button for the footage clip. Detection happens once when you load;
 playback only resolves channels and paints. Frames without a detection remain
 marked absent even though their neighbouring poses are used to condition the
-track. The stage stays 320×200 regardless of the footage dimensions.
+track. The stage stays 320×200 regardless of the footage dimensions. Real
+footage starts at the source picture rate. The **picture fps** buttons sample the
+frozen roto at lower rates while the source track, duration and audio clock stay
+unchanged. Picking frames to trace into cels is a separate future editing step.
 
 MediaPipe's JS, wasm and model are under `public/mediapipe/` and served locally.
 No CDN is used by this app. See that directory's README for provenance.
diff --git a/frontend/public/index.html b/frontend/public/index.html
index 36021fc..782f667 100644
--- a/frontend/public/index.html
+++ b/frontend/public/index.html
@@ -31,6 +31,12 @@
       .scrub { width: 100%; margin: 10px 0 6px; }
       .readout { display: flex; gap: 18px; opacity: .55; font-size: 12px; }
       .readout .warn { color: #d98f5a; opacity: 1; }
+      .picture-rate { display: flex; align-items: center; gap: 6px; margin-top: 7px;
+                      font-size: 12px; }
+      .source-path { display: block; margin-top: 8px; font-size: 12px; opacity: .7; }
+      .source-path input { width: 300px; margin-left: 8px; padding: 3px 5px;
+                           color: var(--fg); background: #1c1f2b; border: 1px solid #2b3040;
+                           font: inherit; }
       .load-status { margin-top: 6px; font-size: 12px; opacity: .75; }
       .note { opacity: .35; font-size: 12px; max-width: 640px; }
     </style>
diff --git a/frontend/src/arthur/clock.cljs b/frontend/src/arthur/clock.cljs
index e5a849f..6b5dcd7 100644
--- a/frontend/src/arthur/clock.cljs
+++ b/frontend/src/arthur/clock.cljs
@@ -97,3 +97,8 @@
   rather than only inside the scene."
   [f expose]
   (node/expose f expose))
+
+(defn picture-frame
+  "The source pose displayed at f after picture-rate sampling and exposure."
+  [f source-fps picture-fps expose]
+  (node/expose (node/sample-frame f source-fps picture-fps) expose))
diff --git a/frontend/src/arthur/db.cljs b/frontend/src/arthur/db.cljs
index ccc3fe3..9e28ad7 100644
--- a/frontend/src/arthur/db.cljs
+++ b/frontend/src/arthur/db.cljs
@@ -23,7 +23,8 @@
   because there is one clip per scene today. Copying them by hand into this table
   is how one of them comes to disagree with the scene it describes."
   [label scene store]
-  (merge {:label label :scene scene :store store :audio "/audio.wav"}
+  (merge {:label label :scene scene :store store :audio "/audio.wav"
+          :display-fps (:fps scene)}
          (select-keys scene [:fps :frames :width :height])))
 
 (def scenes
@@ -50,9 +51,10 @@
    ;; footage's. That is what deleting `makeXform` buys — the framing became a
    ;; transform on a node, so nothing downstream of the freeze knows the frame
    ;; size — and it is why ui/player no longer hardcodes 320x200.
-   :clip    (select-keys (:take scenes) [:fps :frames :width :height :audio])
+   :clip    (select-keys (:take scenes) [:fps :frames :width :height :audio :display-fps])
 
-   :footage {:id nil :label nil :loading? false :status nil}
+   :footage {:id nil :label nil :loading? false :status nil
+             :manifest-path "/manifest.json"}
 
    ;; --- transport ---
    ;;
diff --git a/frontend/src/arthur/domain/node.cljs b/frontend/src/arthur/domain/node.cljs
index 66bf0e7..7603f8c 100644
--- a/frontend/src/arthur/domain/node.cljs
+++ b/frontend/src/arthur/domain/node.cljs
@@ -85,6 +85,22 @@
   [f n]
   (if (and n (> n 1)) (* (js/Math.floor (/ f n)) n) f))
 
