feat: add upstreamVoiceId to voice packs and related services

- Introduced upstreamVoiceId field in voice pack schema and database.
- Updated voice pack service to handle upstreamVoiceId in CRUD operations.
- Modified API routes and tests to accommodate upstreamVoiceId.
- Enhanced UI components to include upstreamVoiceId in forms and displays.
- Adjusted speech processing logic to utilize upstreamVoiceId where applicable.
- Updated related tests to ensure proper functionality with new field.

apps/server/drizzle/0016_tired_dagger.sql

@@ -0,0 +1,3 @@
+ALTER TABLE "voice_packs" ADD COLUMN "upstream_voice_id" text;
+UPDATE "voice_packs" SET "upstream_voice_id" = "voice_id";
+ALTER TABLE "voice_packs" ALTER COLUMN "upstream_voice_id" SET NOT NULL;

apps/server/drizzle/meta/_journal.json

@@ -113,6 +113,13 @@
       "when": 1780498188308,
       "tag": "0015_concerned_piledriver",
       "breakpoints": true
+    },
+    {
+      "idx": 16,
+      "version": "7",
+      "when": 1782847276369,
+      "tag": "0016_tired_dagger",
+      "breakpoints": true
     }
   ]
-}
+}

apps/server/src/routes/admin/voice-packs/route.test.ts

@@ -25,6 +25,7 @@ function createService() {
     provider: 'volcengine',
     model: 'seed-tts-2.0',
     voiceId: 'voice-neuro',
+    upstreamVoiceId: 'voice-neuro-upstream',
     ttsModelId: 'volcengine/neuro-pool',
     params: {},
     costMultiplier: 1.5,
@@ -119,6 +120,7 @@ describe('admin voice packs — CRUD', () => {
       provider: 'volcengine',
       model: 'seed-tts-2.0',
       voiceId: 'voice-neuro',
+      upstreamVoiceId: 'voice-neuro-upstream',
       ttsModelId: 'volcengine/neuro-pool',
       params: { pitch: '+20%' },
       costMultiplier: 1.5,
@@ -149,6 +151,7 @@ describe('admin voice packs — CRUD', () => {
       provider: 'volcengine',
       model: 'seed-tts-2.0',
       voiceId: 'voice-neuro',
+      upstreamVoiceId: 'voice-neuro-upstream',
       ttsModelId: 'volcengine/neuro-pool',
       params: {},
       costMultiplier: -1,

apps/server/src/routes/openai/v1/route.test.ts

@@ -710,7 +710,8 @@ describe('v1CompletionsRoutes', () => {
             description: null,
             provider: 'azure',
             model: 'microsoft/v1',
-            voiceId: 'en-US-AvaMultilingualNeural',
+            voiceId: 'friendly-azure',
+            upstreamVoiceId: 'en-US-AvaMultilingualNeural',
             ttsModelId: 'microsoft/v1',
             params: {},
             costMultiplier: 1.5,
@@ -728,12 +729,11 @@ describe('v1CompletionsRoutes', () => {
           body: JSON.stringify({
             model: 'auto',
             input: 'test',
-            voice: 'en-US-AvaMultilingualNeural',
+            voice: 'friendly-azure',
             speed: 1.2,
             extra_body: {
               voice_pack: {
                 pack_id: 'vp-azure',
-                cost_multiplier: 1.5,
                 pitch: 20,
                 volume: 5,
               },
@@ -784,7 +784,7 @@ describe('v1CompletionsRoutes', () => {
 
     /**
      * @example
-     * POST /api/v1/audio/speech { "input": "hello", "extra_body": { "voice_pack": { "cost_multiplier": 2 } } }
+     * POST /api/v1/audio/speech { "input": "hello", "extra_body": { "voice_pack": { "pack_id": "vp-premium" } } }
      */
     it('uses Voice Pack cost multiplier for affordability and billing units', async () => {
       globalThis.fetch = vi.fn(async () => new Response(new Uint8Array([1]), {
@@ -801,6 +801,7 @@ describe('v1CompletionsRoutes', () => {
           provider: 'azure',
           model: 'microsoft/v1',
           voiceId: 'alloy',
+          upstreamVoiceId: 'upstream-alloy',
           ttsModelId: 'tts-1',
           params: {},
           costMultiplier: 2,
@@ -832,7 +833,6 @@ describe('v1CompletionsRoutes', () => {
             extra_body: {
               voice_pack: {
                 pack_id: 'vp-premium',
-                cost_multiplier: 2,
               },
             },
           }),
@@ -868,6 +868,7 @@ describe('v1CompletionsRoutes', () => {
           provider: 'azure',
           model: 'microsoft/v1',
           voiceId: 'alloy',
+          upstreamVoiceId: 'upstream-alloy',
           ttsModelId: 'tts-1',
           params: {},
           costMultiplier: 2,
@@ -899,7 +900,6 @@ describe('v1CompletionsRoutes', () => {
             extra_body: {
               voice_pack: {
                 pack_id: 'vp-premium',
-                cost_multiplier: 2,
               },
               airi_analytics: {
                 source: 'manual_preview',

