diff --git a/movie_editor/backend/builder.py b/movie_editor/backend/builder.py
index 44676c34..b6f61c61 100644
--- a/movie_editor/backend/builder.py
+++ b/movie_editor/backend/builder.py
@@ -220,9 +220,11 @@ def build(object_info: dict, models_config: dict, params: dict, media: dict | No
         graph["sampler"]["inputs"]["seed"] = params["seed"]
     if params.get("max_scenes") is not None:
         graph["sampler"]["inputs"]["max_scenes"] = params["max_scenes"]
-    # the result must be written to the output dir so the editor can fetch it back
-    # (the reference graph used preview-only save_output=False).
-    graph["vhs"]["inputs"]["save_output"] = True
+    # UI renders are EPHEMERAL: write to ComfyUI's temp dir (cleared on restart), not
+    # the output dir. Persisting a result is the user's job via Export (Save dialog).
+    # save_output=False routes VHS_VideoCombine to temp; history reports type="temp",
+    # which the result proxy + final-render concat resolve against the temp directory.
+    graph["vhs"]["inputs"]["save_output"] = False
     # split_by_transitions MUST be True — Studio defaults to False (single-scene mode)
     # but the Movie Editor always builds multi-scene combined prompts.  Force it here
     # BEFORE the user's studio_inputs override block so it cannot be accidentally
diff --git a/movie_editor/frontend/store.js b/movie_editor/frontend/store.js
index 5dbdffaa..721d1c9d 100644
--- a/movie_editor/frontend/store.js
+++ b/movie_editor/frontend/store.js
@@ -216,13 +216,14 @@
   // ── sync scenes from preview (distribute parsed anchor/transitions back) ──────
   function syncFromPreview() {
     if (!state.project || !state.preview) return;
-    // raw = verbatim split (shortcuts preserved); exp = expanded (transition detection)
+    // structure from whichever split defines more scenes; verbatim text on a tie.
     const raw = state.preview.parsed_raw || state.preview.parsed || {};
     const exp = state.preview.parsed || raw;
-    const parsedScenes = raw.scenes || [];
+    const src = _structureSource(raw, exp);
+    const parsedScenes = src.scenes || [];
     if (!parsedScenes.length) return;
 
-    if (raw.anchor != null) state.project.anchor = raw.anchor;
+    if (src.anchor != null) state.project.anchor = src.anchor;
 
     // Sync scene texts verbatim (only if count matches to avoid destructive mismatches)
     const activeScenes = state.project.scenes.filter((s) => !s.excluded);
@@ -231,7 +232,7 @@
     }
 
     // Transitions: only fill seams the user hasn't set, mapped via the library.
-    const trans = (((exp.scenes || []).length === parsedScenes.length ? exp.transitions : raw.transitions) || []);
+    const trans = (((exp.scenes || []).length === parsedScenes.length ? exp.transitions : src.transitions) || []);
     trans.forEach((t) => {
       const idx = t.after_scene;
       if (idx >= 0 && idx < activeScenes.length && t.visual_effect) {
@@ -257,30 +258,40 @@
     let res;
     try { res = await API.parsePrompt(state.project.id, text); }
     catch (e) { alert("Could not parse the global prompt: " + e.message); return; }
-    // raw = verbatim split (shortcuts kept in scene text); exp = expanded split, used
-    // ONLY to detect transitions (incl. those a shortcut expands into).
-    const raw = res.parsed_raw || res.parsed;
+    // raw = verbatim split (shortcuts kept in text); exp = expanded split. The
+    // STRUCTURE (scene count / anchor / transitions) can be created by shortcut
+    // expansion, so take it from whichever splits MORE. Verbatim text is kept only
+    // when the raw split structures the prompt just as well as the expanded one.
+    const raw = res.parsed_raw || res.parsed || {};
     const exp = res.parsed || raw;
-    if (!raw || !(raw.scenes || []).length) { alert("Nothing parsed — no scenes detected."); return; }
+    const src = _structureSource(raw, exp);
+    if (!(src.scenes || []).length) { alert("Nothing parsed — no scenes detected."); return; }
 
-    if (raw.anchor != null) state.project.anchor = raw.anchor;
+    if (src.anchor != null) state.project.anchor = src.anchor;
     const old = state.project.scenes || [];
-    const next = (raw.scenes || []).map((ps, i) => {
-      const base = old[i] ? JSON.parse(JSON.stringify(old[i])) : { source: { type: "empty" }, excluded: false };
-      base.text = ps.text || "";          // verbatim — shortcuts preserved
+    const next = (src.scenes || []).map((ps, i) => {
+      const base = old[i] ? JSON.parse(JSON.stringify(old[i])) : { source: { type: "carry" }, excluded: false };
+      base.text = ps.text || "";
       base.transition_to_next = "";
       return base;
     });
-    _applyDetectedTransitions(next, raw, exp);
+    _applyDetectedTransitions(next, src, exp);
     state.project.scenes = next;
     state.selectedSceneId = null;  // show project view so the result is visible
     notify(); scheduleSave();
   }
 
