From a1b2eec52346c5a6cefb72d76f05ae5f175979b2 Mon Sep 17 00:00:00 2001
From: digital-garbage <133010426+digital-garbage@users.noreply.github.com>
Date: Thu, 11 Jun 2026 16:10:23 +0300
Subject: [PATCH] Add scene vs video clips with convert and archive round-trip.

Generative scenes can lock to imported video (skipped on Generate); video clips can convert back to v2v with full state restored from scene_archive.
---
 movie_editor/backend/timeline.py      |  25 +++-
 movie_editor/frontend/inspector.js    |  78 ++++++++++-
 movie_editor/frontend/mediabrowser.js |   6 +-
 movie_editor/frontend/player.js       |  42 ++++--
 movie_editor/frontend/store.js        | 190 ++++++++++++++++++++++++--
 movie_editor/frontend/styles.css      |   6 +
 movie_editor/frontend/timeline.js     | 106 ++++++++++++--
 movie_editor/frontend/view_bus.js     |   2 +-
 movie_editor/server.py                |   9 +-
 movie_editor/tests/test_timeline.py   |  18 +++
 10 files changed, 433 insertions(+), 49 deletions(-)

diff --git a/movie_editor/backend/timeline.py b/movie_editor/backend/timeline.py
index 03a3a5d6..ffd1fb77 100644
--- a/movie_editor/backend/timeline.py
+++ b/movie_editor/backend/timeline.py
@@ -382,6 +382,15 @@ def source_type(scene: Scene) -> str:
     return (scene.source.type if scene.source else None) or "carry"
 
 
+def is_video_clip(scene: Scene) -> bool:
+    """Locked timeline video — plays as-is, excluded from generation and global prompt."""
+    return source_type(scene) == "video"
+
+
+def is_generative_scene(scene: Scene) -> bool:
+    return not is_video_clip(scene)
+
+
 def solo_run_wants_prior_guides(scene: Scene) -> bool:
     """Solo i2v runs: only mixed borrows prior-scene guides; image/empty/generated_frame are anchor-only."""
     return is_mixed_source(scene)
@@ -420,8 +429,11 @@ def build_scene_anchors_payload(project: Project) -> Optional[dict]:
 @dataclass
 class SceneSource:
     """How a scene's latent is born. V1 ignores this (uniform chain); Phase 2 maps
-    it onto EmptyLTXVLatent (empty/t2v) or LTXV Image to Video (image/i2v)."""
-    type: str = "carry"  # "carry" | "empty" | "image" | "generated_frame" | "mixed"
+    it onto EmptyLTXVLatent (empty/t2v) or LTXV Image to Video (image/i2v).
+
+    ``video`` = locked NLE clip (imported or converted from a scene); never generated.
+    ``v2v`` = generative scene using a video file as the source (video-to-video)."""
+    type: str = "carry"  # carry | empty | image | generated_frame | mixed | video | v2v
     media_ref: Optional[str] = None              # asset id — image/mixed anchor (Img2Video)
     frame_ref: Optional[dict[str, Any]] = None   # {scene_id, frame_idx}, for "generated_frame"
     target: Optional[str] = None                 # wire dest for the image: "port:<id>" | "node:<slotId>:<input>"
@@ -493,6 +505,9 @@ class Scene:
     # Source trim inside generated media (slip edit): in-point seconds and optional duration.
     source_in: float = 0.0
     source_dur: Optional[float] = None
+    # Saved generative state when this clip is converted to a locked ``video`` clip.
+    # Restored on Convert to scene (scene → video → scene round-trip).
+    scene_archive: Optional[dict] = None
 
     @staticmethod
     def from_dict(d: dict) -> "Scene":
@@ -520,6 +535,7 @@ class Scene:
             character_ids=[str(x) for x in (d.get("character_ids") or []) if x],
             source_in=float(d.get("source_in") or 0),
             source_dur=d.get("source_dur"),
+            scene_archive=d.get("scene_archive"),
         )
 
     def to_dict(self) -> dict:
@@ -707,7 +723,10 @@ def build_combined_prompt(project: Project, include_excluded: bool = False,
     transition_to_next, else the generic 'scene N' label) to force the split. This affects
     only what's sent to Studio, never the displayed global prompt.
     """
-    scenes = [s for s in project.scenes if include_excluded or not s.excluded]
+    scenes = [
+        s for s in project.scenes
+        if (include_excluded or not s.excluded) and is_generative_scene(s)
+    ]
     units = group_generative_units(scenes)
     char_map = _load_character_map()
     parts: list[str] = []
diff --git a/movie_editor/frontend/inspector.js b/movie_editor/frontend/inspector.js
index 8e72c59c..8bd5aca1 100644
--- a/movie_editor/frontend/inspector.js
+++ b/movie_editor/frontend/inspector.js
@@ -9,6 +9,7 @@
     ["empty", "Empty · text-to-video"],
     ["image", "Image · i2v anchor"],
     ["generated_frame", "From generated frame"],
+    ["v2v", "Video · v2v source"],
     ["carry", "Carry i2v guide · continue previous"],
     ["mixed", "Mixed · Img2Video anchor + prior guides"],
   ];
@@ -143,6 +144,71 @@
     return inner;
   }
 
