diff --git a/k8s/configmap.yaml b/k8s/configmap.yaml
index 6c61319..e5040f8 100644
--- a/k8s/configmap.yaml
+++ b/k8s/configmap.yaml
@@ -12,7 +12,7 @@ data:
   LLM_MODEL: "qwen2.5-14b"
   LLM_TIMEOUT: "5m"
   TRANSCRIPTION_PROVIDERS: "voxtral-small"
-  WHISPERX_URL: "http://10.2.3.5:8001"
+  WHISPERX_URL: ""
   WHISPERX_TIMEOUT: "10m"
   WHISPERX_LEAD_SILENCE: "800ms"
   # Fill these after Qwen2-Audio and Voxtral are exposed as OpenAI-compatible