diff --git a/addons/voice/manifest.json b/addons/voice/manifest.json
index a0d6d9277..3a3eb1d68 100644
--- a/addons/voice/manifest.json
+++ b/addons/voice/manifest.json
@@ -14,9 +14,9 @@
       "entry": "ui/voice.js",
       "css": ["ui/voice.css"],
       "sdk": { "min": 2, "max": 2 },
-      "pages": ["recognition", "library", "create", "speech"],
-      "navigation": ["speech", "library"],
-      "mainPage": "recognition",
+      "pages": ["voice", "recognition", "library", "create", "speech"],
+      "navigation": ["voice"],
+      "mainPage": "voice",
       "icon": "ui/icon.png",
       "title": "Voice",
       "description": "Recognise who is speaking, keep the voices that text-to-speech can use, and speak text with them."
diff --git a/addons/voice/ui/src/Speech.jsx b/addons/voice/ui/src/Speech.jsx
index b0f0f10e6..e3550cc0d 100644
--- a/addons/voice/ui/src/Speech.jsx
+++ b/addons/voice/ui/src/Speech.jsx
@@ -6,7 +6,7 @@ export function createSpeech(ctx) {
   const { h, t, ui, hooks, viewer } = ctx
   const { useState, useEffect, useMemo } = ctx.React
 
-  function Speech({ path }) {
+  function Speech({ path, tabs }) {
     const [model, setModel] = useState(ctx.modelFrom(path))
     const [models, setModels] = useState({ status: 'loading', list: [] })
     const [profiles, setProfiles] = useState([])
@@ -57,6 +57,7 @@ export function createSpeech(ctx) {
     return (
       <main className="page vc-speech" data-testid="voice-speech">
         <ui.PageHeader title={t('speech.title')} supporting={t('speech.lede')} />
+        {tabs}
         <label className="vc-field">
           <span>{t('speech.model')}</span>
           <ui.ModelSelector value={model} onChange={setModel} options={models.list.map(m => m.id)} loading={models.status === 'loading'} />
diff --git a/addons/voice/ui/src/VoiceHub.jsx b/addons/voice/ui/src/VoiceHub.jsx
new file mode 100644
index 000000000..4f48cdc5b
--- /dev/null
+++ b/addons/voice/ui/src/VoiceHub.jsx
@@ -0,0 +1,72 @@
+// The one page of the Voice add-on. Its sections are tabs: Recognition,
+// Library, Text to speech and From a recording. The tab is the first segment of
+// the page path (voice/library), the same convention the Studio add-on uses;
+// the profile form is a view of the Library tab (voice/library/new).
+const DEFAULT_TAB = 'recognition'
+const TABS = ['recognition', 'library', 'speech', 'recording']
+
+export function createVoiceHub(ctx, api, identity, sections) {
+  const { React, h, Fragment, t } = ctx
+  const { useEffect, useMemo, useState } = React
+
+  // Where an older page address goes now. `search` is the query string the
+  // old address carried (a model, a selected profile, a voice); it is kept.
+  function redirectFor(legacy, path, search = '') {
+    const params = new URLSearchParams(search)
+    let tab = legacy
+    let rest = path
+    if (legacy === 'recognition' && params.get('tab') === 'recording') { tab = 'recording'; rest = '' }
+    if (legacy === 'recognition') params.delete('tab')
+    if (legacy === 'create') { tab = 'library'; rest = 'new' }
+    const query = params.toString()
+    return `voice/${tab}${rest ? `/${rest}` : ''}${query ? `?${query}` : ''}`
+  }
+
+  function Hub({ path }) {
+    const segments = String(path || '').split('/').filter(Boolean)
+    const tab = TABS.includes(segments[0]) ? segments[0] : DEFAULT_TAB
+    const rest = segments.slice(tab === segments[0] ? 1 : 0).join('/')
+    const isAdmin = ctx.viewer.isAdmin()
+
+    // The counts on the tabs. The people this browser knows come from the
+    // recognition tab; the saved voices are read once for an administrator and
+    // then follow the library tab.
+    const [speakers, setSpeakers] = useState(() => identity.loadList('localai_voice_enrollments').length)
+    const [library, setLibrary] = useState({ loading: true, count: 0 })
+    useEffect(() => {
+      if (!isAdmin) return undefined
+      let live = true
+      api.voiceProfiles.list()
+        .then((body) => { if (live) setLibrary({ loading: false, count: (body?.data || []).length }) })
+        .catch(() => { if (live) setLibrary({ loading: false, count: 0 }) })
+      return () => { live = false }
+    }, [isAdmin])
+    const onLibraryCount = useMemo(() => (count) => setLibrary({ loading: false, count }), [])
+
+    const label = {
+      recognition: <>{t('voices.tab.recognition')} <span className="dk-hubtab-count">{speakers}</span></>,
+      library: <>{t('voices.tab.library')} <span className="dk-hubtab-count">{library.loading ? '' : library.count}</span></>,
+      speech: t('voices.tab.speech'),
+      recording: t('voices.tab.recording'),
+    }
+    const shown = TABS.filter((key) => key !== 'library' || isAdmin)
+    const tabs = (
+      <nav className="dk-tabs idn-tabs vc-tabs" role="tablist" aria-label={t('voices.tabs')} data-testid="voice-tabs">
+        {shown.map((key) => (
+          <ctx.Link key={key} className="dk-tab" role="tab" page={`voice/${key}`} data-tab={key} aria-selected={tab === key} aria-current={tab === key ? 'page' : undefined}>
+            {label[key]}
+          </ctx.Link>
+        ))}
+      </nav>
+    )
+
+    if (tab === 'library') {
+      const Page = rest === 'new' ? sections.create : sections.library
+      return <Page tabs={tabs} onLibraryCount={onLibraryCount} />
+    }
+    if (tab === 'speech') return <sections.speech path={rest} tabs={tabs} />
+    return <sections.recognition path={rest} tab={tab === 'recording' ? 'recording' : 'speakers'} tabs={tabs} onSpeakerCount={setSpeakers} />
+  }
+
+  return { Hub, redirectFor }
+}
diff --git a/addons/voice/ui/src/VoiceLibrary.jsx b/addons/voice/ui/src/VoiceLibrary.jsx
index aaa373e95..87acd2c66 100644
--- a/addons/voice/ui/src/VoiceLibrary.jsx
+++ b/addons/voice/ui/src/VoiceLibrary.jsx
@@ -295,7 +295,7 @@ JSON`
             type="button"
             className="dk-btn dk-btn--primary"
             disabled={cloningModels.length === 0}
-            onClick={() => { close(); ctx.sdk.navigation.navigate(`speech?voice=${encodeURIComponent(profile.id)}`) }}
+            onClick={() => { close(); ctx.sdk.navigation.navigate(`voice/speech?voice=${encodeURIComponent(profile.id)}`) }}
           >
             <Icon name="headphones" /> {t('voiceLibrary.actions.useInTTS')}
           </button>
@@ -314,8 +314,10 @@ JSON`
     )
   }
 