+  function renderVideoSource(st, scene, parent, label) {
+    parent = parent || body;
+    if (!window.MediaPicker) return;
+    const pick = window.MediaPicker.create({
+      mediaBin: st.mediaBin,
+      value: scene.source?.media_ref,
+      filter: (m) => m.kind === "video",
+      noneLabel: "— pick video —",
+      onChange: (mediaRef) => {
+        S.patchScene(scene.id, {
+          source: { ...(scene.source || {}), type: scene.source?.type || "v2v", media_ref: mediaRef },
+        });
+      },
+    });
+    parent.append(field(label || "V2V source video", pick));
+  }
+
+  function renderVideoClip(st, scene) {
+    title.textContent = "Video clip";
+    body.append(el("div", "insp-hint", "Imported or converted video — plays as-is. Not included in Generate or the global prompt montage."));
+    const ref = scene.source?.media_ref;
+    const asset = ref ? (st.mediaBin || []).find((m) => m.id === ref) : null;
+    body.append(field("Source", el("span", null, asset?.name || ref || "From last render")));
+    renderOutgoingSeam(st, scene);
+
+    const actions = el("div", "insp-block");
+    const toScene = el("button", "btn primary", "Convert to scene");
+    toScene.title = scene.scene_archive
+      ? "Restore prompt, source, guides, and settings from before this was locked as video"
+      : "Make this a generative v2v scene (prompt + Generate)";
+    toScene.onclick = () => S.convertToScene(scene.id);
+    actions.append(toScene);
+    const del = el("button", "btn danger", "Delete");
+    del.style.marginLeft = "8px";
+    del.onclick = () => S.removeScene(scene.id);
+    actions.append(del);
+    body.append(actions);
+
+    foldSection("More editing", false, (more) => {
+      const fx = scene.effects || {};
+      const patchFx = (k, v, quiet) => {
+        const next = { ...(scene.effects || {}), [k]: v };
+        quiet ? S.patchSceneQuiet(scene.id, { effects: next }) : S.patchScene(scene.id, { effects: next });
+      };
+      const blur = el("input"); blur.type = "range"; blur.min = "0"; blur.max = "1"; blur.step = "0.05";
+      blur.value = fx.blur || 0;
+      blur.oninput = () => patchFx("blur", parseFloat(blur.value), true);
+      more.append(field(`Blur (${Math.round((fx.blur || 0) * 100)}%)`, blur));
+      const lenRow = el("div", "fields-row");
+      lenRow.append(lengthControl(scene, "frames"));
+      lenRow.append(lengthControl(scene, "fps"));
+      more.append(lenRow);
+      const r = (st.sceneRenders || {})[scene.id];
+      if (r?.media || ref) {
+        const srcTag = el("div", "insp-tag"); srcTag.textContent = "Source trim (slip)"; more.append(srcTag);
+        const inRow = el("div", "fields-row");
+        const iIn = el("input"); iIn.type = "number"; iIn.min = "0"; iIn.step = "0.05";
+        iIn.value = scene.source_in || 0;
+        iIn.oninput = () => S.setSourceTrim(scene.id, { source_in: parseFloat(iIn.value || 0) });
+        inRow.append(field("Source in (s)", iIn));
+        more.append(inRow);
+      }
+    });
+  }
+
   function renderScene(st, scene) {
     const root = S.genUnitRoot(S.genUnitId(scene)) || scene;
     const unitScenes = (st.project.scenes || []).filter((s) => S.genUnitId(s) === S.genUnitId(scene))
@@ -177,6 +243,7 @@
     body.append(field("Source", src));
     if ((root.source?.type) === "image" || (root.source?.type) === "mixed") renderImageSource(st, root, body);
     if ((root.source?.type) === "generated_frame") renderGeneratedFrameSource(st, root);
+    if ((root.source?.type) === "v2v") renderVideoSource(st, root, body, "V2V source video");
 
     const effFrames = effOf(scene, "frames"), effFps = effOf(scene, "fps") || 1;
     body.append(el("div", "insp-hint", `Duration ≈ ${(effFrames / effFps).toFixed(2)}s · trim on timeline · generation splits: global prompt or Outgoing seam below`));
@@ -187,6 +254,11 @@
     const genBtn = el("button", "btn primary", "Generate this scene");
     genBtn.onclick = () => S.generate(scene.id);
     actions.append(genBtn);
+    const toVid = el("button", "btn ghost", "Convert to video");
+    toVid.title = "Lock as a plain video clip — skipped by Generate; settings saved for Convert back to scene";
+    toVid.style.marginLeft = "8px";
+    toVid.onclick = () => S.convertToVideo(scene.id);
+    actions.append(toVid);
     const del = el("button", "btn danger", "Delete"); del.style.marginLeft = "8px"; del.onclick = () => S.removeScene(scene.id);
     actions.append(del);
     body.append(actions);
@@ -638,8 +710,10 @@
     renderGlobalPrompt(st);
     renderSwitch(st, scene);
     renderEngineStrip(st);
-    if (scene) renderScene(st, scene);
-    else renderProject(st);
+    if (scene) {
+      if (S.isVideoClip(scene)) renderVideoClip(st, scene);
+      else renderScene(st, scene);
+    } else renderProject(st);
     renderExposed(st);
     renderSplit(st);
     body.scrollTop = scrollTop;  // restore so editing doesn't jump to the top
diff --git a/movie_editor/frontend/mediabrowser.js b/movie_editor/frontend/mediabrowser.js
index a7d9bf94..ec511a4c 100644
--- a/movie_editor/frontend/mediabrowser.js
+++ b/movie_editor/frontend/mediabrowser.js
@@ -222,7 +222,9 @@
         ? `${m.name}\nClick to preview · drag onto a clip to set anchor`
         : m.kind === "audio"
           ? `${m.name}\nClick to preview · add via timeline + Add → Audio`
-          : `${m.name}\nDrag onto a clip to set anchor`;
+          : m.kind === "video"
+            ? `${m.name}\nDrag onto timeline · + Add → Video`
+            : `${m.name}\nDrag onto a clip to set anchor`;
       const thumb = el("div", "media-thumb");
       if (m.kind === "image") { const img = el("img"); img.src = API.mediaUrl(m.id); img.loading = "lazy"; thumb.append(img); }
       else if (m.kind === "audio") thumb.append(el("span", "media-icon", "♪"));
@@ -234,6 +236,8 @@
       card.append(del);
       if (m.kind === "image" || m.kind === "audio") {
         card.onclick = () => S.previewMedia(m.id);
+      } else if (m.kind === "video") {
+        card.ondblclick = () => S.addVideoClip(m.id);
       }
       grid.append(card);
     });
diff --git a/movie_editor/frontend/player.js b/movie_editor/frontend/player.js
index 2a2cb845..b58832d4 100644
--- a/movie_editor/frontend/player.js
+++ b/movie_editor/frontend/player.js
@@ -39,6 +39,7 @@
   let _currentClip = null;
 
   const _urlFor = (media) => API.resultUrl(S.get().project?.id, media);
+  const _clipUrl = (clip) => clip.binUrl || (clip.media ? _urlFor(clip.media) : "");
   const _clampT = (v, t) => Math.max(0, Math.min(t, (v.duration || (t + 1)) - 0.02));
 