+(defn sample-frame
+  "Pick a source frame for a lower picture rate without changing clip time.
+
+  The input is already an integer source frame from the audio clock. Its time is
+  f/source-fps. Quantise that time to the picture grid, then read the latest
+  source frame at or before it. The result is always an integer and never from
+  the future, including when the rates do not divide (30 source → 24 picture)."
+  [f source-fps picture-fps]
+  (if (and source-fps picture-fps
+           (pos? source-fps) (pos? picture-fps)
+           (< picture-fps source-fps))
+    (min f (js/Math.floor
+            (* (js/Math.floor (/ (* f picture-fps) source-fps))
+               (/ source-fps picture-fps))))
+    f))
+
 (defn local-frame
   "Apply a node's time map to the frame it was handed by its parent.
 
@@ -99,7 +115,8 @@
   as two performances; offset is PER-NODE by design, because mouth lead applies
   to performance nodes and not to the plate, which is the entire point of it."
   [n f]
-  (let [{:keys [mode offset rate] ex :expose :or {mode :inherit}} (:time n)]
+  (let [{:keys [mode offset rate source-fps sample-fps]
+         ex :expose :or {mode :inherit}} (:time n)]
     (if (= mode :inherit)
       f
       (do
@@ -110,7 +127,11 @@
         (when (and rate (not= rate 1.0) (not= rate 1))
           (throw (ex-info "time map :rate is symbol timing and symbols are not built (port-plan step 2 scope)"
                           {:node (:id n) :time (:time n)})))
+        (when (and sample-fps (not (and source-fps (pos? source-fps))))
+          (throw (ex-info "picture sampling needs a positive source fps"
+                          {:node (:id n) :time (:time n)})))
         (cond-> f
+          sample-fps (sample-frame source-fps sample-fps)
           ex     (expose ex)
           offset (+ offset))))))
 
@@ -256,4 +277,3 @@
         (into (for [[path c] (:channels n)
                     p        (ch/problems c)]
                 (str "channel " (pr-str path) ": " p))))))
-
diff --git a/frontend/src/arthur/events/footage.cljs b/frontend/src/arthur/events/footage.cljs
index f616128..3260ab6 100644
--- a/frontend/src/arthur/events/footage.cljs
+++ b/frontend/src/arthur/events/footage.cljs
@@ -43,13 +43,11 @@
 
 (defn- build-clip [manifest {:keys [dense detected dimensions missing first-real]}]
   (let [[w h] dimensions
-        params (merge take/knobs
-                      {:name "footage" :fps (:fps manifest) :aspect (/ w h)
-                       :stage [320 200] :expose 2 :head :as-filmed
-                       :analysis (str "mediapipe:1.0.1/" (:source manifest))})
-        frozen (take/build params {:dense dense :detected detected})
+        frozen (take/footage manifest {:dense dense :detected detected
+                                      :dimensions dimensions})
         scene (:scene frozen)]
     (assoc (select-keys scene [:fps :frames :width :height])
+           :display-fps (:fps scene)
            :scene scene :store (:store frozen)
            ;; Re-extraction often overwrites audio.wav under the same name. A new
            ;; URL makes the element fetch the new sound when this clip is loaded.
@@ -64,8 +62,8 @@
 
 (rf/reg-fx
  ::begin!
- (fn [_]
-   (-> (ingest/manifest!)
+ (fn [path]
+   (-> (ingest/manifest! path)
        (.then (fn [manifest]
                 (rf/dispatch [::progress "loading MediaPipe…"])
                 (-> (detect/landmarker!)
@@ -88,7 +86,11 @@
      {:db (assoc db :footage (assoc (:footage db) :loading? true
                                    :status "reading manifest.json…"))
       ::pb/pause! nil
-      ::begin! nil})))
+      ::begin! (get-in db [:footage :manifest-path])})))
+
+(rf/reg-event-db
+ ::set-manifest-path
+ (fn [db [_ path]] (assoc-in db [:footage :manifest-path] path)))
 