-  function VoiceLibrary() {
+  function VoiceLibrary({ tabs, onLibraryCount }) {
     const { profiles, loading, error, refetch } = useProfiles()
+    // The count on the Library tab follows the list this tab shows.
+    useEffect(() => { if (!loading && !error) onLibraryCount?.(profiles.length) }, [loading, error, profiles.length, onLibraryCount])
     const setup = useModelSetup()
     const { cloningModels, modelsLoading } = setup
     const [search, setSearch] = useState('')
@@ -356,7 +358,7 @@ JSON`
     }, [selectedParam, profiles, openProfile])
 
     const newButton = (className) => admin && (
-      <ctx.Link page="create" className={className} data-testid="voice-library-new">
+      <ctx.Link page="voice/library/new" className={className} data-testid="voice-library-new">
         <Icon name="plus" /> {t('voiceLibrary.actions.create')}
       </ctx.Link>
     )
@@ -375,6 +377,7 @@ JSON`
             </div>
           )}
         />
+        {tabs}
 
         {error && (
           <div className="idn-error" role="alert">
diff --git a/addons/voice/ui/src/VoiceProfileCreate.jsx b/addons/voice/ui/src/VoiceProfileCreate.jsx
index 467b3f28d..08b0bc312 100644
--- a/addons/voice/ui/src/VoiceProfileCreate.jsx
+++ b/addons/voice/ui/src/VoiceProfileCreate.jsx
@@ -19,7 +19,7 @@ export function createVoiceProfileCreate(ctx, api) {
     )
   }
 