   function _segmentSourceBadge(sceneId) {
@@ -77,13 +78,21 @@
       }
       const sc = seg.scene;
       const r = sr[sc.id];
+      const fx = sc.effects || {};
+      const vol = sc.audio_separated ? 0 : (sc.audio_volume != null ? sc.audio_volume : 1);
+      const inSec = (sc.source_in || 0) + (r?.inSec || 0);
       if (r && r.media) {
         out.push({
-          media: r.media, sceneId: sc.id, startSec, durationSec: dur, inSec: r.inSec || 0,
-          fx: sc.effects || {},
-          vol: sc.audio_separated ? 0 : (sc.audio_volume != null ? sc.audio_volume : 1),
+          media: r.media, sceneId: sc.id, startSec, durationSec: dur, inSec,
+          fx, vol,
         });
-      } else if (!sc.excluded) {
+      } else if (S.isVideoClip && S.isVideoClip(sc) && sc.source?.media_ref) {
+        out.push({
+          binUrl: API.mediaUrl(sc.source.media_ref),
+          sceneId: sc.id, startSec, durationSec: dur, inSec: sc.source_in || 0,
+          fx, vol,
+        });
+      } else if (!sc.excluded && S.isGenerativeScene && S.isGenerativeScene(sc)) {
         out.push({
           pending: true, sceneId: sc.id, startSec, durationSec: dur,
           slate: "Not generated yet",
@@ -204,7 +213,7 @@
   }
 
   function _syncPool() {
-    const urls = new Set(_clips.filter((c) => c.media).map((c) => _urlFor(c.media)));
+    const urls = new Set(_clips.filter((c) => c.media || c.binUrl).map((c) => _clipUrl(c)));
     urls.forEach((u) => _ensureVideo(u));
     for (const [u, v] of [...pool]) {
       if (urls.has(u)) continue;
@@ -275,8 +284,8 @@
       if (play) _startTick();
       return;
     }
-    if (!clip.media) return;
-    const v = _ensureVideo(_urlFor(clip.media));
+    if (!clip.media && !clip.binUrl) return;
+    const v = _ensureVideo(_clipUrl(clip));
     _currentClip = clip; _setActive(v);
     _showSlate("");
     const badge = clip.ghost ? (clip.slate || "Removed — preview only") : _segmentSourceBadge(clip.sceneId);
@@ -304,7 +313,8 @@
     const prev = _currentClip;
     const sameSource = next.media && prev.media
       && _urlFor(next.media) === _urlFor(prev.media);
-    const contiguous = sameSource
+    const sameBin = next.binUrl && prev.binUrl && next.binUrl === prev.binUrl;
+    const contiguous = (sameSource || sameBin)
       && Math.abs((next.inSec || 0) - ((prev.inSec || 0) + prev.durationSec)) < 0.05;
     _phSec = next.startSec;
     _currentClip = next;
@@ -548,7 +558,7 @@
     _syncPool();  // preload current clips, drop stale ones (timeline is dynamic)
 
     // Re-resolve the clip under the playhead each render (positions change with edits).
-    const hash = _clips.map((c) => (c.pending ? "pending:" + c.sceneId : c.ghost ? "ghost:" + c.ghostId : c.media?.filename) + "@" + c.startSec.toFixed(2) + "+" + (c.inSec || 0).toFixed(2)).join("|");
+    const hash = _clips.map((c) => (c.pending ? "pending:" + c.sceneId : c.ghost ? "ghost:" + c.ghostId : (_clipUrl(c) || "x")) + "@" + c.startSec.toFixed(2) + "+" + (c.inSec || 0).toFixed(2)).join("|");
     if (hash !== _lastSegHash) {
       _lastSegHash = hash;
       const clip = _clipAt(_phSec);
@@ -660,15 +670,19 @@
         } else {
           const sc = seg.scene;
           const stype = sc.source?.type || "empty";
-          const rendered = !!((st.sceneRenders || {})[sc.id] || {}).media;
-          const stale = rendered && S.renderIsStale?.(sc.id);
+          const isVideo = S.isVideoClip && S.isVideoClip(sc);
+          const rendered = isVideo
+            ? !!(sc.source?.media_ref || ((st.sceneRenders || {})[sc.id] || {}).media)
+            : !!((st.sceneRenders || {})[sc.id] || {}).media;
+          const stale = !isVideo && rendered && S.renderIsStale?.(sc.id);
           cls += rendered ? " rendered" : " pending";
-          if (stale) cls += " stale";
-          if (stype === "mixed") { cls += " mixed"; title = "Mixed · anchor + prior guides"; }
+          if (isVideo) { cls += " video"; title = "Video clip (locked)"; }
+          else if (stale) cls += " stale";
+          else if (stype === "mixed") { cls += " mixed"; title = "Mixed · anchor + prior guides"; }
           else if (stype === "carry") { cls += " carry"; title = "Carry · prior guides only"; }
           const idx = (p.scenes || []).indexOf(sc) + 1;
           if (!title) title = `Scene ${idx}${rendered ? (stale ? " (previous generation)" : " (rendered)") : " (not rendered)"}`;
-          else title = `Scene ${idx} · ${title}${rendered ? (stale ? " · previous generation" : "") : " · not rendered"}`;
+          else if (!isVideo) title = `Scene ${idx} · ${title}${rendered ? (stale ? " · previous generation" : "") : " · not rendered"}`;
         }
         const chip = el("div", cls);
         chip.style.width = (d / _totalSecCur * 100).toFixed(3) + "%";
diff --git a/movie_editor/frontend/store.js b/movie_editor/frontend/store.js
index 637192fe..e1040da3 100644
--- a/movie_editor/frontend/store.js
+++ b/movie_editor/frontend/store.js
@@ -146,7 +146,7 @@
     const parts = [];
     const anchor = (project.anchor || "").trim();
     if (anchor) parts.push(anchor);
-    const active = (project.scenes || []).filter((s) => !s.excluded);
+    const active = (project.scenes || []).filter((s) => !s.excluded && isGenerativeScene(s));
     const units = _groupGenerativeUnits(active);
     units.forEach(([, group], i) => {
       const root = _rootFromGroup(group);
@@ -299,10 +299,28 @@
     return out;
   }
 
