diff --git a/src/livekit/MatrixAudioRenderer.test.tsx b/src/livekit/MatrixAudioRenderer.test.tsx
index bb79a64d..eb6c306f 100644
--- a/src/livekit/MatrixAudioRenderer.test.tsx
+++ b/src/livekit/MatrixAudioRenderer.test.tsx
@@ -52,9 +52,16 @@ afterEach(() => {
 vi.mock("@livekit/components-react", async (importOriginal) => {
   return {
     ...(await importOriginal()),
-    AudioTrack: (props: { trackRef: TrackReference }): ReactNode => {
+    AudioTrack: (props: {
+      trackRef: TrackReference;
+      muted?: boolean;
+    }): ReactNode => {
       return (
-        <audio data-testid={"audio"}>
+        <audio
+          data-testid={"audio"}
+          data-source={props.trackRef.publication.source}
+          data-muted={String(!!props.muted)}
+        >
           {getTrackReferenceId(props.trackRef)}
         </audio>
       );
@@ -83,6 +90,7 @@ function renderTestComponent(
     kind: Track.Kind;
     source: Track.Source;
   }[],
+  props: { muted?: boolean; screenshareAudioMuted?: boolean } = {},
 ): RenderResult {
   const liveKitParticipants = livekitParticipantIdentities.map((identity) =>
     mockRemoteParticipant({ identity }),
@@ -117,6 +125,7 @@ function renderTestComponent(
         validIdentities={participants.map((p) => p.identity)}
         livekitRoom={livekitRoom}
         url={""}
+        {...props}
       />
     </MediaDevicesProvider>,
   );
@@ -286,3 +295,55 @@ it("should setup audioContext gain and pan", () => {
   expect(testAudioContext.gain.gain.value).toEqual(0.1);
   expect(testAudioContext.pan.pan.value).toEqual(1);
 });
+
+// [lotus] The host's "Mute Screenshare Audio" mutes ONLY screenshare-audio
+// elements, via the same `muted` prop path as deafen (so it survives re-renders
+// and later-published shares); deafen still mutes everything.
+it("screenshareAudioMuted mutes only ScreenShareAudio tracks", () => {
+  const explicitTracks = [
+    {
+      participantId: "@alice:DEV0",
+      kind: Track.Kind.Audio,
+      source: Track.Source.Microphone,
+    },
+    {
+      participantId: "@alice:DEV0",
+      kind: Track.Kind.Audio,
+      source: Track.Source.ScreenShareAudio,
+    },
+  ];
+  const mutedBySource = (r: RenderResult): Record<string, string> =>
+    Object.fromEntries(
+      r
+        .queryAllByTestId("audio")
+        .map((el) => [el.dataset.source, el.dataset.muted]),
+    );
+
+  expect(
+    mutedBySource(
+      renderTestComponent(
+        [{ userId: "@alice", deviceId: "DEV0" }],
+        ["@alice:DEV0"],
+        explicitTracks,
+        { screenshareAudioMuted: true },
+      ),
+    ),
+  ).toEqual({
+    [Track.Source.Microphone]: "false",
+    [Track.Source.ScreenShareAudio]: "true",
+  });
+
+  expect(
+    mutedBySource(
+      renderTestComponent(
+        [{ userId: "@alice", deviceId: "DEV0" }],
+        ["@alice:DEV0"],
+        explicitTracks,
+        { muted: true },
+      ),
+    ),
+  ).toEqual({
+    [Track.Source.Microphone]: "true",
+    [Track.Source.ScreenShareAudio]: "true",
+  });
+});
diff --git a/src/livekit/MatrixAudioRenderer.tsx b/src/livekit/MatrixAudioRenderer.tsx
index e3970e9f..cf92f361 100644
--- a/src/livekit/MatrixAudioRenderer.tsx
+++ b/src/livekit/MatrixAudioRenderer.tsx
@@ -38,6 +38,12 @@ export interface MatrixAudioRendererProps {
    * If set to `true`, the server will stop sending audio track data to the client.
    */
   muted?: boolean;
+  /**
+   * [lotus] If set to `true`, mutes only the `Track.Source.ScreenShareAudio`
+   * tracks (the host's "Mute Screenshare Audio" control) — same mechanism as
+   * `muted`, so it holds across re-renders and later-published shares.
+   */
+  screenshareAudioMuted?: boolean;
 }
 