apps/server/src/routes/voice-packs/index.ts

@@ -5,6 +5,20 @@ import { Hono } from 'hono'
 
 import { authGuard } from '../../middlewares/auth'
 
+function publicVoicePack(pack: Awaited<ReturnType<VoicePackService['listEnabled']>>[number]) {
+  return {
+    id: pack.id,
+    name: pack.name,
+    description: pack.description,
+    voiceId: pack.voiceId,
+    params: pack.params,
+    costMultiplier: pack.costMultiplier,
+    enabled: pack.enabled,
+    createdAt: pack.createdAt,
+    updatedAt: pack.updatedAt,
+  }
+}
+
 /**
  * User-facing Voice Pack routes.
  *
@@ -17,6 +31,6 @@ export function createVoicePackRoutes(service: VoicePackService) {
     .use('*', authGuard)
     .get('/', async (c) => {
       const packs = await service.listEnabled()
-      return c.json(packs)
+      return c.json(packs.map(publicVoicePack))
     })
 }

apps/server/src/routes/voice-packs/route.test.ts

@@ -23,7 +23,21 @@ function createTestApp(service: VoicePackService, user: { id: string } | null) {
 
 function createService() {
   return {
-    listEnabled: vi.fn(async () => [{ id: 'vp-1', name: 'Enabled', enabled: true }]),
+    listEnabled: vi.fn(async () => [{
+      id: 'vp-1',
+      name: 'Enabled',
+      description: 'Public description',
+      provider: 'azure',
+      model: 'microsoft/v1',
+      voiceId: 'friendly-voice',
+      upstreamVoiceId: 'en-US-AvaMultilingualNeural',
+      ttsModelId: 'microsoft/v1',
+      params: { pitch: '+10%' },
+      costMultiplier: 2,
+      enabled: true,
+      createdAt: new Date('2026-01-01T00:00:00.000Z'),
+      updatedAt: new Date('2026-01-02T00:00:00.000Z'),
+    }]),
     list: vi.fn(),
     create: vi.fn(),
     update: vi.fn(),
@@ -50,7 +64,17 @@ describe('voice packs routes', () => {
     const res = await app.request('/api/v1/voice-packs')
 
     expect(res.status).toBe(200)
-    expect(await res.json()).toEqual([{ id: 'vp-1', name: 'Enabled', enabled: true }])
+    expect(await res.json()).toEqual([{
+      id: 'vp-1',
+      name: 'Enabled',
+      description: 'Public description',
+      voiceId: 'friendly-voice',
+      params: { pitch: '+10%' },
+      costMultiplier: 2,
+      enabled: true,
+      createdAt: '2026-01-01T00:00:00.000Z',
+      updatedAt: '2026-01-02T00:00:00.000Z',
+    }])
     expect(service.listEnabled).toHaveBeenCalled()
   })
 })

apps/server/src/schemas/voice-packs.ts

@@ -16,6 +16,7 @@ export const voicePacks = pgTable(
     provider: text('provider').notNull(),
     model: text('model').notNull(),
     voiceId: text('voice_id').notNull(),
+    upstreamVoiceId: text('upstream_voice_id').notNull(),
     ttsModelId: text('tts_model_id').notNull(),
     params: jsonb('params').notNull().$type<VoicePackParams>().default({}),
     costMultiplier: real('cost_multiplier').notNull().default(1),

apps/server/src/services/domain/openai-speech/index.ts

@@ -90,20 +90,23 @@ export function createOpenAiSpeechService(deps: OpenAiSpeechServiceDeps) {
 
   async function handleSpeechRequest(input: OpenAiSpeechRequest): Promise<Response> {
     const requestId = nanoid()
-    let requestModel = typeof input.body.model === 'string' ? input.body.model : 'auto'
+    const requestedModel = typeof input.body.model === 'string' ? input.body.model : 'auto'
+    let requestModel = requestedModel
+    const requestVoice = typeof input.body.voice === 'string' ? input.body.voice : undefined
     const inputText = typeof input.body.input === 'string' ? input.body.input : ''
     const analytics = ttsAnalyticsContext(input.body)
 