+  function _mergeGenerativeWithVideoClips(oldScenes, generativeNext) {
+    const oldOrder = (oldScenes || []).filter((s) => !s.excluded);
+    const videoIds = new Set(oldOrder.filter(isVideoClip).map((s) => s.id));
+    if (!videoIds.size) return generativeNext;
+    const byId = new Map((oldScenes || []).map((s) => [s.id, s]));
+    const out = [];
+    let g = 0;
+    for (const sc of oldOrder) {
+      if (videoIds.has(sc.id)) {
+        out.push(JSON.parse(JSON.stringify(byId.get(sc.id) || sc)));
+      } else if (isGenerativeScene(sc)) {
+        if (g < generativeNext.length) out.push(generativeNext[g++]);
+      }
+    }
+    while (g < generativeNext.length) out.push(generativeNext[g++]);
+    return out;
+  }
+
   // Global prompt is authoritative: parsed scenes become the timeline; unmatched clips
   // are ghosted (if rendered) or dropped. Matching is by normalized text only.
   function _alignScenesFromParsed(oldScenes, parsedScenes, ghosts) {
-    const old = (oldScenes || []).filter((s) => !s.excluded);
+    const old = (oldScenes || []).filter((s) => !s.excluded && isGenerativeScene(s));
     const usedOld = new Set();
     const usedGhost = new Set();
     const next = [];
@@ -340,7 +358,8 @@
     }
     ghostsOut = ghostsOut.filter((g) => !usedGhost.has(g.id));
 
-    return { next, ghosts: ghostsOut, usedGhost };
+    const merged = _mergeGenerativeWithVideoClips(oldScenes, next);
+    return { next: merged, ghosts: ghostsOut, usedGhost };
   }
 
   function _normalizeGlobalPromptParse(trimmed, v) {
@@ -676,6 +695,8 @@
   function scene(id) { return state.project?.scenes.find((s) => s.id === id) || null; }
 
   function genUnitId(sc) { return (sc && sc.gen_unit_id) || (sc && sc.id) || ""; }
+  function isVideoClip(sc) { return (sc && sc.source && sc.source.type) === "video"; }
+  function isGenerativeScene(sc) { return !!sc && !isVideoClip(sc); }
   function isGenSubclip(sc) { return !!((sc && sc.cut_offset_frames) > 0); }
   function genUnitRoot(unitId) {
     const group = (state.project?.scenes || []).filter((s) => genUnitId(s) === unitId);
@@ -872,6 +893,127 @@
     notify(); scheduleSave(); // server assigns id; reselect after commit
   }
 