-  function VoiceProfileCreate() {
+  function VoiceProfileCreate({ tabs }) {
     const [audio, setAudio] = useState(null)
     const [audioProcessing, setAudioProcessing] = useState(false)
     const [audioError, setAudioError] = useState('')
@@ -116,7 +116,7 @@ export function createVoiceProfileCreate(ctx, api) {
         })
         const profile = await api.voiceProfiles.create(formData)
         ctx.toast(t('voiceCreate.toasts.created', { name: profile.name }), 'success')
-        ctx.sdk.navigation.navigate(`library?selected=${encodeURIComponent(profile.id)}`)
+        ctx.sdk.navigation.navigate(`voice/library?selected=${encodeURIComponent(profile.id)}`)
       } catch (err) {
         ctx.toast(err.message, 'error')
       } finally {
@@ -129,8 +129,9 @@ export function createVoiceProfileCreate(ctx, api) {
         <ui.PageHeader
           title={<><Icon name="mic" /> {t('voiceCreate.title')}</>}
           supporting={t('voiceCreate.subtitle')}
-          actions={<ctx.Link page="library" className="dk-btn dk-btn--secondary"><Icon name="arrow-left" /> {t('voiceCreate.actions.back')}</ctx.Link>}
+          actions={<ctx.Link page="voice/library" className="dk-btn dk-btn--secondary"><Icon name="arrow-left" /> {t('voiceCreate.actions.back')}</ctx.Link>}
         />
+        {tabs}
 
         <form className="voice-create-grid" onSubmit={submit}>
           <div className="voice-create-form">
@@ -218,7 +219,7 @@ export function createVoiceProfileCreate(ctx, api) {
             </section>
 
             <div className="voice-create-actions">
-              <ctx.Link page="library" className="dk-btn dk-btn--secondary">{t('voiceCreate.actions.cancel')}</ctx.Link>
+              <ctx.Link page="voice/library" className="dk-btn dk-btn--secondary">{t('voiceCreate.actions.cancel')}</ctx.Link>
               <button type="submit" className="dk-btn dk-btn--primary" disabled={!formReady || submitting}>
                 {submitting ? <><ui.Spinner size="sm" /> {t('voiceCreate.actions.saving')}</> : <><Icon name="save" /> {t('voiceCreate.actions.save')}</>}
               </button>
diff --git a/addons/voice/ui/src/VoiceRecognition.jsx b/addons/voice/ui/src/VoiceRecognition.jsx
index 2e00fc7b9..9cda25511 100644
--- a/addons/voice/ui/src/VoiceRecognition.jsx
+++ b/addons/voice/ui/src/VoiceRecognition.jsx
@@ -1,9 +1,7 @@
-// The Voices page: who is speaking. Ported from pages/VoiceRecognition.jsx.
-// Two unrelated things carry the word voice and are kept apart: Speakers
-// (voiceprints used to recognise who is speaking) and Speech voices (reference
-// recordings text-to-speech speaks in). Speech voices is the library page of
-// this add-on, so its tab is a link to that page; From a recording is the
-// diarization hand-off, a tab of this page.
+// The Recognition and From a recording tabs of the Voice page. Ported from
+// pages/VoiceRecognition.jsx. Speakers are voiceprints used to recognise who is
+// speaking; the speech voices (reference recordings text-to-speech speaks in)
+// are the Library tab. The tab strip comes from VoiceHub as `tabs`.
 import { createClipInput } from '../../../internal/identity-ui/src/ClipInput.jsx'
 import { createWorkbench } from '../../../internal/identity-ui/src/Workbench.jsx'
 import { createKnownList } from '../../../internal/identity-ui/src/KnownList.jsx'
@@ -37,9 +35,10 @@ export function createVoiceRecognition(ctx, api, identity) {
 