 /**
@@ -58,6 +64,7 @@ export function LivekitRoomAudioRenderer({
   livekitRoom,
   validIdentities,
   muted,
+  screenshareAudioMuted,
 }: MatrixAudioRendererProps): ReactNode {
   const logger = rootLogger.getChild("[MatrixAudioRenderer]");
   const tracks = useTracks(
@@ -143,7 +150,11 @@ export function LivekitRoomAudioRenderer({
         <AudioTrackWithAudioNodes
           key={getTrackReferenceId(trackRef)}
           trackRef={trackRef}
-          muted={muted}
+          muted={
+            muted ||
+            (screenshareAudioMuted &&
+              trackRef.publication.source === Track.Source.ScreenShareAudio)
+          }
           audioContext={shouldUseAudioContext ? audioContext : undefined}
           audioNodes={audioNodes}
         />
diff --git a/src/lotus/lotusDeafen.test.ts b/src/lotus/lotusDeafen.test.ts
index b9a11ee6..2d4fb7ea 100644
--- a/src/lotus/lotusDeafen.test.ts
+++ b/src/lotus/lotusDeafen.test.ts
@@ -7,13 +7,11 @@ Please see LICENSE in the repository root for full details.
 
 import { EventEmitter } from "events";
 import { afterEach, beforeEach, expect, test, vi } from "vitest";
-import { of } from "rxjs";
-import { Track } from "livekit-client";
 
-import { type CallViewModel } from "../state/CallViewModel/CallViewModel";
 import { setAudioEnabled$ } from "../controls";
 import { startLotusDeafen } from "./lotusDeafen";
 import { LotusWidgetActions } from "./lotusActions";
+import { setScreenshareAudioMuted$ } from "./lotusScreenshareAudio";
 
 const lazyActions = new EventEmitter();
 
@@ -27,18 +25,6 @@ vi.mock("../widget", () => ({
   },
 }));
 
-/** Minimal CallViewModel stub: no connections, so no livekit rooms. */
-function mockVm(participants: unknown[] = []): CallViewModel {
-  const livekitRoom = {
-    remoteParticipants: new Map(participants.map((p, i) => [String(i), p])),
-    on: vi.fn(),
-    off: vi.fn(),
-  };
-  return {
-    allConnections$: of({ getConnections: () => [{ livekitRoom }] }),
-  } as unknown as CallViewModel;
-}
-
 function send(data: unknown): void {
   lazyActions.emit(LotusWidgetActions.SetDeafen, {
     detail: { data },
@@ -60,7 +46,7 @@ afterEach(() => {
 });
 
 test("deafen mutes, and undeafen unmutes, EC's global audio output", () => {
-  const stop = startLotusDeafen(mockVm());
+  const stop = startLotusDeafen();
 
   send({ deafened: true, screenshareAudioMuted: false });
   expect(emissions).toEqual([false]);
@@ -73,7 +59,7 @@ test("deafen mutes, and undeafen unmutes, EC's global audio output", () => {
 });
 
 test("re-sending the same state is idempotent (host resendForkState)", () => {
-  const stop = startLotusDeafen(mockVm());
+  const stop = startLotusDeafen();
 
   send({ deafened: true, screenshareAudioMuted: false });
   send({ deafened: true, screenshareAudioMuted: false });
@@ -86,7 +72,7 @@ test("re-sending the same state is idempotent (host resendForkState)", () => {
 });
 
 test("undeafen does not re-enable audio the user had muted themselves", () => {
-  const stop = startLotusDeafen(mockVm());
+  const stop = startLotusDeafen();
 
   // The user mutes all audio through EC's own control first.
   setAudioEnabled$.next(false);
@@ -101,48 +87,44 @@ test("undeafen does not re-enable audio the user had muted themselves", () => {
   expect(emissions).toEqual([false]);
 });
 
-test("deafen never touches per-participant volume", () => {
-  const participant = { setVolume: vi.fn() };
-  const stop = startLotusDeafen(mockVm([participant]));
+test("deafen never touches the screenshare-audio flag", () => {
+  const stop = startLotusDeafen();
 
   send({ deafened: true, screenshareAudioMuted: false });
   send({ deafened: false, screenshareAudioMuted: false });
-  expect(participant.setVolume).not.toHaveBeenCalled();
+  expect(setScreenshareAudioMuted$.value).toBe(false);
 
   stop();
 });
 