-    if (requestModel === 'auto')
-      requestModel = await deps.configKV.getOrThrow('DEFAULT_TTS_MODEL')
-
     const voicePackRequest = await voicePackRequestOptions(input.body, {
-      model: requestModel,
+      requestedModel,
-      voice: typeof input.body.voice === 'string' ? input.body.voice : undefined,
+      voice: requestVoice,
       voicePackService: deps.voicePackService,
     })
+    requestModel = voicePackRequest.model ?? requestModel
+    if (requestModel === 'auto')
+      requestModel = await deps.configKV.getOrThrow('DEFAULT_TTS_MODEL')
+    const routedVoice = voicePackRequest.voice ?? requestVoice
     const voiceMetadata = ttsVoiceMetadata({
-      voice: typeof input.body.voice === 'string' ? input.body.voice : undefined,
+      voice: requestVoice,
       voicePackId: voicePackRequest.voicePackId,
       voiceType: analytics.voiceType,
     })
@@ -114,7 +117,7 @@ export function createOpenAiSpeechService(deps: OpenAiSpeechServiceDeps) {
       userId: input.userId,
       model: requestModel,
       inputChars: inputText.length,
-      voice: typeof input.body.voice === 'string' ? input.body.voice : undefined,
+      voice: requestVoice,
     }).log('tts speech request')
 
     void deps.productEventService.track({
@@ -171,7 +174,7 @@ export function createOpenAiSpeechService(deps: OpenAiSpeechServiceDeps) {
 
     const ttsInput = {
       text: inputText,
-      voice: typeof input.body.voice === 'string' ? input.body.voice : undefined,
+      voice: routedVoice,
       speed: typeof input.body.speed === 'number' ? input.body.speed : undefined,
       responseFormat: typeof input.body.response_format === 'string' ? input.body.response_format : undefined,
       extraOptions: voicePackRequest.extraOptions,
@@ -405,19 +408,22 @@ function ttsVoiceMetadata(input: {
 async function voicePackRequestOptions(
   body: Record<string, unknown>,
   context: {
-    model: string
+    requestedModel: string
     voice?: string
     voicePackService: VoicePackService
   },
-): Promise<{ extraOptions: Record<string, unknown> | undefined, costMultiplier: number, voicePackId?: string }> {
+): Promise<{
+  extraOptions: Record<string, unknown> | undefined
+  costMultiplier: number
+  voicePackId?: string
+  model?: string
+  voice?: string
+}> {
   const extraBody = asRecord(body.extra_body)
   const voicePackOptions = asRecord(extraBody?.voice_pack)
   const pitch = readOptionalNumber(voicePackOptions, 'pitch')
   const volume = readOptionalNumber(voicePackOptions, 'volume')
-  const costMultiplier = await resolveVoicePackCostMultiplier(voicePackOptions, context)
+  const voicePack = await resolveVoicePackRequest(voicePackOptions, context)
-  const voicePackId = typeof voicePackOptions?.pack_id === 'string' && voicePackOptions.pack_id.trim()
-    ? voicePackOptions.pack_id
-    : undefined
   const extraOptions: Record<string, unknown> = {}
   if (pitch != null)
     extraOptions.pitch = pitch
@@ -426,40 +432,49 @@ async function voicePackRequestOptions(
 
   return {
     extraOptions: Object.keys(extraOptions).length > 0 ? extraOptions : undefined,
-    costMultiplier,
+    costMultiplier: voicePack?.costMultiplier ?? 1,
-    voicePackId,
+    voicePackId: voicePack?.id,
+    model: voicePack?.ttsModelId,
+    voice: voicePack?.upstreamVoiceId,
   }
 }
 