 (rf/reg-event-db
  ::progress
@@ -106,9 +108,9 @@
    (let [clip (store/entry id)]
      {:db (-> db
               (assoc :scene/current id
-                     :clip (select-keys clip [:fps :frames :width :height :audio])
-                     :footage {:id id :label (:label clip)
-                               :loading? false :status summary})
+                     :clip (select-keys clip [:fps :frames :width :height :audio :display-fps])
+                     :footage (assoc (:footage db) :id id :label (:label clip)
+                                     :loading? false :status summary))
               (assoc-in [:playback :frame] 0)
               (assoc-in [:playback :playing?] false))
       ::pb/pause! nil})))
diff --git a/frontend/src/arthur/events/playback.cljs b/frontend/src/arthur/events/playback.cljs
index 2b69389..af32d8e 100644
--- a/frontend/src/arthur/events/playback.cljs
+++ b/frontend/src/arthur/events/playback.cljs
@@ -65,6 +65,13 @@
    {:db (assoc-in db [:playback :rate] r)
     ::rate! r}))
 
+(rf/reg-event-db
+ ::set-picture-fps
+ (fn [db [_ target]]
+   (if (and (number? target) (pos? target) (<= target (fps db)))
+     (assoc-in db [:clip :display-fps] target)
+     db)))
+
 ;; --- effects: every DOM touch on the audio element is one of these ---
 
 (rf/reg-fx ::play!  (fn [_] (clock/play!)))
@@ -98,7 +105,7 @@
               ;; The stage travels with the clip: two clips may be different
               ;; sizes, and the raster the loop paints into is the clip's, not
               ;; the app's.
-              (assoc :clip (select-keys clip [:fps :frames :width :height :audio]))
+              (assoc :clip (select-keys clip [:fps :frames :width :height :audio :display-fps]))
               (assoc-in [:playback :frame] 0)
               (assoc-in [:playback :playing?] false))
       ::pause! nil
diff --git a/frontend/src/arthur/flow/freeze.cljs b/frontend/src/arthur/flow/freeze.cljs
index 4681479..0035185 100644
--- a/frontend/src/arthur/flow/freeze.cljs
+++ b/frontend/src/arthur/flow/freeze.cljs
@@ -275,6 +275,31 @@
 ;; ---------------------------------------------------------------------------
 ;; the face, onto the stage
 
+(defn- motion-placement
+  "An editable default framing for a real take. Fit the observed eye/nose and
+  mouth motion within the stage; otherwise a downward head move can push the
+  mouth below a fixed 320×200 canvas even while it stays in the source image."
+  [[w h] {:keys [rigid transforms outer detected]}]
+  (let [points (mapcat (fn [i]
+                         (when (or (nil? detected) (nth detected i))
+                           (concat (nth rigid i)
+                                   (geom/apply-sim-all (invert (nth transforms i))
+                                                       (nth outer i)))))
+                       (range (count outer)))
+        xs     (map :x points)
+        ys     (map :y points)
+        x0     (reduce min xs)
+        x1     (reduce max xs)
+        y0     (reduce min ys)
+        y1     (reduce max ys)
+        cx     (/ (+ x0 x1) 2)
+        cy     (/ (+ y0 y1) 2)
+        k      (min (/ (* 0.8 w) (max 1e-9 (- x1 x0)))
+                    (/ (* 0.8 h) (max 1e-9 (- y1 y0))))]
+    {[:xform :anchor] (ch/framed [cx cy])
+     [:xform :scale]  (ch/framed [k k])
+     [:xform :pos]    (ch/framed [(- (/ w 2) cx) (- (/ h 2) cy)])}))
+
 (defn face-placement
   "The face's transform on the stage, as FRAMED channels.
 