+  function _snapshotSceneForArchive(sc) {
+    const r = state.sceneRenders[sc.id];
+    return {
+      text: sc.text || "",
+      source: JSON.parse(JSON.stringify(sc.source || {})),
+      guides: JSON.parse(JSON.stringify(sc.guides || [])),
+      character_ids: [...(sc.character_ids || [])],
+      effects: JSON.parse(JSON.stringify(sc.effects || {})),
+      frames: sc.frames,
+      frames_mode: sc.frames_mode,
+      fps: sc.fps,
+      fps_mode: sc.fps_mode,
+      transition_to_next: sc.transition_to_next || "",
+      video_transition: sc.video_transition || "",
+      transition_frames: sc.transition_frames,
+      rating: sc.rating || "",
+      audio_volume: sc.audio_volume,
+      audio_separated: sc.audio_separated,
+      gen_unit_id: sc.gen_unit_id,
+      cut_offset_frames: sc.cut_offset_frames,
+      source_in: sc.source_in,
+      source_dur: sc.source_dur,
+      width: sc.width,
+      height: sc.height,
+      render: r ? JSON.parse(JSON.stringify(r)) : null,
+    };
+  }
+
+  function _applySceneArchive(sc, archive) {
+    if (!archive) return;
+    sc.text = archive.text || "";
+    sc.source = JSON.parse(JSON.stringify(archive.source || { type: "carry" }));
+    sc.guides = JSON.parse(JSON.stringify(archive.guides || []));
+    sc.character_ids = [...(archive.character_ids || [])];
+    sc.effects = JSON.parse(JSON.stringify(archive.effects || {}));
+    sc.frames = archive.frames;
+    sc.frames_mode = archive.frames_mode;
+    sc.fps = archive.fps;
+    sc.fps_mode = archive.fps_mode;
+    sc.transition_to_next = archive.transition_to_next || "";
+    sc.video_transition = archive.video_transition || "";
+    sc.transition_frames = archive.transition_frames;
+    sc.rating = archive.rating || "";
+    sc.audio_volume = archive.audio_volume;
+    sc.audio_separated = !!archive.audio_separated;
+    sc.gen_unit_id = archive.gen_unit_id;
+    sc.cut_offset_frames = archive.cut_offset_frames || 0;
+    sc.source_in = archive.source_in || 0;
+    sc.source_dur = archive.source_dur;
+    sc.width = archive.width;
+    sc.height = archive.height;
+    sc.scene_archive = null;
+    if (archive.render) state.sceneRenders[sc.id] = JSON.parse(JSON.stringify(archive.render));
+  }
+
+  function convertToVideo(sceneId) {
+    if (!state.project) return;
+    const sc = scene(sceneId);
+    if (!sc || isVideoClip(sc)) return;
+    _historyRecord();
+    const uid = genUnitId(sc);
+    const root = genUnitRoot(uid) || sc;
+    const archive = _snapshotSceneForArchive(root);
+    const mediaRef = (root.source && root.source.media_ref) || null;
+    state.project.scenes = state.project.scenes.filter(
+      (s) => genUnitId(s) !== uid || s.id === root.id
+    );
+    Object.assign(root, {
+      scene_archive: archive,
+      source: { type: "video", media_ref: mediaRef },
+      gen_unit_id: root.id,
+      cut_offset_frames: 0,
+    });
+    if (state.selectedSceneId && !scene(state.selectedSceneId)) {
+      state.selectedSceneId = root.id;
+      state.selectedSceneIds = [root.id];
+      _selectionAnchorId = root.id;
+    }
+    _afterTimelineStructureChange();
+    notify(); scheduleSave();
+  }
+
+  function convertToScene(sceneId) {
+    if (!state.project) return;
+    const sc = scene(sceneId);
+    if (!sc || !isVideoClip(sc)) return;
+    _historyRecord();
+    if (sc.scene_archive) {
+      _applySceneArchive(sc, sc.scene_archive);
+    } else {
+      const mediaRef = sc.source?.media_ref || null;
+      sc.scene_archive = null;
+      sc.source = { type: "v2v", media_ref: mediaRef };
+      if (!sc.text) sc.text = "";
+      sc.gen_unit_id = sc.gen_unit_id || sc.id;
+    }
+    _afterTimelineStructureChange();
+    notify(); scheduleSave();
+  }
+
+  function addVideoClip(mediaId) {
+    if (!state.project || !mediaId) return;
+    const asset = (state.mediaBin || []).find((m) => m.id === mediaId);
+    if (!asset || asset.kind !== "video") return;
+    _historyRecord();
+    const id = _newSceneId();
+    state.project.scenes.push({
+      id,
+      gen_unit_id: id,
+      text: "",
+      transition_to_next: "",
+      source: { type: "video", media_ref: mediaId },
+      excluded: false,
+    });
+    state.selectedSceneId = id;
+    state.selectedSceneIds = [id];
+    _selectionAnchorId = id;
+    window.Timeline?.requestAutoFit?.();
+    notify(); scheduleSave();
+  }
+
   function _removeSceneNoHistory(id) {
     if (!state.project) return;
     const arr = state.project.scenes;
@@ -1362,19 +1504,19 @@
   // back to carry / i2v guides, not a broken anchor).
   function _anchorAvailable(s) {
     const t = s.source && s.source.type;
-    if (t !== "image" && t !== "generated_frame" && t !== "mixed") return false;
+    if (t !== "image" && t !== "generated_frame" && t !== "mixed" && t !== "v2v") return false;
     const ref = s.source.media_ref;
     return !!(ref && (state.mediaBin || []).some((m) => m.id === ref));
   }
   function _runs() {
-    const active = state.project.scenes.filter((s) => !s.excluded);
+    const active = state.project.scenes.filter((s) => !s.excluded && isGenerativeScene(s));
     const runs = [];
     for (const s of active) {
       const t = (s.source && s.source.type) || "empty";
       // Editorial subclips always continue the current run; carry / broken anchors too.
       const isCarry = isGenSubclip(s)
         || t === "carry"
-        || ((t === "image" || t === "generated_frame" || t === "mixed") && !_anchorAvailable(s));
+        || ((t === "image" || t === "generated_frame" || t === "mixed" || t === "v2v") && !_anchorAvailable(s));
       if (isCarry && runs.length) runs[runs.length - 1].push(s.id);
       else runs.push([s.id]);
     }
@@ -1716,7 +1858,7 @@
     if (!state.project) return;
     const ids = (state.selectedSceneIds || []).filter((sid) => {
       const s = scene(sid);
-      return s && !s.excluded;
+      return s && !s.excluded && isGenerativeScene(s);
     });
     if (!ids.length) {
       set({ gen: { state: "error", promptId: null, media: [], msg: "No scenes selected." } });
@@ -1782,12 +1924,10 @@
     for (const sc of (p.scenes || [])) {
       if (sc.excluded) continue;
       const r = state.sceneRenders[sc.id];
-      if (!(r && r.media)) continue;
       const fps = (sc.fps_mode !== "project" && sc.fps != null ? sc.fps : p.frame_rate) || 25;
       const tFrames = sc.transition_frames || 0;
-      out.push({
-        media: r.media,
-        inSec: (r.inSec || 0) + (sc.source_in || 0),
+      const base = {
+        inSec: (r?.inSec || 0) + (sc.source_in || 0),
         durationSec: sc.source_dur != null ? sc.source_dur : sceneDurationSec(sc),
         fx: sc.effects || {}, fps,
         w: (sc.width != null ? sc.width : p.width) || null,
@@ -1796,7 +1936,12 @@
         tdur: tFrames > 0 ? tFrames / fps : 0,
         sceneId: sc.id,
         volume: sc.audio_separated ? 0 : (sc.audio_volume != null ? sc.audio_volume : 1),
-      });
+      };
+      if (r && r.media) {
+        out.push({ media: r.media, ...base });
+      } else if (isVideoClip(sc) && sc.source?.media_ref) {
+        out.push({ bin_media_ref: sc.source.media_ref, ...base });
+      }
     }
     return out;
   }
@@ -1808,9 +1953,15 @@
     if (!id) { alert("Select a clip to export."); return; }
     const sc = scene(id);
     const r = state.sceneRenders[id];
-    if (!r || !r.media) { alert("That clip hasn't been generated yet — generate it first."); return; }
     const idx = state.project.scenes.findIndex((s) => s.id === id);
     const proj = (state.project.name || "montage").replace(/[^\w.-]+/g, "_");
+    if (isVideoClip(sc) && sc.source?.media_ref && !r?.media) {
+      const asset = (state.mediaBin || []).find((m) => m.id === sc.source.media_ref);
+      const name = asset?.name || `${proj}_video_${_stamp()}.mp4`;
+      await _saveBlobAs(API.mediaUrl(sc.source.media_ref), name.replace(/[^\w.-]+/g, "_"));
+      return;
+    }
+    if (!r || !r.media) { alert("That clip hasn't been generated yet — generate it first."); return; }
     const inSec = (r.inSec || 0) + (sc?.source_in || 0);
     const dur = sc?.source_dur != null ? sc.source_dur : sceneDurationSec(sc);
     try {
@@ -1836,7 +1987,8 @@
       `Stitch ${rc.length} clip(s) into one final video?\n\nThis replaces per-scene renders with a single stitched file on the timeline.`
     )) return;
     const clips = rc.map((c) => ({
-      filename: c.media.filename, subfolder: c.media.subfolder || "", type: c.media.type || "output",
+      filename: c.media?.filename, subfolder: c.media?.subfolder || "", type: c.media?.type || "output",
+      bin_media_ref: c.bin_media_ref || "",
       in: +c.inSec.toFixed(3), dur: +c.durationSec.toFixed(3),
       fx: c.fx, fps: c.fps, w: c.w, h: c.h, transition: c.transition, tdur: +(c.tdur || 0).toFixed(3),
       volume: c.volume, scene_id: c.sceneId,
@@ -2052,6 +2204,15 @@
 