   // Shown in place of a page whose feature the signed-in user may not use. It
   // replaces a silent redirect: the person learns what is off and what turns it on.
-  function FeatureOff({ feature }) {
+  function FeatureOff({ feature, tabs }) {
     return (
       <main className="page idn-page idn-off" data-testid="voice-off">
+        {tabs}
         <section className="idn-needed dk-card">
           <span className="dk-badge dk-badge--warn"><Icon name="lock" /> {t('off.badge')}</span>
           <h1 className="idn-h2">{t(`off.${feature}.title`)}</h1>
@@ -71,27 +70,13 @@ export function createVoiceRecognition(ctx, api, identity) {
     return { ...state, refetch: useCallback(() => load(true), [load]) }
   }
 
-  function VoicesPage({ path }) {
-    const tab = ctx.useSearch().get('tab') === 'recording' ? 'recording' : 'speakers'
+  function VoicesPage({ path, tab = 'speakers', tabs, onSpeakerCount }) {
     const isAdmin = viewer.isAdmin()
     const [model, setModel] = useState(() => ctx.modelFrom(path) || '')
     const { models, loading: modelsLoading, refetch } = useSpeakerModels()
     const names = useMemo(() => models.map(m => m.id), [models])
-    const [library, setLibrary] = useState({ loading: true, count: 0 })
-    const [savedCount, setSavedCount] = useState(() => identity.loadList('localai_voice_enrollments').length)
     const needModel = !modelsLoading && names.length === 0
 
-    // The speech voices count on the tab. Only an administrator sees the tab,
-    // so only an administrator's page asks for the list.
-    useEffect(() => {
-      if (!isAdmin) return undefined
-      let live = true
-      api.voiceProfiles.list()
-        .then((body) => { if (live) setLibrary({ loading: false, count: (body?.data || []).length }) })
-        .catch(() => { if (live) setLibrary({ loading: false, count: 0 }) })
-      return () => { live = false }
-    }, [isAdmin])
-
     return (
       <main className="page idn-page" data-testid="voices-page">
         <ui.PageHeader
@@ -104,19 +89,11 @@ export function createVoiceRecognition(ctx, api, identity) {
           ) : null}
         />
 
-        <nav className="dk-tabs idn-tabs" role="tablist" aria-label={t('voices.tabs')}>
-          <ctx.Link className="dk-tab" role="tab" page="recognition" aria-selected={tab === 'speakers'} aria-current={tab === 'speakers' ? 'page' : undefined}>{t('voices.speakers')} <span className="dk-hubtab-count">{savedCount}</span></ctx.Link>
-          {isAdmin && (
-            <ctx.Link className="dk-tab" role="tab" page="library" aria-selected="false">
-              {t('voices.speech')} <span className="dk-hubtab-count">{library.loading ? '' : library.count}</span>
-            </ctx.Link>
-          )}
-          <ctx.Link className="dk-tab" role="tab" page="recognition" query="tab=recording" aria-selected={tab === 'recording'} aria-current={tab === 'recording' ? 'page' : undefined}>{t('voices.recording')}</ctx.Link>
-        </nav>
+        {tabs}
 
         {tab === 'speakers' && (
           <Workbench
-            kind="voice" model={model} canSpeech={isAdmin} onCount={setSavedCount}
+            kind="voice" model={model} canSpeech={isAdmin} onCount={onSpeakerCount}
             needed={needModel ? <ModelNeeded kind="voice" isAdmin={isAdmin} onInstalled={refetch} /> : null}
             more={needModel ? null : (
               <details className="idn-disclosure idn-disclosure--page idn-more">
@@ -150,7 +127,7 @@ export function createVoiceRecognition(ctx, api, identity) {
   // the gate of the page is the gate of the core route). The server checks every
   // /v1/voice/* call again. A viewer without it gets the notice and no call.
   function VoiceRecognition(props) {
-    if (!viewer.hasFeature('voice_recognition')) return <FeatureOff feature="voice" />
+    if (!viewer.hasFeature('voice_recognition')) return <FeatureOff feature="voice" tabs={props.tabs} />
     return <VoicesPage {...props} />
   }
 