@@ -284,9 +309,14 @@
   ordinary node. `makeXform` made the same decision and then baked it into every
   vertex, where nothing could ever revise it.
 
-  Derived from the reference rigid configuration, which is what the freeze has:
-  the face oval is not measured, because its only consumers in the prototype were
-  that transform and the placeholder plate outline. Two numbers:
+  The synthetic take uses the reference rigid configuration for its default.
+  Real footage can request `:fit-motion?`: its default fits the observed rigid
+  and lip motion in the stage so a head move does not send the mouth off canvas.
+  Both are ordinary editable transforms on :face, never baked into the geometry.
+  The face oval is not measured, because its only consumers in the prototype were
+  the old baked framing transform and the placeholder plate outline.
+
+  For the synthetic default, two numbers:
 
   SCALE is stage pixels per image height, set so the reference's eye-corner span
   is 40% of the stage width. Landmark-free — it is the rigid configuration's own
@@ -304,15 +334,17 @@
   landmarks are eyes and nose — the upper middle of a face — so a quarter down
   leaves the jaw and the mouth on the stage. Whatever hangs off is clipped, which
   is not a feature to add: every fill in `domain/raster` clamps already."
-  [{:keys [stage]} {:keys [ref]}]
-  (let [[w h] stage
-        c     (geom/centroid ref)
-        span  (- (reduce max (map :x ref)) (reduce min (map :x ref)))
-        k     (/ (* 0.4 w) span)]
-    {[:xform :anchor] (ch/framed [(:x c) (:y c)])
-     [:xform :scale]  (ch/framed [k k])
-     [:xform :pos]    (ch/framed [(- (/ w 2) (:x c))
-                                  (- (* 0.25 h) (:y c))])}))
+  [{:keys [stage fit-motion?]} {:keys [ref] :as inputs}]
+  (if fit-motion?
+    (motion-placement stage inputs)
+    (let [[w h] stage
+          c     (geom/centroid ref)
+          span  (- (reduce max (map :x ref)) (reduce min (map :x ref)))
+          k     (/ (* 0.4 w) span)]
+      {[:xform :anchor] (ch/framed [(:x c) (:y c)])
+       [:xform :scale]  (ch/framed [k k])
+       [:xform :pos]    (ch/framed [(- (/ w 2) (:x c))
+                                    (- (* 0.25 h) (:y c))])})))
 
 ;; ---------------------------------------------------------------------------
 ;; the clip
@@ -327,6 +359,8 @@
                    `(count outer)`, because a freeze that could disagree with its
                    own input about the length of the take would.
     :stage [w h]   project dimensions, INDEPENDENT of the footage
+    :fit-motion?   choose an editable real-footage default that keeps the
+                   observed feature motion on stage
     :expose        the clip root's exposure grid, inherited by everything
     :verts         the lip rings' vertex budget
     :aperture-cut  fraction of the take's peak aperture below which the mouth
diff --git a/frontend/src/arthur/flow/ingest.cljs b/frontend/src/arthur/flow/ingest.cljs
index dcf47db..1382ecc 100644
--- a/frontend/src/arthur/flow/ingest.cljs
+++ b/frontend/src/arthur/flow/ingest.cljs
@@ -13,8 +13,8 @@
     (assoc m :fps fps :frames frames)))
 
