diff --git a/apps/server/src/routes/openai/v1/index.ts b/apps/server/src/routes/openai/v1/index.ts
index 22d23b607..d67f0eb15 100644
--- a/apps/server/src/routes/openai/v1/index.ts
+++ b/apps/server/src/routes/openai/v1/index.ts
@@ -647,11 +647,16 @@ export function createV1Routes(
    * configured yet so the client can render "no voices" instead of
    * exploding.
    */
-  async function handleListStreamingVoices(_c: Context<HonoEnv>) {
+  async function handleListStreamingVoices(c: Context<HonoEnv>) {
     const upstream = await configKV.getOptional('STREAMING_TTS_UPSTREAM')
     if (!upstream || !upstream.baseURL)
       return Response.json({ voices: [], recommended: {} })
 
+    // Pass through the api_resource_id (e.g. `seed-tts-2.0`). unspeech
+    // filters the embedded Volcengine catalogue server-side; absent model
+    // means "return everything streaming-safe".
+    const model = c.req.query('model')
+
     let voicesURL: string
     try {
       const u = new URL(upstream.baseURL)
@@ -659,7 +664,10 @@ export function createV1Routes(
       // stream and the REST voices endpoint on the same listener.
       u.protocol = u.protocol === 'wss:' ? 'https:' : 'http:'
       u.pathname = '/api/voices'
-      u.search = '?provider=volcengine'
+      const params = new URLSearchParams({ provider: 'volcengine' })
+      if (model)
+        params.set('model', model)
+      u.search = `?${params.toString()}`
       voicesURL = u.toString()
     }
     catch (err) {
diff --git a/packages/stage-pages/src/pages/settings/providers/speech/official-provider-speech-streaming.vue b/packages/stage-pages/src/pages/settings/providers/speech/official-provider-speech-streaming.vue
index c6f23c0e1..cc2a8c283 100644
--- a/packages/stage-pages/src/pages/settings/providers/speech/official-provider-speech-streaming.vue
+++ b/packages/stage-pages/src/pages/settings/providers/speech/official-provider-speech-streaming.vue
@@ -48,7 +48,7 @@ const voicesLoading = ref(false)
 async function loadVoices() {
   voicesLoading.value = true
   try {
-    await speechStore.loadVoicesForProvider(providerId)
+    await speechStore.loadVoicesForProvider(providerId, model.value)
   }
   finally {
     voicesLoading.value = false
@@ -62,11 +62,9 @@ onMounted(async () => {
   await loadVoices()
 })
 
-// Reload voices when the model variant changes. The streaming provider
-// shares one voice catalogue across model variants (both Seed-TTS 2.0 and
-// 1.0 expose the same `zh_female_*` ids), so this is mostly a refresh —
-// but it keeps the flow consistent with provider pages where the variant
-// actually does change the catalogue.
+// Volcengine TTS 1.0 and 2.0 ship different voice catalogues (mars/moon/ICL
+// vs uranus/saturn; see unspeech voices.go). Re-fetch on model change so the
+// list switches accordingly.
 watch(model, async () => {
   await loadVoices()
 })
diff --git a/packages/stage-ui/src/components/scenarios/providers/speech-playground-openai-compatible.vue b/packages/stage-ui/src/components/scenarios/providers/speech-playground-openai-compatible.vue
index e5c6c1aa3..8993fd512 100644
--- a/packages/stage-ui/src/components/scenarios/providers/speech-playground-openai-compatible.vue
+++ b/packages/stage-ui/src/components/scenarios/providers/speech-playground-openai-compatible.vue
@@ -178,11 +178,6 @@ defineExpose({
         {{ errorMessage }}
       </div>
       <audio v-if="audioUrl" ref="audioPlayer" :src="audioUrl" controls class="mt-2 w-full" />
-      <SpeechStreamingPlayground
-        :text="testText"
-        :voice="voice"
-        :generate-speech="generateSpeech"
-      />
     </div>
     <!-- Slot for additional provider-specific UI in the playground -->
     <slot />
diff --git a/packages/stage-ui/src/components/scenarios/providers/speech-playground.vue b/packages/stage-ui/src/components/scenarios/providers/speech-playground.vue
index 4617b907a..dcb39b0c5 100644
--- a/packages/stage-ui/src/components/scenarios/providers/speech-playground.vue
+++ b/packages/stage-ui/src/components/scenarios/providers/speech-playground.vue
@@ -5,8 +5,6 @@ import { FieldCheckbox, FieldCombobox } from '@proj-airi/ui'
 import { computed, onUnmounted, ref, watch } from 'vue'
 import { useI18n } from 'vue-i18n'
 