   function assignMediaToScene(sceneId, mediaId) {
     const s = scene(sceneId); if (!s) return;
+    const asset = (state.mediaBin || []).find((m) => m.id === mediaId);
+    if (asset?.kind === "video") {
+      if (isVideoClip(s)) {
+        patchScene(sceneId, { source: { ...(s.source || {}), type: "video", media_ref: mediaId } });
+      } else if (confirm("Replace this scene with the imported video clip? Use Convert to video in the inspector to keep a backup of scene settings.")) {
+        patchScene(sceneId, { source: { type: "video", media_ref: mediaId } });
+      }
+      return;
+    }
     // Drag-drop is an explicit new anchor — use image i2v (not mixed/carry guides).
     const patch = { source: { ...(s.source || {}), type: "image", media_ref: mediaId } };
     if ((s.source?.media_ref || null) !== mediaId) patch.guides = [];
@@ -2083,6 +2244,7 @@
     scheduleSaveFromHistory, notifyHistoryState,
     refreshProjectList, loadProject, newProject, deleteProject, downloadProject, importProject,
     patchProject, patchProjectQuiet, patchScene, patchSceneQuiet, flushSave, selectScene, addScene, removeScene, removeSelectedScenes, dismissGhost, moveScene, moveSceneTo, scene,
+    addVideoClip, convertToVideo, convertToScene, isVideoClip, isGenerativeScene,
     sceneCharacterIds, toggleSceneCharacter,
     genUnitId, isGenSubclip, genUnitRoot, genUnitSceneIds,
     renderPromptForScene, renderPromptMismatch, renderAnchorMismatch, renderMediaLabel, renderIsStale,
diff --git a/movie_editor/frontend/styles.css b/movie_editor/frontend/styles.css
index 7f68c010..5bc555a2 100644
--- a/movie_editor/frontend/styles.css
+++ b/movie_editor/frontend/styles.css
@@ -377,6 +377,7 @@ input[type=number] { font-family: var(--font-mono); }
 .pm-chip.ghost { background: repeating-linear-gradient(-45deg, #2a2418, #2a2418 4px, #3d3420 4px, #3d3420 8px); opacity: .65; }
 .pm-chip.mixed { background: linear-gradient(180deg, color-mix(in srgb, var(--amber) 55%, #3ab3a8), #2a6a62); opacity: .72; }
 .pm-chip.carry { background: repeating-linear-gradient(-45deg, #1e2a28, #1e2a28 3px, #243532 3px, #243532 6px); opacity: .5; }
+.pm-chip.video { background: linear-gradient(180deg, #3a5a8a, #2a4468); opacity: .7; }
 .pm-chip:hover { opacity: 1 !important; }
 .pm-slate { position: absolute; inset: 0; z-index: 2; display: flex; flex-direction: column; align-items: center; justify-content: center; gap: 6px; background: #080808ee; color: var(--muted); pointer-events: none; text-align: center; padding: 16px; }
 .pm-slate-title { font-size: 14px; color: var(--amber-2); font-family: var(--font-mono); }
@@ -639,6 +640,11 @@ body.tl-reordering { cursor: grabbing; user-select: none; }
 .tl-aud-clip.scene-aud.selected:not(.focus) { border-color: var(--teal-soft); box-shadow: inset 0 0 0 1px rgba(45, 212, 191, 0.35); }
 .clip.excluded { opacity: .42; }
 .clip.excluded::before { background: var(--faint); }
+.clip.clip-video { background: linear-gradient(#141c28, #0f141c); border-color: #2a4468; }
+.clip.clip-video::before { background: #5b8def; opacity: .75; }
+.clip.clip-video .clip-no { color: #7eb0ff; font-weight: 600; }
+.clip.src-v2v::before { background: #7c6cf0; }
+.tl-track2.drop-target-track { outline: 1px dashed var(--teal); outline-offset: -2px; background: rgba(45, 212, 191, 0.04); }
 .clip.pending:not(.rendered) { border-style: dashed; border-color: color-mix(in srgb, var(--muted) 55%, var(--line)); }
 .clip.ghost { border-style: dashed; border-color: color-mix(in srgb, var(--amber) 45%, var(--line)); opacity: .72; cursor: default; }
 .clip.ghost::before { background: var(--amber); opacity: .35; }
diff --git a/movie_editor/frontend/timeline.js b/movie_editor/frontend/timeline.js
index 2837a512..77c5a49b 100644
--- a/movie_editor/frontend/timeline.js
+++ b/movie_editor/frontend/timeline.js
@@ -7,7 +7,7 @@
   const body = document.getElementById("timeline-body");
   const meta = document.getElementById("timeline-meta");
 
-  const SRC_ICON = { empty: "▦", image: "◐", generated_frame: "⛶", carry: "⇥", mixed: "◑" };
+  const SRC_ICON = { empty: "▦", image: "◐", generated_frame: "⛶", carry: "⇥", mixed: "◑", video: "▶", v2v: "⟳" };
 
   function genUnitRootScene(scene, p) {
     const uid = scene.gen_unit_id || scene.id;
@@ -23,7 +23,7 @@
   function anchorMediaRef(scene, p) {
     const src = sceneSourceForClip(scene, p);
     const t = src.type;
-    if (t === "image" || t === "mixed" || t === "generated_frame") return src.media_ref || null;
+    if (t === "image" || t === "mixed" || t === "generated_frame" || t === "video" || t === "v2v") return src.media_ref || null;
     return null;
   }
 