-test("screenshare-audio mute is applied per source and only undone for participants we muted", () => {
-  const participant = { setVolume: vi.fn() };
-  const stop = startLotusDeafen(mockVm([participant]));
+test("screenshare-audio mute is a global flag the renderer reads, reset on teardown", () => {
+  const seen: boolean[] = [];
+  const flagSub = setScreenshareAudioMuted$.subscribe((v) => seen.push(v));
+  const stop = startLotusDeafen();
 
-  // Not muted yet: no volume writes at all.
+  // Not muted yet: nothing pushed beyond the BehaviorSubject's initial value.
   send({ deafened: false, screenshareAudioMuted: false });
-  expect(participant.setVolume).not.toHaveBeenCalled();
+  expect(seen).toEqual([false]);
 
   send({ deafened: false, screenshareAudioMuted: true });
-  expect(participant.setVolume).toHaveBeenCalledWith(
-    0,
-    Track.Source.ScreenShareAudio,
-  );
+  expect(setScreenshareAudioMuted$.value).toBe(true);
+  // Re-applying the same state (host resend after reconnect) is a no-op.
+  send({ deafened: false, screenshareAudioMuted: true });
+  expect(seen).toEqual([false, true]);
 
-  participant.setVolume.mockClear();
-  send({ deafened: false, screenshareAudioMuted: false });
-  expect(participant.setVolume).toHaveBeenCalledWith(
-    1,
-    Track.Source.ScreenShareAudio,
-  );
-
-  // Releasing again writes nothing: we no longer own that participant.
-  participant.setVolume.mockClear();
-  send({ deafened: false, screenshareAudioMuted: false });
-  expect(participant.setVolume).not.toHaveBeenCalled();
+  // Deafen + undeafen while screenshare audio is muted leaves it muted.
+  send({ deafened: true, screenshareAudioMuted: true });
+  send({ deafened: false, screenshareAudioMuted: true });
+  expect(setScreenshareAudioMuted$.value).toBe(true);
 
+  // Leaving the call clears it so the next call starts clean.
   stop();
+  expect(setScreenshareAudioMuted$.value).toBe(false);
+  flagSub.unsubscribe();
 });
 
 test("a partial payload only moves the flag it names", () => {
-  const stop = startLotusDeafen(mockVm());
+  const stop = startLotusDeafen();
 
   send({ deafened: true });
   expect(emissions).toEqual([false]);
diff --git a/src/lotus/lotusDeafen.ts b/src/lotus/lotusDeafen.ts
index 9f2e998b..4d821ebc 100644
--- a/src/lotus/lotusDeafen.ts
+++ b/src/lotus/lotusDeafen.ts
@@ -5,19 +5,13 @@ SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
 Please see LICENSE in the repository root for full details.
 */
 
-import {
-  type RemoteParticipant,
-  type Room as LivekitRoom,
-  RoomEvent,
-  Track,
-} from "livekit-client";
 import { logger } from "matrix-js-sdk/lib/logger";
 import { type IWidgetApiRequest } from "matrix-widget-api";
 
-import { type CallViewModel } from "../state/CallViewModel/CallViewModel";
 import { setAudioEnabled$ } from "../controls";
 import { widget } from "../widget";
 import { LotusWidgetActions } from "./lotusActions";
+import { setScreenshareAudioMuted$ } from "./lotusScreenshareAudio";
 