-import SpeechStreamingPlayground from './speech-streaming-playground.vue'
-
 import { TestDummyMarker } from '../../gadgets'
 
 const props = defineProps<{
@@ -195,12 +193,6 @@ defineExpose({
         {{ errorMessage }}
       </div>
       <audio v-if="audioUrl" ref="audioPlayer" :src="audioUrl" controls class="mt-2 w-full" />
-
-      <SpeechStreamingPlayground
-        :text="testText"
-        :voice="selectedVoice"
-        :generate-speech="generateSpeech"
-      />
     </div>
     <!-- Slot for additional provider-specific UI in the playground -->
     <slot />
diff --git a/packages/stage-ui/src/components/scenarios/providers/speech-streaming-playground.vue b/packages/stage-ui/src/components/scenarios/providers/speech-streaming-playground.vue
deleted file mode 100644
index e2b0b89b9..000000000
--- a/packages/stage-ui/src/components/scenarios/providers/speech-streaming-playground.vue
+++ /dev/null
@@ -1,165 +0,0 @@
-<script setup lang="ts">
-import type { TTSInputChunk } from '../../../utils/tts'
-
-import { createQueue } from '@proj-airi/stream-kit'
-import { animate } from 'animejs'
-import { ref } from 'vue'
-
-import { useAudioContext } from '../../../stores/audio'
-import { chunkTTSInput } from '../../../utils/tts'
-
-const props = defineProps<{
-  text: string
-  // Provider-specific handlers (provided from parent)
-  generateSpeech: (input: string, voice: string, useSSML: boolean) => Promise<ArrayBuffer>
-  voice: string
-}>()
-
-const { audioContext } = useAudioContext()
-const nowSpeaking = ref(false)
-const ttsInputChunks = ref<TTSInputChunk[]>([])
-const speechGenerationIndex = ref(-1)
-
-const audioQueue = createQueue<{ audioBuffer: AudioBuffer, text: string }>({
-  handlers: [
-    (ctx) => {
-      return new Promise((resolve) => {
-        const source = audioContext.createBufferSource()
-        source.buffer = ctx.data.audioBuffer
-        source.connect(audioContext.destination)
-
-        nowSpeaking.value = true
-        source.start(0)
-        source.onended = () => {
-          nowSpeaking.value = false
-          resolve()
-        }
-      })
-    },
-  ],
-})
-
-async function handleSpeechGeneration(ctx: { data: string }) {
-  speechGenerationIndex.value++
-
-  try {
-    const input = ctx.data
-
-    const res = await props.generateSpeech(input, props.voice, false)
-
-    const audioBuffer = await audioContext.decodeAudioData(res)
-    audioQueue.enqueue({ audioBuffer, text: ctx.data })
-  }
-  catch (error) {
-    console.error('Speech generation failed:', error)
-  }
-}
-
-const ttsQueue = createQueue<string>({ handlers: [handleSpeechGeneration] })
-
-async function testStreaming() {
-  speechGenerationIndex.value = -1
-  for await (const chunk of chunkTTSInput(props.text, { boost: 1, minimumWords: 4, maximumWords: 12 })) {
-    if (!chunk.text)
-      continue
-    ttsQueue.enqueue(chunk.text)
-  }
-}
-
-async function testChunking() {
-  const chunks: TTSInputChunk[] = []
-  const stream = new ReadableStream<Uint8Array>({
-    start(controller) {
-      controller.enqueue(new TextEncoder().encode(props.text))
-      controller.close()
-    },
-  })
-
-  for await (const chunk of chunkTTSInput(stream.getReader(), { boost: 1, minimumWords: 4, maximumWords: 12 })) {
-    chunks.push(chunk)
-  }
-
-  ttsInputChunks.value = chunks
-}
-</script>
-
-<template>
-  <div class="flex items-center gap-1 text-sm font-medium">
-    Streaming Playground
-  </div>
-  <div flex="~ row" gap-4>
-    <button
-      border="neutral-800 dark:neutral-200 solid 2" transition="border duration-250 ease-in-out"
-      rounded-lg px-4 text="neutral-100 dark:neutral-900" py-2 text-sm
-      bg="neutral-700 dark:neutral-300" @click="testChunking"
-    >
-      <div flex="~ row" items-center gap-2>
-        <div i-solar:round-double-alt-arrow-right-bold-duotone />
-        <span>Test chunking</span>
-      </div>
-    </button>