@@ -234,7 +234,7 @@
     const sep = bar.querySelector("[data-separate-audio]");
     if (sep) {
       const sc = hasSel ? S.scene(st.selectedSceneId) : null;
-      sep.disabled = !(sc && hasRender(st, st.selectedSceneId) && !sc.audio_separated);
+      sep.disabled = !(sc && S.isGenerativeScene(sc) && hasRender(st, st.selectedSceneId) && !sc.audio_separated);
     }
     const oldRating = bar.querySelector(".tl-rating-block");
     const freshRating = toolbarRatingBlock(st, st.project);
@@ -379,11 +379,12 @@
     const src = sceneSourceForClip(scene, p);
     const srcType = src.type || "empty";
     const clip = el("div", "clip" + clipSelClass(st, scene.id)
+      + (S.isVideoClip(scene) ? " clip-video" : "")
       + (scene.excluded ? " excluded" : "")
-      + (hasRender(st, scene.id) ? " rendered" : (!scene.excluded ? " pending" : ""))
+      + (hasRender(st, scene.id) ? " rendered" : (!scene.excluded && S.isGenerativeScene(scene) ? " pending" : ""))
       + (hasRender(st, scene.id) && S.renderIsStale?.(scene.id) ? " stale-render" : "")
       + (unitCuts > 1 ? " gen-cut" : "") + (subclip ? " subclip" : "")
-      + (srcType === "mixed" ? " src-mixed" : srcType === "carry" ? " src-carry" : srcType === "image" ? " src-image" : ""));
+      + (srcType === "mixed" ? " src-mixed" : srcType === "carry" ? " src-carry" : srcType === "image" ? " src-image" : srcType === "video" ? " src-video" : srcType === "v2v" ? " src-v2v" : ""));
     clip.dataset.sceneId = scene.id;
     clip.style.left = leftPx + "px";
     clip.style.width = Math.max(widthPx, 8) + "px";
@@ -464,7 +465,7 @@
     }
 
     const head = el("div", "clip-head");
-    head.append(el("span", "clip-no", p2(index + 1)));
+    head.append(el("span", "clip-no", S.isVideoClip(scene) ? "V" : p2(index + 1)));
     appendSrcBadge(head, srcType);
     head.append(el("span", "clip-dur", timecode(sDur(scene, p), sFps(scene, p))));
     clip.append(head);
@@ -472,8 +473,11 @@
     const root = unitCuts > 1
       ? (p.scenes || []).find((s) => (s.gen_unit_id || s.id) === (scene.gen_unit_id || scene.id) && !(s.cut_offset_frames > 0))
       : null;
-    const label = scene.text || (root && root.text) || (subclip ? "cut" : "empty scene");
-    clip.append(el("div", "clip-text" + (label && label !== "empty scene" && label !== "cut" ? "" : " empty"), label));
+    const mref = anchorMediaRef(scene, p);
+    const label = S.isVideoClip(scene)
+      ? ((mref && (st.mediaBin || []).find((m) => m.id === mref)?.name) || "Video clip")
+      : (scene.text || (root && root.text) || (subclip ? "cut" : "empty scene"));
+    clip.append(el("div", "clip-text" + (label && label !== "empty scene" && label !== "cut" && label !== "Video clip" ? "" : " empty"), label));
 
     const charIds = S.sceneCharacterIds(scene.id);
     if (charIds.length) {
@@ -690,14 +694,15 @@
       panel.append(row);
     };
 
-    addRow("Clip", "Append a new scene clip to the timeline", () => S.addScene());
+    addRow("Clip", "Append a new generative scene clip to the timeline", () => S.addScene());
+    addRow("Video", "Add an imported video clip from the Media Browser", () => openVideoClipModal(st));
     addRow("Effects", "Post-render clip effect (zoom, blur, fade…)", () => openNleSettingsModal("effect", st));
     addRow("Transitions", "Video blend on the outgoing edge of the clip", () => openNleSettingsModal("transition", st));
     addRow("Text", "Coming soon", null, true);
     addRow("Image", "Coming soon", null, true);
     addRow("Audio", "Add an audio track from the Media Browser", () => openAudioTrackModal(st));
     const selSc = st.selectedSceneId ? S.scene(st.selectedSceneId) : null;
-    const canSepAud = !!(selSc && hasRender(st, st.selectedSceneId) && !selSc.audio_separated);
+    const canSepAud = !!(selSc && hasRender(st, st.selectedSceneId) && !selSc.audio_separated && S.isGenerativeScene(selSc));
     addRow("Separate audio", "Move selected clip's audio to its own track", () => S.separateSceneAudio(st.selectedSceneId), !canSepAud);
 
     btn.onclick = (e) => {
@@ -713,6 +718,66 @@
     return wrap;
   }
 
+  function openVideoClipModal(st) {
+    closeAddModal();
+    const videoAssets = (st.mediaBin || []).filter((m) => m.kind === "video");
+
+    _addModal = el("div", "modal-overlay");
+    const box = el("div", "modal");
+    const head = el("div", "modal-head");
+    head.append(el("div", "modal-title", "Add video clip"));
+    const headRight = el("div", "modal-head-right");
+    const closeBtn = el("button", "btn ghost tiny", "✕");
+    closeBtn.onclick = closeAddModal;
+    headRight.append(closeBtn);
+    head.append(headRight);
+    box.append(head);
+
+    const content = el("div", "modal-content");
+    const pick = (mediaId) => {
+      S.addVideoClip(mediaId);
+      closeAddModal();
+    };
+
+    if (!videoAssets.length) {
+      content.append(el("div", "pj-meta", "Upload video in the Media Browser (mp4, mov, webm…), then pick it here."));
+      const uploadBtn = el("button", "btn primary tiny", "Upload video file");
+      const fileIn = el("input");
+      fileIn.type = "file";
+      fileIn.accept = "video/*,.mp4,.mov,.webm,.mkv";
+      fileIn.style.display = "none";
+      uploadBtn.onclick = () => fileIn.click();
+      fileIn.onchange = async () => {
+        const files = [...(fileIn.files || [])];
+        fileIn.value = "";
+        if (!files.length) return;
+        await S.uploadMedia(files);
+        const fresh = S.get();
+        const added = (fresh.mediaBin || []).filter((m) => m.kind === "video");
+        if (added[0]) pick(added[0].id);
+        else closeAddModal();
+      };
+      content.append(uploadBtn);
+      content.append(fileIn);
+    } else {
+      content.append(el("div", "pj-meta", "Video clips play as-is and are skipped by Generate."));
+      videoAssets.forEach((m) => {
+        const row = el("button", "tl-add-row aud-pick-row", m.name || m.id);
+        row.type = "button";
+        row.onclick = () => pick(m.id);
+        content.append(row);
+      });
+    }
+
+    const cancel = el("button", "btn ghost tiny", "Cancel");
+    cancel.onclick = closeAddModal;
+    content.append(cancel);
+    box.append(content);
+    _addModal.append(box);
+    _addModal.addEventListener("click", (e) => { if (e.target === _addModal) closeAddModal(); });
+    document.body.append(_addModal);
+  }
+
   // ── audio lanes (NLE-style, below video) ───────────────────────────────────────
   function openAudioTrackModal(st) {
     closeAddModal();
@@ -953,7 +1018,7 @@
     const Picker = window.MovieRatingPicker;
     if (!studioCond || !hasSel || !hasRender(st, st.selectedSceneId) || !Picker) return wrap;
     const sc = S.scene(st.selectedSceneId);
-    if (!sc) return wrap;
+    if (!sc || !S.isGenerativeScene(sc)) return wrap;
     const sceneNo = p.scenes.indexOf(sc) + 1;
     const raw = sceneRatingRaw(st, sc);
     const rlabel = el("span", "tl-keys", `Scene ${sceneNo}`);
@@ -1000,7 +1065,7 @@
     sepAud.dataset.separateAudio = "1";
     sepAud.title = "Detach this clip's audio onto its own track (video keeps picture only)";
     const selSc = hasSel ? S.scene(st.selectedSceneId) : null;
-    sepAud.disabled = !(selSc && hasRender(st, st.selectedSceneId) && !selSc?.audio_separated);
+    sepAud.disabled = !(selSc && hasRender(st, st.selectedSceneId) && !selSc?.audio_separated && S.isGenerativeScene(selSc));
     sepAud.onclick = () => S.separateSceneAudio(st.selectedSceneId);
     bar.append(split); bar.append(del); bar.append(exp); bar.append(sepAud);
     bar.append(toolbarRatingBlock(st, p));
@@ -1101,6 +1166,23 @@
 