 (defn manifest!
-  []
-  (-> (js/fetch "/manifest.json" #js {:cache "no-store"})
+  [path]
+  (-> (js/fetch (str "/" (str/replace path #"^/+" "")) #js {:cache "no-store"})
       (.then (fn [response]
                (when-not (.-ok response)
                  (throw (ex-info "manifest.json was not found; run extract.sh first"
diff --git a/frontend/src/arthur/flow/take.cljs b/frontend/src/arthur/flow/take.cljs
index c3ac5c0..414148a 100644
--- a/frontend/src/arthur/flow/take.cljs
+++ b/frontend/src/arthur/flow/take.cljs
@@ -24,3 +24,16 @@
 (defn build
   [params inputs]
   (freeze/clip params (measure params inputs)))
+
+(defn footage
+  "A real manifest and its detected landmarks through the same measurement and
+  freeze path as the synthetic take. The source cadence stays in :fps; picture
+  sampling is a root time map applied only after this artifact exists."
+  [manifest {:keys [dense detected dimensions]}]
+  (let [[w h] dimensions
+        params (merge knobs
+                      {:name "footage" :fps (:fps manifest) :aspect (/ w h)
+                       :stage [320 200] :fit-motion? true
+                       :expose 1 :head :as-filmed
+                       :analysis (str "mediapipe:1.0.1/" (:source manifest))})]
+    (build params {:dense dense :detected detected})))
diff --git a/frontend/src/arthur/subs/playback.cljs b/frontend/src/arthur/subs/playback.cljs
index 451f365..41f7102 100644
--- a/frontend/src/arthur/subs/playback.cljs
+++ b/frontend/src/arthur/subs/playback.cljs
@@ -10,6 +10,7 @@
 (rf/reg-sub ::loop?    (fn [db _] (get-in db [:playback :loop?])))
 (rf/reg-sub ::muted?   (fn [db _] (get-in db [:playback :muted?])))
 (rf/reg-sub ::fps      (fn [db _] (get-in db [:clip :fps])))
+(rf/reg-sub ::display-fps (fn [db _] (get-in db [:clip :display-fps])))
 (rf/reg-sub ::frames   (fn [db _] (get-in db [:clip :frames])))
 ;; The stage, in pixels. On the clip because project dimensions are independent
 ;; of the footage — see flow/freeze/face-placement — so the canvas and the raster
diff --git a/frontend/src/arthur/subs/render.cljs b/frontend/src/arthur/subs/render.cljs
index 12dbb5e..5a54881 100644
--- a/frontend/src/arthur/subs/render.cljs
+++ b/frontend/src/arthur/subs/render.cljs
@@ -10,11 +10,28 @@
   (:require [arthur.domain.palette :as pal]
             [arthur.domain.scene :as scene]
             [arthur.footage.store :as footage]
+            [arthur.subs.playback :as playback]
             [re-frame.core :as rf]))
 
 (rf/reg-sub ::scene-id (fn [db _] (:scene/current db)))
 
-(rf/reg-sub ::scene (fn [db _] (:scene (footage/entry (:scene/current db)))))
+(rf/reg-sub
+ ::base-scene
+ :<- [::scene-id]
+ (fn [id _] (:scene (footage/entry id))))
+
+(rf/reg-sub
+ ::scene
+ :<- [::base-scene]
+ :<- [::playback/display-fps]
+ (fn [[scene picture-fps] _]
+   ;; This changes only the scene's root time map. The dense source track stays
+   ;; at its native rate, and the audio clock still advances through source time.
+   (if (and scene picture-fps (< picture-fps (:fps scene)))
+     (-> scene
+         (assoc-in [:nodes :root :time :source-fps] (:fps scene))
+         (assoc-in [:nodes :root :time :sample-fps] picture-fps))
+     scene)))
 
 (rf/reg-sub
  ::exposure
@@ -45,11 +62,12 @@
 
 (rf/reg-sub
  ::store
- (fn [db _]
+ :<- [::scene-id]
+ (fn [id _]
    ;; Tier 2, behind a handle, and never in app-db itself — what is in the db is
    ;; the id of the clip whose blocks these are. The hand-written demo has none;
    ;; the swarm is entirely dense.
-   (:store (footage/entry (:scene/current db)))))
+   (:store (footage/entry id))))
 
