diff --git a/tools/machine-learning/mujoco/packages/mujoco-interactive-viewer/src/mujoco_interactive_viewer/viewer.py b/tools/machine-learning/mujoco/packages/mujoco-interactive-viewer/src/mujoco_interactive_viewer/viewer.py
index 4665f7d695..ea92c9570f 100644
--- a/tools/machine-learning/mujoco/packages/mujoco-interactive-viewer/src/mujoco_interactive_viewer/viewer.py
+++ b/tools/machine-learning/mujoco/packages/mujoco-interactive-viewer/src/mujoco_interactive_viewer/viewer.py
@@ -589,6 +589,12 @@ def _handle_perturbation(self, mods: int) -> None:
                 )
         self.perturbation.active = perturbation_kind
 
+    def track_with_camera(self, body: int | str) -> None:
+        bodyid = self.model.body(body).id
+        self.camera.type = mujoco.mjtCamera.mjCAMERA_TRACKING.value
+        self.camera.trackbodyid = bodyid
+        self.camera.fixedcamid = -1
+
     def _handle_selection(self, mods: int) -> None:
         class Mode(Enum):
             Select = 1
@@ -603,7 +609,10 @@ class Mode(Enum):
             mode = None
             if self._interaction_state.left_double_click_pressed:
                 mode = Mode.Select
-            elif self._interaction_state.right_double_click_pressed:
+            elif (
+                self._interaction_state.right_double_click_pressed
+                and mods != glfw.MOD_CONTROL
+            ):
                 mode = Mode.LookAt
             elif (
                 self._interaction_state.right_double_click_pressed
diff --git a/tools/machine-learning/mujoco/scripts/mujoco-walking.py b/tools/machine-learning/mujoco/scripts/mujoco-walking.py
index c28507374e..d5297e040a 100644
--- a/tools/machine-learning/mujoco/scripts/mujoco-walking.py
+++ b/tools/machine-learning/mujoco/scripts/mujoco-walking.py
@@ -53,7 +53,7 @@ def main(*, throw_tomatoes: bool, load_policy: str | None) -> None:
 
     while viewer.is_alive:
         start_time = time.time()
-        # viewer.camera.lookat[:] = env.data.site("Robot").xpos
+        viewer.track_with_camera("Nao")
         observation, reward, _terminated, _truncated, infos = env.step(action)
         if model:
             action, _ = model.predict(observation, deterministic=True)