+  // Pick the split that defines the most scenes — prefer raw (verbatim) on a tie so
+  // descriptive shortcuts stay unexpanded; fall back to expanded when it splits more.
+  function _structureSource(raw, exp) {
+    const rawN = (raw.scenes || []).length, expN = (exp.scenes || []).length;
+    return (rawN >= expN && rawN > 0) ? raw : exp;
+  }
+
   // Map detected transitions (visual_effect) onto scene seams via the library. Prefers
   // the expanded split (catches shortcut→transition) when its scene count matches.
-  function _applyDetectedTransitions(scenes, raw, exp) {
-    const trans = (((exp.scenes || []).length === scenes.length ? exp.transitions : raw.transitions) || []);
+  function _applyDetectedTransitions(scenes, src, exp) {
+    const trans = (((exp.scenes || []).length === scenes.length ? exp.transitions : src.transitions) || []);
     trans.forEach((t) => {
       const idx = t.after_scene;
       if (idx >= 0 && idx < scenes.length && t.visual_effect) {
@@ -425,6 +436,40 @@
     set({ gen: { state: "done", promptId: null, media: state.gen.media, msg: `${runs.length} run(s) generated — use Render Final Video to stitch them.` } });
   }
 
+  // Fetch a render and offer a Save dialog (File System Access API where available,
+  // else a normal download). Renders live in ComfyUI's temp dir, so persisting one
+  // is the user's job — this is how they do it.
+  async function _saveBlobAs(url, name) {
+    let blob;
+    try { const r = await fetch(url); if (!r.ok) throw new Error(r.statusText); blob = await r.blob(); }
+    catch (e) { alert("Could not fetch the render: " + e.message); return; }
+    if (window.showSaveFilePicker) {
+      try {
+        const h = await window.showSaveFilePicker({
+          suggestedName: name,
+          types: [{ description: "MP4 video", accept: { "video/mp4": [".mp4"] } }],
+        });
+        const w = await h.createWritable(); await w.write(blob); await w.close();
+        return;
+      } catch (e) { if (e && e.name === "AbortError") return; }  // cancelled, or unsupported → fall through
+    }
+    const a = document.createElement("a");
+    a.href = URL.createObjectURL(blob); a.download = name;
+    document.body.appendChild(a); a.click(); a.remove();
+    setTimeout(() => URL.revokeObjectURL(a.href), 4000);
+  }
+
+  // Export the render covering the selected clip via a Save dialog.
+  async function exportSelected() {
+    if (!state.project) return;
+    const id = state.selectedSceneId;
+    if (!id) { alert("Select a clip to export."); return; }
+    const seg = (state.renderedSegments || []).find((s) => s.media && (s.sceneIds || []).includes(id));
+    if (!seg) { alert("That clip hasn't been generated yet — generate it first."); return; }
+    const idx = state.project.scenes.findIndex((s) => s.id === id);
+    await _saveBlobAs(API.resultUrl(state.project.id, seg.media), `scene_${idx >= 0 ? idx + 1 : "x"}.mp4`);
+  }
+
   // Stitch the rendered runs (hard cut, video + audio) into one final file.
   async function renderFinal() {
     if (!state.project) return;
@@ -439,7 +484,9 @@
       const r = await API.renderFinal(state.project.id, clips);
       // Show the stitched file across the whole timeline in the program monitor.
       state.renderedSegments = [{ sceneIds: order, media: r.media }];
-      set({ gen: { state: "done", promptId: null, media: [r.media], msg: `Final video rendered from ${r.clips} run(s).` } });
+      set({ gen: { state: "done", promptId: null, media: [r.media], msg: `Final video rendered from ${r.clips} run(s) — saving…` } });
+      const name = (state.project.name || "montage").replace(/\s+/g, "_") + ".mp4";
+      await _saveBlobAs(API.resultUrl(state.project.id, r.media), name);
     } catch (e) {
       set({ gen: { state: "error", promptId: null, media: [], msg: "Render failed: " + e.message } });
     }
@@ -547,7 +594,7 @@
     refreshProjectList, loadProject, newProject, deleteProject, downloadProject, importProject,
     patchProject, patchProjectQuiet, patchScene, patchSceneQuiet, selectScene, addScene, removeScene, moveScene, scene,
     resizeScene, splitScene, snapFrames,
-    refreshPreview, syncFromPreview, applyGlobalPrompt, generate, generateMontage, renderFinal, loadModels, loadImageTargets, setModelInput, setModelLink,
+    refreshPreview, syncFromPreview, applyGlobalPrompt, generate, generateMontage, renderFinal, exportSelected, loadModels, loadImageTargets, setModelInput, setModelLink,
     setConditioningSlot, setSamplerSlot, setSamplerInput, setSamplerInputNow, setStudioInput, setStudioInputNow,
     loadMedia, uploadMedia, deleteMedia, assignMediaToScene,
     loadShortcuts, saveShortcut, deleteShortcut, importShortcuts, loadTransitions, saveTransition, deleteTransition, importTransitions,
diff --git a/movie_editor/frontend/styles.css b/movie_editor/frontend/styles.css
index 5903807c..09d67da8 100644
--- a/movie_editor/frontend/styles.css
+++ b/movie_editor/frontend/styles.css
@@ -346,6 +346,9 @@ input[type=number] { font-family: var(--font-mono); }
 .clip.selected::before { background: var(--teal); opacity: 1; }
 .clip.excluded { opacity: .42; }
 .clip.excluded::before { background: var(--faint); }
+/* a clip that has a rendered video available (in temp) */
+.clip.rendered::after { content: "●"; position: absolute; top: 4px; right: 6px; font-size: 7px;
+  color: #57c98e; text-shadow: 0 0 5px rgba(87,201,142,.7); pointer-events: none; z-index: 4; }
 .clip.trimming { border-color: var(--amber); z-index: 5; }
 .clip-head { display: flex; align-items: center; gap: 6px; }
 .clip-no { font-family: var(--font-mono); font-size: 11px; color: var(--amber); font-weight: 600; }
diff --git a/movie_editor/frontend/timeline.js b/movie_editor/frontend/timeline.js
index eeb433e9..16cb6491 100644
--- a/movie_editor/frontend/timeline.js
+++ b/movie_editor/frontend/timeline.js
@@ -25,6 +25,9 @@
   const sFrames = (sc, p) => ((sc.frames_mode !== "project" && sc.frames != null) ? sc.frames : p.num_frames_per_scene) || 1;
   const sDur = (sc, p) => sFrames(sc, p) / sFps(sc, p);
 