diff --git a/addons/voice/ui/src/i18n/en.js b/addons/voice/ui/src/i18n/en.js
index e61c82392..d3acd54a8 100644
--- a/addons/voice/ui/src/i18n/en.js
+++ b/addons/voice/ui/src/i18n/en.js
@@ -15,9 +15,10 @@ export default {
   "voices.title": "Voices",
   "voices.lede": "Recognise who is speaking, and keep voices that text-to-speech can use.",
   "voices.tabs": "Voices sections",
-  "voices.speakers": "Speakers",
-  "voices.speech": "Speech voices",
-  "voices.recording": "From a recording",
+  "voices.tab.recognition": "Recognition",
+  "voices.tab.library": "Library",
+  "voices.tab.speech": "Text to speech",
+  "voices.tab.recording": "From a recording",
   "tool.aria": "Recognition tool",
   "tool.sameTab": "Same person?",
   "tool.clip": "Clip",
diff --git a/addons/voice/ui/src/index.jsx b/addons/voice/ui/src/index.jsx
index 5dea36844..11a141668 100644
--- a/addons/voice/ui/src/index.jsx
+++ b/addons/voice/ui/src/index.jsx
@@ -3,6 +3,7 @@ import { createSpeech } from './Speech.jsx'
 import { createApi } from './api.js'
 import { createVoiceLibrary } from './VoiceLibrary.jsx'
 import { createVoiceProfileCreate } from './VoiceProfileCreate.jsx'
+import { createVoiceHub } from './VoiceHub.jsx'
 import { createVoiceRecognition } from './VoiceRecognition.jsx'
 import * as identity from '../../../internal/identity-ui/src/identity.js'
 
@@ -14,23 +15,40 @@ export function createAddon(sdk, options = {}) {
   // The library and the profile form are for administrators (the server
   // refuses everyone else). A viewer who is not one gets this notice and the
   // page below is never rendered, so it makes no call and shows no profile data.
+  // The library and the profile form are for administrators (the server
+  // refuses everyone else). A viewer who is not one gets this notice and the
+  // page below is never rendered, so it makes no call and shows no profile data.
   const adminOnly = (Page) => function AdminPage(props) {
     if (!ctx.viewer.isAdmin()) {
       return (
         <main className="page vc-off" data-testid="voice-admin-only">
           <ctx.ui.PageHeader title={ctx.t('voice.adminOnly.title')} supporting={ctx.t('voice.adminOnly.text')} />
+          {props.tabs}
         </main>
       )
     }
     return <Page {...props} />
   }
-  const pages = {
+  const sections = {
     recognition: createVoiceRecognition(ctx, api, identity).VoiceRecognition,
     library: adminOnly(createVoiceLibrary(ctx, api).VoiceLibrary),
     create: adminOnly(createVoiceProfileCreate(ctx, api).VoiceProfileCreate),
     speech: createSpeech(ctx).Speech,
     ...options.pages,
   }
+  const { Hub, redirectFor } = createVoiceHub(ctx, api, identity, sections)
+  // One page, "voice", holds every section as a tab. The older page names stay
+  // declared in the manifest so a saved link still opens, and each one is
+  // sent to its tab (see redirectFor).
+  const pages = { voice: Hub }
+  for (const legacy of ['recognition', 'library', 'create', 'speech']) {
+    pages[legacy] = function LegacyPage({ path }) {
+      const { useEffect } = ctx.React
+      const target = redirectFor(legacy, path || '', sdk.navigation.current().search)
+      useEffect(() => { sdk.navigation.navigate(target, { replace: true }) }, [target])
+      return null
+    }
+  }
 
   function Component({ page, path }) {
     const Page = pages[page]
diff --git a/addons/voice/ui/src/style.css b/addons/voice/ui/src/style.css
index 4220a7e85..4c0929ee5 100644
--- a/addons/voice/ui/src/style.css
+++ b/addons/voice/ui/src/style.css
@@ -262,3 +262,7 @@
 .idn-sheet__need { flex: 1 1 160px; min-width: 0; margin-right: 0; }
 .idn-again { align-self: flex-start; margin-top: var(--dk-space-2); }
 .idn-sheet__foot { display: flex; align-items: center; justify-content: flex-end; gap: var(--dk-space-3); flex-wrap: wrap; padding-top: var(--dk-space-3); border-top: 1px solid var(--dk-edge); }
+
+/* The tab strip of the Voice page. The recognition page is a flex column with
+   its own gap; the other pages are block layouts and need the space here. */
+.vc-speech > .vc-tabs, .voice-library-page > .vc-tabs, .voice-create-page > .vc-tabs, .vc-off > .vc-tabs { margin: var(--dk-space-4) 0 var(--dk-space-5); }