-
-    <button
-      v-if="ttsInputChunks.length > 0"
-      border="neutral-800 dark:neutral-200 solid 2" transition="border duration-250 ease-in-out"
-      rounded-lg px-4 text="neutral-100 dark:neutral-900" py-2 text-sm
-      bg="neutral-700 dark:neutral-300" @click="testStreaming"
-    >
-      <div flex="~ row" items-center gap-2>
-        <div i-solar:round-double-alt-arrow-right-bold-duotone />
-        <span>Test streaming</span>
-      </div>
-    </button>
-  </div>
-
-  <div flex="~ col gap-2 items-start" py-4>
-    <div
-      v-for="(chunk, i) in ttsInputChunks"
-      :key="i"
-      flex="~ row gap-2 items-center"
-    >
-      <div
-        flex="~ row gap-2 items-center"
-        rounded-xl px-2 py-1.5
-        :class="{
-          'bg-neutral-100 dark:bg-neutral-800': speechGenerationIndex < i,
-          'bg-neutral-200 dark:bg-neutral-700': speechGenerationIndex >= i,
-        }"
-      >
-        <span ml-1>{{ chunk.text }}</span>
-        <span
-          rounded-full px-2 py-.5 text-nowrap text-xs
-          b="~ dashed"
-          :class="{
-            'b-green text-green': chunk.reason === 'boost',
-            'b-orange text-orange': chunk.reason === 'limit',
-            'b-red text-red': chunk.reason === 'hard',
-            'b-purple text-purple': chunk.reason === 'flush',
-          }"
-        >
-          {{ chunk.words }} words,
-          {{ chunk.reason }}
-        </span>
-      </div>
-      <Transition
-        :css="false"
-        @enter="(el) => animate(el, {
-          opacity: [0, 1],
-          translateX: [10, 0],
-          duration: 200,
-          ease: 'inOut',
-        })"
-      >
-        <div
-          v-if="speechGenerationIndex >= i"
-          tag="div"
-          flex="~ row items-center gap-1"
-          text-sm
-        >
-          <div i-solar-check-circle-line-duotone />
-          <div>Queued</div>
-        </div>
-      </Transition>
-    </div>
-  </div>
-</template>
diff --git a/packages/stage-ui/src/libs/providers/providers/official/index.ts b/packages/stage-ui/src/libs/providers/providers/official/index.ts
index 9f6d04cd2..b3a78096a 100644
--- a/packages/stage-ui/src/libs/providers/providers/official/index.ts
+++ b/packages/stage-ui/src/libs/providers/providers/official/index.ts
@@ -223,15 +223,23 @@ export const providerOfficialSpeechStreaming = defineProvider({
         },
       ]
     },
-    listVoices: async (): Promise<VoiceInfo[]> => {
+    listVoices: async (_config, _provider, model): Promise<VoiceInfo[]> => {
       // Streaming voices live behind a dedicated endpoint
       // (`/audio/voices/streaming`) because they come from a separate
       // configKV entry (`STREAMING_TTS_UPSTREAM`) than the HTTP TTS
       // `?model=...` lookup. The server proxies to unspeech's
       // `/api/voices?provider=volcengine`, which ships an embed-time
       // catalogue without requiring credentials.
+      //
+      // `model` here is the unspeech-routed id (e.g. `volcengine/seed-tts-2.0`).
+      // unspeech expects the bare `api_resource_id` for its filter, so we
+      // strip the backend prefix before forwarding.
+      const apiResourceId = model?.includes('/') ? model.split('/', 2)[1] : model
+      const voicesURL = new URL(`${SERVER_URL}/api/v1/audio/voices/streaming`)
+      if (apiResourceId)
+        voicesURL.searchParams.set('model', apiResourceId)
       const res = await globalThis.fetch(
-        `${SERVER_URL}/api/v1/audio/voices/streaming`,
+        voicesURL.toString(),
         { headers: authHeaders() },
       )
       if (!res.ok)
diff --git a/packages/stage-ui/src/libs/providers/types.ts b/packages/stage-ui/src/libs/providers/types.ts
index 59b67411a..84e4fd433 100644
--- a/packages/stage-ui/src/libs/providers/types.ts
+++ b/packages/stage-ui/src/libs/providers/types.ts
@@ -47,7 +47,12 @@ export interface ProviderOnboardingField {
 
