diff --git a/hosts/idols-aquamarine/asr.nix b/hosts/idols-aquamarine/asr.nix
index 1eaeae9e..89ed94c4 100644
--- a/hosts/idols-aquamarine/asr.nix
+++ b/hosts/idols-aquamarine/asr.nix
@@ -66,17 +66,24 @@ in
 {
   # vinput on the desktop sends raw PCM here and gets text back; the endpoint is
   # the OpenAI transcription API, so any client of that shape works.
+  # FreeBind: the LAN address may not exist yet at boot.
+  systemd.sockets.whisper-asr = {
+    wantedBy = [ "sockets.target" ];
+    listenStreams = [ "${ipv4}:${toString port}" ];
+    socketConfig.FreeBind = true;
+  };
+
   systemd.services.whisper-asr = {
     description = "OpenAI-compatible Whisper transcription service (Intel NPU)";
+    # Start eagerly so the ~70 s NPU compile does not land on the first request.
     wantedBy = [ "multi-user.target" ];
-    after = [ "network.target" ];
+    requires = [ "whisper-asr.socket" ];
+    after = [ "whisper-asr.socket" ];
 
     environment = {
       WHISPER_MODEL_DIR = "${whisperModel}";
       WHISPER_MODEL_NAME = "whisper-large-v3-turbo";
       WHISPER_DEVICE = "NPU";
-      WHISPER_HOST = ipv4;
-      WHISPER_PORT = toString port;
       # Compiling the graph for the NPU takes ~70 s; cache it so restarts and
       # reboots reuse the compiled blob instead of paying that again.
       OV_CACHE_DIR = "/var/cache/whisper-asr";
diff --git a/hosts/idols-aquamarine/asr/server.py b/hosts/idols-aquamarine/asr/server.py
index f6decac1..c94a1f24 100644
--- a/hosts/idols-aquamarine/asr/server.py
+++ b/hosts/idols-aquamarine/asr/server.py
@@ -27,8 +27,6 @@ DEVICE = os.environ.get("WHISPER_DEVICE", "NPU")
 # Empty auto-detects the language per request; set it to pin one language.
 DEFAULT_LANGUAGE = os.environ.get("WHISPER_LANGUAGE") or ""
 CACHE_DIR = os.environ.get("OV_CACHE_DIR", "")
-LISTEN_HOST = os.environ.get("WHISPER_HOST", "127.0.0.1")
-LISTEN_PORT = int(os.environ.get("WHISPER_PORT", "18080"))
 SAMPLE_RATE = 16000
 MAX_AUDIO_SECONDS = int(os.environ.get("WHISPER_MAX_AUDIO_SECONDS") or "120")
 
@@ -197,4 +195,5 @@ async def transcriptions(
 
 
 if __name__ == "__main__":
-    uvicorn.run(app, host=LISTEN_HOST, port=LISTEN_PORT, log_level="info")
+    # Listens on the socket systemd passes in (whisper-asr.socket).
+    uvicorn.run(app, fd=3, log_level="info")