diff --git a/video/out/wayland_common.c b/video/out/wayland_common.c
index 78ce86ca74..3b69f60947 100644
--- a/video/out/wayland_common.c
+++ b/video/out/wayland_common.c
@@ -306,18 +306,21 @@ static void pointer_handle_button(void *data, struct wl_pointer *wl_pointer,
     if (button)
         mp_input_put_key(wl->vo->input_ctx, button | state | s->mpmod);
 
+    uint32_t edges;
     if (!mp_input_test_dragging(wl->vo->input_ctx, wl->mouse_x, wl->mouse_y) &&
-        !wl->locked_size && (button == MP_MBTN_LEFT) && (state == MP_KEY_STATE_DOWN))
+        !wl->locked_size && (button == MP_MBTN_LEFT) && (state == MP_KEY_STATE_DOWN) &&
+        !wl->vo_opts->border && check_for_resize(wl, wl->opts->edge_pixels_pointer, &edges))
     {
-        uint32_t edges;
         // Implement an edge resize zone if there are no decorations
-        if (!wl->vo_opts->border && check_for_resize(wl, wl->opts->edge_pixels_pointer, &edges)) {
-            xdg_toplevel_resize(wl->xdg_toplevel, s->seat, serial, edges);
-        } else {
-            xdg_toplevel_move(wl->xdg_toplevel, s->seat, serial);
-        }
-        // Explicitly send an UP event after the client finishes a move/resize
+        xdg_toplevel_resize(wl->xdg_toplevel, s->seat, serial, edges);
+        // Explicitly send an UP event after the client finishes a resize
         mp_input_put_key(wl->vo->input_ctx, button | MP_KEY_STATE_UP);
+    } else if (state == MP_KEY_STATE_DOWN) {
+        // Save the serial and seat for voctrl-initialized dragging requests.
+        s->pointer_serial = serial;
+        wl->last_button_seat = s;
+    } else {
+        wl->last_button_seat = NULL;
     }
 }
 
@@ -427,12 +430,14 @@ static void touch_handle_down(void *data, struct wl_touch *wl_touch,
     wl->mouse_y = wl_fixed_to_int(y_w) * wl->scaling;
 
     enum xdg_toplevel_resize_edge edge;
-    if (!mp_input_test_dragging(wl->vo->input_ctx, wl->mouse_x, wl->mouse_y)) {
-        if (check_for_resize(wl, wl->opts->edge_pixels_touch, &edge)) {
-            xdg_toplevel_resize(wl->xdg_toplevel, s->seat, serial, edge);
-        } else  {
-            xdg_toplevel_move(wl->xdg_toplevel, s->seat, serial);
-        }
+    if (!mp_input_test_dragging(wl->vo->input_ctx, wl->mouse_x, wl->mouse_y) &&
+        check_for_resize(wl, wl->opts->edge_pixels_touch, &edge))
+    {
+        xdg_toplevel_resize(wl->xdg_toplevel, s->seat, serial, edge);
+    } else {
+        // Save the serial and seat for voctrl-initialized dragging requests.
+        s->pointer_serial = serial;
+        wl->last_button_seat = s;
     }
 
     mp_input_set_mouse_pos(wl->vo->input_ctx, wl->mouse_x, wl->mouse_y);
@@ -445,6 +450,7 @@ static void touch_handle_up(void *data, struct wl_touch *wl_touch,
     struct vo_wayland_seat *s = data;
     struct vo_wayland_state *wl = s->wl;
     mp_input_put_key(wl->vo->input_ctx, MP_MBTN_LEFT | MP_KEY_STATE_UP);
+    wl->last_button_seat = NULL;
 }
 
 static void touch_handle_motion(void *data, struct wl_touch *wl_touch,
@@ -1908,6 +1914,8 @@ static void remove_seat(struct vo_wayland_seat *seat)
 
     MP_VERBOSE(seat->wl, "Deregistering seat 0x%x\n", seat->id);
     wl_list_remove(&seat->link);
+    if (seat == seat->wl->last_button_seat)
+        seat->wl->last_button_seat = NULL;
     if (seat->keyboard)
         wl_keyboard_destroy(seat->keyboard);
     if (seat->pointer)
@@ -2199,6 +2207,18 @@ static void wayland_dispatch_events(struct vo_wayland_state *wl, int nfds, int64
     wl_display_dispatch_pending(wl->display);
 }
 
+static void begin_dragging(struct vo_wayland_state *wl)
+{
+    struct vo_wayland_seat *s = wl->last_button_seat;
+    if (!mp_input_test_dragging(wl->vo->input_ctx, wl->mouse_x, wl->mouse_y) &&
+        !wl->locked_size && s)
+    {
+        xdg_toplevel_move(wl->xdg_toplevel, s->seat, s->pointer_serial);
+        wl->last_button_seat = NULL;
+        mp_input_put_key(wl->vo->input_ctx, MP_INPUT_RELEASE_ALL);
+    }
+}
+
 /* Non-static */
 int vo_wayland_allocate_memfd(struct vo *vo, size_t size)
 {
@@ -2369,6 +2389,9 @@ int vo_wayland_control(struct vo *vo, int *events, int request, void *arg)
         *(double *)arg = wl->scaling;
         return VO_TRUE;
     }
+    case VOCTRL_BEGIN_DRAGGING:
+        begin_dragging(wl);
+        return VO_TRUE;
     case VOCTRL_UPDATE_WINDOW_TITLE:
         return update_window_title(wl, (const char *)arg);
     case VOCTRL_SET_CURSOR_VISIBILITY:
diff --git a/video/out/wayland_common.h b/video/out/wayland_common.h
index 0554d308ba..721d4fb2eb 100644
--- a/video/out/wayland_common.h
+++ b/video/out/wayland_common.h
@@ -22,6 +22,8 @@
 #include "input/event.h"
 #include "vo.h"
 
+struct vo_wayland_seat;
+
 typedef struct {
     uint32_t format;
     uint32_t padding;
@@ -155,6 +157,7 @@ struct vo_wayland_state {
     struct wl_surface      *cursor_surface;
     bool                    cursor_visible;
     int                     allocated_cursor_scale;
+    struct vo_wayland_seat *last_button_seat;
 };
 
 bool vo_wayland_check_visible(struct vo *vo);