src/app/features/settings/general/General.tsx

@@ -2037,7 +2037,7 @@ export function Calls() {
                 <Text size="L400">Model</Text>
                 <SettingTile
                   title="ML model"
-                  description="Choose the machine learning model used for noise removal. Heavier models clean more aggressively at a higher CPU cost. CPU figures are the share of one processor core during a call on a mid-range desktop; Task Manager shows that divided by your number of cores, and slower computers use more."
+                  description="Choose the machine learning model used for noise removal. DTLN is the best all-rounder; DeepFilterNet removes the most noise at a higher CPU cost. None of the models remove other people talking in the background. CPU figures are the share of one processor core during a call on a mid-range desktop; Task Manager shows that divided by your number of cores, and slower computers use more."
                   after={
                     <SettingsSelect<DenoiseModelId>
                       value={callDenoiseModel}
@@ -2066,7 +2066,7 @@ export function Calls() {
                         [
                           { label: 'CPU (one core)', value: selectedDenoiseModel.cpuUsage },
                           { label: 'Voice quality', value: selectedDenoiseModel.voiceQuality },
-                          { label: 'Transients', value: selectedDenoiseModel.transients },
+                          { label: 'Typing & clicks', value: selectedDenoiseModel.transients },
                           { label: 'Download', value: selectedDenoiseModel.binarySize },
                         ] as const
                       ).map((stat) => (
@@ -2113,7 +2113,7 @@ export function Calls() {
                       </Box>
                       <Box grow="Yes">
                         <Text size="T200" priority="300">
-                          Transients
+                          Typing & clicks
                         </Text>
                       </Box>
                     </Box>

src/app/state/settings.ts

@@ -377,9 +377,11 @@ const defaultSettings: Settings = {
 
   cameraOnJoin: false,
   // Tier default stays browser-native (known-good; best-perceived in testing so
-  // far). If a user opts into the ML tier, default to the highest-quality model.
+  // far). If a user opts into the ML tier, default to DTLN: in our measurements
+  // it keeps voices the most natural, removes fans and typing well, and costs a
+  // third of DeepFilterNet's CPU.
   callNoiseSuppression: 'browser',
-  callDenoiseModel: 'deepfilternet',
+  callDenoiseModel: 'dtln',
   // "Series suppression" (stack the browser's native NS before the ML model) is
   // off by default — best practice is a single NS stage; it's an opt-in test aid.
   callDenoiseNativeNS: false,

src/app/utils/lotusDenoiseUtils.test.ts

@@ -5,10 +5,10 @@ import { DENOISE_MODELS, ML_DENOISE_REQUIREMENTS, isMLDenoiseSupported } from '.
 
 // ── Model catalog (data integrity) ──────────────────────────────────────────
 
-test('DENOISE_MODELS lists the four models ordered best-quality (highest CPU) first', () => {
+test('DENOISE_MODELS lists the four models most-recommended first', () => {
   assert.deepEqual(
     DENOISE_MODELS.map((m) => m.id),
-    ['deepfilternet', 'dtln', 'rnnoise', 'speex'],
+    ['dtln', 'deepfilternet', 'rnnoise', 'speex'],
   );
 });
 

src/app/utils/lotusDenoiseUtils.ts

@@ -1,8 +1,8 @@
 /**
  * Detection utilities + model catalog for Lotus ML noise suppression
- * (DeepFilterNet 3 / DTLN / RNNoise / Speex). The catalog is ordered by
- * quality (and, correspondingly, CPU cost) — highest first — and drives the
- * order of the model dropdown in settings.
+ * (DTLN / DeepFilterNet 3 / RNNoise / Speex). The catalog is ordered by
+ * recommendation — most recommended first — and drives the order of the model
+ * dropdown in settings.
  */
 
 import { DenoiseModelId } from '../state/settings';
@@ -10,6 +10,9 @@ import { DenoiseModelId } from '../state/settings';
 // CPU = share of ONE core while in a call, measured with each model running
 // exactly as Lotus Call runs it (Ryzen 7 PRO 4750GE, Chromium); download = bytes
 // actually fetched on first use. Re-measure when a model or its settings change.
+// Ratings come from scoring each model against the clean voice (PESQ/STOI) on
+// real speech mixed with fan, keyboard, street and babble noise at SNR 5 and
+// 15 dB; `transients` is how well it removes typing and clicks.
 export type DenoiseModel = {
   id: DenoiseModelId;
   name: string;
@@ -20,25 +23,25 @@ export type DenoiseModel = {
   voiceQuality: 'Moderate' | 'High' | 'Very High';
 };
 
-// Ordered best-quality (highest CPU) first — this is the dropdown order.
+// Ordered most-recommended first — this is the dropdown order.
 export const DENOISE_MODELS: DenoiseModel[] = [
   {
-    id: 'deepfilternet',
-    name: 'DeepFilterNet 3 (beta)',
+    id: 'dtln',
+    name: 'DTLN (recommended)',
     description:
-      'Studio-grade deep-learning model (48 kHz fullband, ONNX). Best quality; highest CPU and a larger one-time download.',
-    cpuUsage: '~30%',
-    binarySize: '~18 MB',
+      'Keeps voices the most natural of the four. Removes fans and typing well, but not traffic or people talking in the background.',
+    cpuUsage: '~9%',
+    binarySize: '~20 MB',
     transients: 'Excellent',
     voiceQuality: 'Very High',
   },
   {
-    id: 'dtln',
-    name: 'DTLN (beta)',
+    id: 'deepfilternet',
+    name: 'DeepFilterNet 3 (strongest)',
     description:
-      'Dual-signal deep-learning model (16 kHz). Strong on transient noise; moderate CPU.',
-    cpuUsage: '~9%',
-    binarySize: '~20 MB',
+      'Removes the most noise, including loud typing and traffic, but voices can sound slightly muffled. Uses about three times the CPU of DTLN.',
+    cpuUsage: '~30%',
+    binarySize: '~18 MB',
     transients: 'Excellent',
     voiceQuality: 'High',
   },
@@ -46,17 +49,17 @@ export const DENOISE_MODELS: DenoiseModel[] = [
     id: 'rnnoise',
     name: 'RNNoise',
     description:
-      'Lightweight hybrid model (48 kHz). Very low CPU; good for steady noise like fans, but can sound processed at full strength.',
+      'Very light. Handles steady noise like fans and traffic; does little for typing and can sound slightly processed.',
     cpuUsage: '~3%',
     binarySize: '< 1 MB',
-    transients: 'Good',
+    transients: 'Poor',
     voiceQuality: 'Moderate',
   },
   {
     id: 'speex',
-    name: 'Speex (Legacy)',
+    name: 'Speex (legacy)',
     description:
-      'Classic DSP noise suppressor. Minimal CPU, gentlest on voice; weakest suppression.',
+      'Classic suppressor kept for very slow computers. Barely reduces noise in our tests; prefer any other option.',
     cpuUsage: '~2%',
     binarySize: '< 1 MB',
     transients: 'Poor',