 /**
  * Handle the host's `io.lotus.set_deafen` toWidget action, replacing cinny's
@@ -41,22 +35,23 @@ import { LotusWidgetActions } from "./lotusActions";
  * volume/mute state.
  *
  * `screenshareAudioMuted` is a NARROWER, independent host control (drop shared
- * tab/game audio while still hearing voices). EC has no global per-source mute,
- * so that one still has to go through
- * `RemoteParticipant.setVolume(volume, Track.Source.ScreenShareAudio)` — whose
- * verified signature in livekit-client ^2.18.1 is
- * `setVolume(volume, source?: Track.Source.Microphone | Track.Source.ScreenShareAudio)`.
- * It is re-applied to late joiners via `RoomEvent.ParticipantConnected`, and we
- * only ever restore participants we muted ourselves. Known limitation: EC's own
- * screenshare volume slider writes the same `volumeMap`, so a user who moves
- * that slider while screenshare audio is host-muted wins; the mic path (the
- * actual deafen) is no longer affected by that race at all.
+ * tab/game audio while still hearing voices). It goes through the same
+ * mechanism as deafen — a global behavior (`muteScreenshareAudio$`) that
+ * `LivekitRoomAudioRenderer` turns into the `muted` prop of every
+ * `Track.Source.ScreenShareAudio` element — rather than
+ * `RemoteParticipant.setVolume(0, ScreenShareAudio)`: EC's own
+ * `createVolumeControls` writes volume 1 through that very setter whenever a
+ * new screenshare media item resolves, so a sharer who stopped and re-shared
+ * (or a late joiner's share) came back at full volume while the host's button
+ * still said "Unmute Screenshare Audio".
  *
  * Undeafen restores the user's OWN output-enabled state as it was before the
  * deafen (and never touches the `mute-all-audio` setting), so a user who had
  * already muted all audio themselves stays muted.
  *
- * State is closure-scoped (per invocation, matching the sibling lotus modules).
+ * State is closure-scoped (per invocation, matching the sibling lotus modules);
+ * the screenshare-audio flag additionally lives in `setScreenshareAudioMuted$`
+ * so the renderer can read it, and is reset on teardown.
  * Applying the same state twice is a no-op, so the host's
  * `CallControl.resendForkState()` after a reconnect is safe. The host re-sends
  * the current state on every call join (CallControl.forceState), so a fresh
@@ -64,7 +59,7 @@ import { LotusWidgetActions } from "./lotusActions";
  *
  * No effect unless the host sends the action. Returns a teardown function.
  */
-export function startLotusDeafen(vm: CallViewModel): () => void {
+export function startLotusDeafen(): () => void {
   const w = widget;
   if (!w) return () => undefined;
 
@@ -107,53 +102,6 @@ export function startLotusDeafen(vm: CallViewModel): () => void {
     }
   });
 
-  // Participants whose screenshare audio WE turned down, so undoing the host's
-  // screenshare mute never writes a volume to anyone else.
-  const screenshareMuted = new Set<RemoteParticipant>();
-
-  const applyToParticipant = (p: RemoteParticipant): void => {
-    if (screenshareAudioMuted) {
-      p.setVolume(0, Track.Source.ScreenShareAudio);
-      screenshareMuted.add(p);
-    } else if (screenshareMuted.delete(p)) {
-      p.setVolume(1, Track.Source.ScreenShareAudio);
-    }
-  };
-
-  const applyToRoom = (room: LivekitRoom): void =>
-    room.remoteParticipants.forEach(applyToParticipant);
-
-  // Per-room ParticipantConnected listeners, so LATE JOINERS pick up the
-  // current screenshare-audio mute the moment they connect. Drive off the
-  // local participant's connection(s), not `livekitRoomItems$` — that stream is
-  // empty until a matrix-validated REMOTE member resolves, so listeners would
-  // be attached too late for the first joiner (the sibling lotus modules all
-  // use `allConnections$` for the same reason).
-  const roomListeners = new Map<LivekitRoom, () => void>();
-  let rooms: LivekitRoom[] = [];
-
-  const sub = vm.allConnections$.subscribe((data) => {
-    const next = data.getConnections().map((c) => c.livekitRoom);
-    rooms = next;
-    // Detach listeners for rooms that went away.
-    for (const [room, off] of roomListeners) {
-      if (!next.includes(room)) {
-        off();
-        roomListeners.delete(room);
-      }
-    }
-    // Attach to new rooms + apply the current state to their participants.
-    for (const room of next) {
-      if (!roomListeners.has(room)) {
-        room.on(RoomEvent.ParticipantConnected, applyToParticipant);
-        roomListeners.set(room, () =>
-          room.off(RoomEvent.ParticipantConnected, applyToParticipant),
-        );
-        applyToRoom(room);
-      }
-    }
-  });
-
   const handler = (ev: CustomEvent<IWidgetApiRequest>): void => {
     w.api.transport.reply(ev.detail, {});
     const data = ev.detail.data as
@@ -168,20 +116,18 @@ export function startLotusDeafen(vm: CallViewModel): () => void {
       `[lotus] set_deafen: deafened=${deafened} screenshareAudioMuted=${screenshareAudioMuted}`,
     );
     applyGlobalMute();
-    rooms.forEach(applyToRoom);
+    if (setScreenshareAudioMuted$.value !== screenshareAudioMuted)
+      setScreenshareAudioMuted$.next(screenshareAudioMuted);
   };
 