 export interface ProviderExtraMethods<TConfig> {
   listModels?: (config: TConfig, provider: ProviderInstance) => Promise<ModelInfo[]>
-  listVoices?: (config: TConfig, provider: ProviderInstance) => Promise<VoiceInfo[]>
+  /**
+   * Returns the voice catalogue. `model` lets providers whose voices vary by
+   * model variant (Volcengine streaming TTS 1.0 vs 2.0 differ in catalogue)
+   * narrow the result. Providers with a single catalogue ignore it.
+   */
+  listVoices?: (config: TConfig, provider: ProviderInstance, model?: string) => Promise<VoiceInfo[]>
   loadModel?: (config: TConfig, provider: ProviderInstance, hooks?: { onProgress?: (progress: ProgressInfo) => Promise<void> | void }) => Promise<void>
 }
 
diff --git a/packages/stage-ui/src/stores/modules/speech.ts b/packages/stage-ui/src/stores/modules/speech.ts
index 7c09c5c03..e65538211 100644
--- a/packages/stage-ui/src/stores/modules/speech.ts
+++ b/packages/stage-ui/src/stores/modules/speech.ts
@@ -83,7 +83,7 @@ export const useSpeechStore = defineStore('speech', () => {
     return ['elevenlabs', 'microsoft-speech', 'azure-speech'].includes(activeSpeechProvider.value)
   })
 
-  async function loadVoicesForProvider(provider: string) {
+  async function loadVoicesForProvider(provider: string, model?: string) {
     if (!provider) {
       return []
     }
@@ -92,7 +92,7 @@ export const useSpeechStore = defineStore('speech', () => {
     speechProviderError.value = null
 
     try {
-      const voices = await providersStore.getProviderMetadata(provider).capabilities.listVoices?.(providersStore.getProviderConfig(provider)) || []
+      const voices = await providersStore.getProviderMetadata(provider).capabilities.listVoices?.(providersStore.getProviderConfig(provider), model) || []
       // Reassign to trigger reactivity when adding/updating provider entries
       availableVoices.value = {
         ...availableVoices.value,
diff --git a/packages/stage-ui/src/stores/providers.ts b/packages/stage-ui/src/stores/providers.ts
index a948beff6..157cdfbca 100644
--- a/packages/stage-ui/src/stores/providers.ts
+++ b/packages/stage-ui/src/stores/providers.ts
@@ -137,7 +137,7 @@ export interface ProviderMetadata {
     | Promise<TranscriptionProviderWithExtraOptions>
   capabilities: {
     listModels?: (config: Record<string, unknown>) => Promise<ModelInfo[]>
-    listVoices?: (config: Record<string, unknown>) => Promise<VoiceInfo[]>
+    listVoices?: (config: Record<string, unknown>, model?: string) => Promise<VoiceInfo[]>
     loadModel?: (config: Record<string, unknown>, hooks?: { onProgress?: (progress: ProgressInfo) => Promise<void> | void }) => Promise<void>
   }
   validators: {
diff --git a/packages/stage-ui/src/stores/providers/converters.ts b/packages/stage-ui/src/stores/providers/converters.ts
index fb0fd18d6..e8ba8bc1b 100644
--- a/packages/stage-ui/src/stores/providers/converters.ts
+++ b/packages/stage-ui/src/stores/providers/converters.ts
@@ -161,10 +161,10 @@ export function convertProviderDefinitionToMetadata(
           }
         },
       listVoices: definition.extraMethods?.listVoices
-        ? async (config) => {
+        ? async (config, model) => {
           const provider = await definition.createProvider(config as any)
           try {
-            return await definition.extraMethods!.listVoices!(config as any, provider)
+            return await definition.extraMethods!.listVoices!(config as any, provider, model)
           }
           finally {
             await (provider as { dispose?: () => Promise<void> | void }).dispose?.()