+  const hasRender = (st, sceneId) =>
+    (st.renderedSegments || []).some((s) => s.media && (s.sceneIds || []).includes(sceneId));
+
   const p2 = (n) => String(n).padStart(2, "0");
   function timecode(sec, fps) {
     sec = Math.max(0, sec);
@@ -70,7 +73,7 @@
 
   // ── clip ───────────────────────────────────────────────────────────────────────
   function clipEl(st, p, scene, index, leftPx, widthPx) {
-    const clip = el("div", "clip" + (scene.id === st.selectedSceneId ? " selected" : "") + (scene.excluded ? " excluded" : ""));
+    const clip = el("div", "clip" + (scene.id === st.selectedSceneId ? " selected" : "") + (scene.excluded ? " excluded" : "") + (hasRender(st, scene.id) ? " rendered" : ""));
     clip.style.left = leftPx + "px";
     clip.style.width = Math.max(widthPx, 8) + "px";
     clip.onclick = () => S.selectScene(scene.id);
@@ -191,7 +194,11 @@
     const del = el("button", "btn ghost tiny danger", "✕ Remove");
     del.title = "Remove the selected clip (Delete / Backspace)"; del.disabled = !hasSel;
     del.onclick = () => { if (st.selectedSceneId) S.removeScene(st.selectedSceneId); };
-    bar.append(split); bar.append(del);
+    const exp = el("button", "btn ghost tiny", "⤓ Export");
+    exp.title = "Save the selected clip's rendered video to disk (renders are temporary)";
+    exp.disabled = !(hasSel && hasRender(st, st.selectedSceneId));
+    exp.onclick = () => S.exportSelected();
+    bar.append(split); bar.append(del); bar.append(exp);
 
     const spacer = el("div", "tl-spacer"); bar.append(spacer);
     const keys = el("span", "tl-keys", "S split · ⌫ remove"); keys.title = "Select a clip, then: S splits it at the playhead · Delete/Backspace removes it";
diff --git a/movie_editor/server.py b/movie_editor/server.py
index 678cac52..f0229ae7 100644
--- a/movie_editor/server.py
+++ b/movie_editor/server.py
@@ -535,8 +535,10 @@ if web is not None and PromptServer is not None:
         if missing:
             return web.json_response({"detail": f"{len(missing)} clip file(s) not found on disk — regenerate then render."}, status=400)
 
+        # The final render is ALSO ephemeral (temp dir) — the user persists it via the
+        # Export Save dialog, not by us writing to the output directory.
         out_name = f"funpack_final_{int(_time.time())}.mp4"
-        out_path = os.path.join(outdir, out_name)
+        out_path = os.path.join(tempdir, out_name)
         # concat demuxer + re-encode → tolerates differing codecs/params between runs
         # and keeps BOTH the video and the LTXAV audio stream.
         list_fd, list_path = tempfile.mkstemp(suffix=".txt")
@@ -561,7 +563,7 @@ if web is not None and PromptServer is not None:
             except OSError:
                 pass
         return web.json_response({
-            "media": {"filename": out_name, "subfolder": "", "type": "output", "kind": "videos"},
+            "media": {"filename": out_name, "subfolder": "", "type": "temp", "kind": "videos"},
             "clips": len(paths),
         })