   w.lazyActions.on(LotusWidgetActions.SetDeafen, handler);
   return () => {
-    sub.unsubscribe();
-    for (const off of roomListeners.values()) off();
-    roomListeners.clear();
     // Leave the user's own output state as they had it before deafen.
     if (deafened) {
       deafened = false;
       applyGlobalMute();
     }
-    screenshareMuted.clear();
+    if (setScreenshareAudioMuted$.value) setScreenshareAudioMuted$.next(false);
     audioSub.unsubscribe();
     w.lazyActions.off(LotusWidgetActions.SetDeafen, handler);
   };
diff --git a/src/lotus/lotusScreenshareAudio.ts b/src/lotus/lotusScreenshareAudio.ts
new file mode 100644
index 00000000..ab9f404d
--- /dev/null
+++ b/src/lotus/lotusScreenshareAudio.ts
@@ -0,0 +1,27 @@
+/*
+Copyright 2026 Lotus Guild
+
+SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-Element-Commercial
+Please see LICENSE in the repository root for full details.
+*/
+
+import { BehaviorSubject } from "rxjs";
+
+import { globalScope } from "../state/ObservableScope";
+
+/**
+ * Whether the host has muted remote screenshare AUDIO (`io.lotus.set_deafen`
+ * `screenshareAudioMuted`). Consumed by `LivekitRoomAudioRenderer` exactly the
+ * way `muteAllAudio$` is: it becomes the `muted` prop of every
+ * `Track.Source.ScreenShareAudio` element, so it survives re-renders, applies
+ * to tracks that are published LATER (a sharer stopping and re-sharing, a late
+ * joiner) and never fights EC's per-tile volume controls — which was the
+ * failure mode of the old `RemoteParticipant.setVolume(0, ScreenShareAudio)`
+ * approach: `createVolumeControls` writes its own volume (1) through the same
+ * sink the moment a new screenshare media item appears, un-muting it.
+ */
+export const setScreenshareAudioMuted$ = new BehaviorSubject(false);
+
+export const muteScreenshareAudio$ = globalScope.behavior(
+  setScreenshareAudioMuted$,
+);
diff --git a/src/room/InCallView.tsx b/src/room/InCallView.tsx
index 110863e1..ac987193 100644
--- a/src/room/InCallView.tsx
+++ b/src/room/InCallView.tsx
@@ -73,6 +73,7 @@ import { matrixRTCMode as matrixRTCModeSetting } from "../settings/settings";
 import { ReactionsReader } from "../reactions/ReactionsReader";
 import { LivekitRoomAudioRenderer } from "../livekit/MatrixAudioRenderer.tsx";
 import { muteAllAudio$ } from "../state/MuteAllAudioModel.ts";
+import { muteScreenshareAudio$ } from "../lotus/lotusScreenshareAudio";
 import { useMediaDevices } from "../MediaDevicesContext.ts";
 import { EarpieceOverlay } from "./EarpieceOverlay.tsx";
 import {
@@ -261,6 +262,8 @@ export const InCallView: FC<InCallViewProps> = ({
   const { showControls, header: headerStyle } = useUrlParams();
 
   const muteAllAudio = useBehavior(muteAllAudio$);
+  // [lotus] host-driven "Mute Screenshare Audio" (io.lotus.set_deafen).
+  const muteScreenshareAudio = useBehavior(muteScreenshareAudio$);
   const toggleAudio = useBehavior(muteStates.audio.toggle$);
   const toggleVideo = useBehavior(muteStates.video.toggle$);
   const setAudioEnabled = useBehavior(muteStates.audio.setEnabled$);
@@ -309,7 +312,7 @@ export const InCallView: FC<InCallViewProps> = ({
   // [lotus] Handle the host's io.lotus.set_deafen action to silence remote
   // audio (and optionally screenshare audio) at the LiveKit source. No-op
   // unless the host sends the action.
-  useEffect(() => startLotusDeafen(vm), [vm]);
+  useEffect(() => startLotusDeafen(), []);
 
   const fatalCallError = useBehavior(vm.fatalError$);
   // Stop the rendering and throw for the error boundary
@@ -657,6 +660,7 @@ export const InCallView: FC<InCallViewProps> = ({
           livekitRoom={livekitRoom}
           validIdentities={participants}
           muted={muteAllAudio}
+          screenshareAudioMuted={muteScreenshareAudio}
         />
       ))}
       {renderContent()}