     const tracks = el("div", "tl-tracks");
     const track = el("div", "tl-track2");
+    track.addEventListener("dragover", (e) => {
+      if (e.dataTransfer?.types?.includes("application/funpack-media")) {
+        e.preventDefault();
+        track.classList.add("drop-target-track");
+      }
+    });
+    track.addEventListener("dragleave", () => track.classList.remove("drop-target-track"));
+    track.addEventListener("drop", (e) => {
+      track.classList.remove("drop-target-track");
+      const id = e.dataTransfer?.getData("application/funpack-media");
+      if (!id) return;
+      const asset = (st.mediaBin || []).find((m) => m.id === id);
+      if (asset?.kind === "video") {
+        e.preventDefault();
+        S.addVideoClip(id);
+      }
+    });
     track.addEventListener("click", (e) => {
       const clip = e.target.closest(".clip:not(.ghost)[data-scene-id]");
       if (!clip) return;
diff --git a/movie_editor/frontend/view_bus.js b/movie_editor/frontend/view_bus.js
index c9c74ae5..d4778436 100644
--- a/movie_editor/frontend/view_bus.js
+++ b/movie_editor/frontend/view_bus.js
@@ -13,7 +13,7 @@
         s.id, s.text, s.excluded, s.frames, s.fps, s.frames_mode, s.fps_mode,
         s.video_transition, s.transition_frames, s.transition_to_next, s.effects,
         s.gen_unit_id, s.cut_offset_frames, s.source?.type, s.source?.media_ref,
-        s.source_in, s.source_dur, s.rating, s.audio_volume, s.audio_separated,
+        s.source_in, s.source_dur, s.rating, s.audio_volume, s.audio_separated, s.scene_archive,
       ]),
       renders: st.sceneRenders,
       ghosts: st.sceneGhosts,
diff --git a/movie_editor/server.py b/movie_editor/server.py
index 3fa8df96..abf4dc92 100644
--- a/movie_editor/server.py
+++ b/movie_editor/server.py
@@ -840,7 +840,7 @@ if web is not None and PromptServer is not None:
             return web.json_response({"detail": f"Node registry unavailable: {e}"}, status=502)
         bridge.reset_progress()
         bridge.current_progress()  # ensure the sampler step-progress hook is installed
-        active_scenes = [s for s in target.scenes if not s.excluded]
+        active_scenes = [s for s in target.scenes if not s.excluded and not is_video_clip(s)]
         active_scene_count = len(active_scenes)
         # V1 uniform chain: if trimmed scenes all agree on a frame count, use it.
         # This makes the timeline trim handle actually affect generation length,
@@ -856,7 +856,7 @@ if web is not None and PromptServer is not None:
             if trimmed and all(f == trimmed[0] for f in trimmed)
             else target.num_frames_per_scene
         )
-        from .backend.timeline import continuity_media_refs
+        from .backend.timeline import continuity_media_refs, is_video_clip
         media_pack = _prepare_media(target, continuity_media_refs(p, target))
         sampler_inputs = _run_sampler_inputs(target, active_scene_count, full=p)
         _attach_scene_anchors(sampler_inputs, media_pack, target)
@@ -1003,6 +1003,11 @@ if web is not None and PromptServer is not None:
             return web.json_response({"detail": f"Output directory unavailable: {e}"}, status=500)
 
         def _resolve(c):
+            if c.get("bin_media_ref"):
+                mp = media.path_for(c.get("bin_media_ref") or "")
+                if mp is None:
+                    return ""
+                return str(mp)
             base = outdir if c.get("type", "output") == "output" else tempdir
             return os.path.join(base, c.get("subfolder", ""), c.get("filename", ""))
 
diff --git a/movie_editor/tests/test_timeline.py b/movie_editor/tests/test_timeline.py
index 03a80d75..dc9747cb 100644
--- a/movie_editor/tests/test_timeline.py
+++ b/movie_editor/tests/test_timeline.py
@@ -103,6 +103,24 @@ def test_missing_markers_fall_back_to_scene_labels():
     assert "scene 2" in prompt
 
 
+def test_video_clips_excluded_from_combined_prompt():
+    p = _project(scenes=[
+        {"text": "gen scene", "source": {"type": "carry"}},
+        {"text": "", "source": {"type": "video", "media_ref": "vid1"}},
+    ])
+    prompt = build_combined_prompt(p)
+    assert "gen scene" in prompt
+    assert build_combined_prompt(p, include_excluded=True).count("gen scene") >= 1
+
+
+def test_is_video_clip():
+    from movie_editor.backend.timeline import Scene, SceneSource, is_generative_scene, is_video_clip
+
+    sc = Scene(source=SceneSource(type="video", media_ref="v1"))
+    assert is_video_clip(sc)
+    assert not is_generative_scene(sc)
+
+
 def test_excluded_scenes_dropped_unless_requested():
     p = _project(
         anchor="a cat",