diff --git a/kubernetes/apps/music-assistant/deployment.yaml b/kubernetes/apps/music-assistant/deployment.yaml
index fe2f655e..c69229e3 100644
--- a/kubernetes/apps/music-assistant/deployment.yaml
+++ b/kubernetes/apps/music-assistant/deployment.yaml
@@ -75,9 +75,15 @@ spec:
           resources:
             requests:
               cpu: 250m
-              memory: 512Mi
+              memory: 1Gi
             limits:
-              memory: 2Gi
+              # The Smart Fades audio-analysis provider (MA 2.9+) loads an ML
+              # model and refuses to start if it detects < 4GB via the pod's
+              # cgroup memory limit ("at least 4GB of RAM is required, 2.0GB
+              # detected"). Upstream recommends 6GB+; the 4GB minimum risks
+              # OOM under load. k8s-edge (12GB VM) has the headroom, so the
+              # limit is set to the recommended 6Gi rather than the 4GB floor.
+              memory: 6Gi
           securityContext:
             # Matches the LXC docker-compose: SYS_ADMIN + DAC_READ_SEARCH
             # and AppArmor unconfined are required by the upstream image