 (rf/reg-sub
  ::resolver
diff --git a/frontend/src/arthur/ui/player.cljs b/frontend/src/arthur/ui/player.cljs
index 308395e..0c67c0e 100644
--- a/frontend/src/arthur/ui/player.cljs
+++ b/frontend/src/arthur/ui/player.cljs
@@ -167,12 +167,18 @@
         ;; the next one lands where the audio already is, so the failure mode is
         ;; a visible stutter rather than an invisible slide out of sync.
         f      (if live? (clock/frame fps frames) frame)]
+    ;; A paused scrub is an explicit seek, not a playback sample. Clear the
+    ;; meter before the next play so its first frame cannot count the seek as
+    ;; dropped frames or leave a stale low rate in the transport.
+    (when (and (not live?) (:t0 @meter-state))
+      (reset! meter-state {:t0 nil :paints 0 :samples 0 :advanced 0 :prev nil})
+      (reset! meter {:fps 0 :drop 0}))
     ;; `ready?` gates the bookkeeping as well as the draw: recording a frame as
     ;; painted when it was not is how the canvas stays empty forever.
     (when (and ready? f (not= f (:last @state)))
       (swap! state assoc :last f)
       (paint! f)
-      (meter! f)
+      (when live? (meter! f))
       (when live? (rf/dispatch [::pb/tick f])))
     ;; The audio ending is the authority on playback having stopped; nothing
     ;; counts frames to notice it.
diff --git a/frontend/src/arthur/ui/shell.cljs b/frontend/src/arthur/ui/shell.cljs
index d43ba80..924ce42 100644
--- a/frontend/src/arthur/ui/shell.cljs
+++ b/frontend/src/arthur/ui/shell.cljs
@@ -35,8 +35,10 @@
         frame    @(rf/subscribe [::sub/frame])
         frames   @(rf/subscribe [::sub/frames])
         fps      @(rf/subscribe [::sub/fps])
+        picture-fps @(rf/subscribe [::sub/display-fps])
+        current  @(rf/subscribe [::render/scene-id])
         expose   @(rf/subscribe [::render/exposure])
-        {:keys [id label loading? status]} @(rf/subscribe [::sub/footage])]
+        {:keys [id label loading? status manifest-path]} @(rf/subscribe [::sub/footage])]
     [:div.transport
      [:div.row
       [:button {:on-click #(rf/dispatch [::pb/toggle])}
@@ -77,8 +79,9 @@
        :on-change #(rf/dispatch [::pb/seek (js/parseInt (.. % -target -value) 10)])}]
      [:div.readout
       [:span (str "frame " frame " / " frames)]
-      [:span (str fps " fps")]
-      [:span (str "exposure " expose " → holds " (clock/exposed-frame frame expose))]
+      [:span (str "source " fps " fps")]
+      [:span (str "picture " picture-fps " fps")]
+      [:span (str "pose " (clock/picture-frame frame fps picture-fps expose))]
       [:span (str (js/Math.round (* 100 rate)) "%")]
       ;; Measured in the loop, not derived from the clock: the whole question
       ;; while profiling is whether the painting keeps up with the clock, so a
@@ -88,6 +91,19 @@
          (str (.toFixed (or fps 0) 1) " paint/s"
               (when (and drop (pos? drop))
                 (str " · " (.toFixed drop 2) " frames/paint")))])]
+     (when (= id current)
+       [:div.picture-rate
+        [:span "picture fps "]
+        (doall
+         (for [r (distinct (filter #(<= % fps) [8 12 15 24 fps]))]
+           ^{:key r}
+           [:button {:class (when (= r picture-fps) "on")
+                     :on-click #(rf/dispatch [::pb/set-picture-fps r])}
+            (if (= r fps) "source" (str r))]))])
+     [:label.source-path "source manifest "
+      [:input {:type "text" :value manifest-path :disabled loading?
+               :on-change #(rf/dispatch [::footage/set-manifest-path
+                                         (.. % -target -value)])}]]
      (when status [:div.load-status status])]))
 
 (defn- stage []