Describe clearly you don't need to download models on the server.

docs/Decisions.md

@@ -290,3 +290,7 @@ starts the weight download. Weights load exactly when they are needed or asked f
 (`loadStarted`, progress); the next load starts from clean progress. The explicit
 download is a full pipeline load, so it also warms up inference and surfaces the
 "Running on …" backend badge.
+**Amendment (2026-07-14):** while the engine toggle is on "server" (signed in,
+available), `ModelPicker` states explicitly that nothing is downloaded to the
+device; the Download button stays (with a matching tooltip) for on-device /
+offline preparation.

src/App.tsx

@@ -644,6 +644,7 @@ export function App() {
         disabled={busy || liveActive}
         onDownloadSelected={handleDownloadModel}
         downloading={loadStarted && !modelReady}
+        serverEngine={serverSelected}
       />
 
       {model?.multilingual && (

src/components/ModelPicker.tsx

@@ -23,6 +23,8 @@ interface ModelPickerProps {
   onDownloadSelected: () => void;
   /** True while a model load/download is in flight. */
   downloading: boolean;
+  /** True when transcription will run on the server — no download needed (D-020). */
+  serverEngine: boolean;
 }
 
 const DEVICE_OPTIONS: readonly { value: DevicePreference; label: string }[] = [
@@ -56,6 +58,7 @@ export function ModelPicker({
   disabled = false,
   onDownloadSelected,
   downloading,
+  serverEngine,
 }: ModelPickerProps) {
   // Size estimates depend on the dtype the worker will pick, which depends on
   // the device. Until the worker has reported one, assume WASM unless the
@@ -116,6 +119,13 @@ export function ModelPicker({
           Model switching is disabled while a transcription is running.
         </p>
       )}
+      {serverEngine && (
+        <p className="model-picker-hint server-engine-hint" role="status">
+          Transcription is set to run on the server — the selected model runs
+          there, so nothing is downloaded to this device. Downloading is only
+          useful if you plan to transcribe on-device or offline.
+        </p>
+      )}
       <p
         className={`offline-status${offlineReady && downloaded.size > 0 ? " offline-status-ready" : ""}`}
       >
@@ -164,7 +174,11 @@ export function ModelPicker({
                           type="button"
                           className="model-download-button"
                           disabled={disabled || downloading}
-                          title="Download this model now so it is available offline. Otherwise it downloads automatically when an on-device transcription starts."
+                          title={
+                            serverEngine
+                              ? "Not needed while transcribing on the server — download only to prepare on-device or offline use."
+                              : "Download this model now so it is available offline. Otherwise it downloads automatically when an on-device transcription starts."
+                          }
                           onClick={onDownloadSelected}
                         >
                           {downloading ? "Downloading…" : "Download"}

src/index.css

@@ -207,6 +207,11 @@ body {
   font-size: 0.85rem;
 }
 
+/* Server engine active — no model download needed on this device (D-020). */
+.server-engine-hint {
+  color: #9ad6ae;
+}
+
 .device-select,
 .transcription-options label {
   display: flex;