-async function resolveVoicePackCostMultiplier(
+async function resolveVoicePackRequest(
   voicePackOptions: Record<string, unknown> | undefined,
   context: {
-    model: string
+    requestedModel: string
     voice?: string
     voicePackService: VoicePackService
   },
-): Promise<number> {
+): Promise<Awaited<ReturnType<VoicePackService['findById']>> | null> {
   const packId = voicePackOptions?.pack_id
-  const value = voicePackOptions?.cost_multiplier
+  if (voicePackOptions?.cost_multiplier != null) {
-  if (packId == null && value == null)
+    throw createBadRequestError('voice_pack.cost_multiplier is server-managed', 'INVALID_VOICE_PACK', {
-    return 1
+      field: 'voice_pack.cost_multiplier',
+    })
+  }
+  if (packId == null)
+    return null
   if (typeof packId !== 'string' || !packId.trim())
     throw createBadRequestError('voice_pack.pack_id is required when Voice Pack billing metadata is provided', 'INVALID_VOICE_PACK')
 
   const pack = await context.voicePackService.findById(packId)
   if (!pack)
     throw createBadRequestError('Voice Pack not found', 'INVALID_VOICE_PACK', { packId })
-  if (pack.ttsModelId !== context.model || pack.voiceId !== context.voice) {
+  if (context.requestedModel !== 'auto' && pack.ttsModelId !== context.requestedModel) {
+    throw createBadRequestError('Voice Pack does not match requested model and voice', 'INVALID_VOICE_PACK', {
+      packId,
+      actualModel: context.requestedModel,
+    })
+  }
+  if (pack.voiceId !== context.voice) {
     throw createBadRequestError('Voice Pack does not match requested model and voice', 'INVALID_VOICE_PACK', {
       packId,
-      expectedModel: pack.ttsModelId,
-      actualModel: context.model,
-      expectedVoice: pack.voiceId,
       actualVoice: context.voice,
     })
   }
 
-  return pack.costMultiplier
+  return pack
 }
 
 function routerFailure(error: unknown): { status: number, reason: string, message: string } {

apps/server/src/services/domain/voice-packs/index.test.ts

@@ -27,6 +27,7 @@ describe('voicePackService', () => {
       provider: 'volcengine',
       model: 'seed-tts-2.0',
       voiceId: 'voice-neuro',
+      upstreamVoiceId: 'voice-neuro-upstream',
       ttsModelId: 'volcengine/neuro-pool',
       params: { pitch: '+20%', volume: '+5%' },
       costMultiplier: 1.5,
@@ -37,6 +38,7 @@ describe('voicePackService', () => {
     expect(pack.provider).toBe('volcengine')
     expect(pack.model).toBe('seed-tts-2.0')
     expect(pack.voiceId).toBe('voice-neuro')
+    expect(pack.upstreamVoiceId).toBe('voice-neuro-upstream')
     expect(pack.ttsModelId).toBe('volcengine/neuro-pool')
     expect(pack.params).toEqual({ pitch: '+20%', volume: '+5%' })
     expect(pack.costMultiplier).toBe(1.5)
@@ -50,6 +52,7 @@ describe('voicePackService', () => {
       provider: 'volcengine',
       model: 'seed-tts-2.0',
       voiceId: 'voice-a',
+      upstreamVoiceId: 'voice-a-upstream',
       ttsModelId: 'volcengine/pool',
       params: {},
       costMultiplier: 1,
@@ -60,6 +63,7 @@ describe('voicePackService', () => {
       provider: 'volcengine',
       model: 'seed-tts-2.0',
       voiceId: 'voice-a',
+      upstreamVoiceId: 'voice-a-upstream',
       ttsModelId: 'volcengine/pool',
       params: { pitch: '+20%' },
       costMultiplier: 1,
@@ -78,6 +82,7 @@ describe('voicePackService', () => {
       provider: 'azure',
       model: 'v1',
       voiceId: 'en-US-AvaMultilingualNeural',
+      upstreamVoiceId: 'en-US-AvaMultilingualNeural',
       ttsModelId: 'microsoft/v1',
       params: {},
       costMultiplier: 1,
@@ -103,6 +108,7 @@ describe('voicePackService', () => {
       provider: 'dashscope-cosyvoice',
       model: 'cosyvoice-v2',
       voiceId: 'longxiaochun_v2',
+      upstreamVoiceId: 'longxiaochun_v2',
       ttsModelId: 'alibaba/cosyvoice-v2',
       params: {},
       costMultiplier: 1,

apps/server/src/services/domain/voice-packs/index.ts

@@ -24,6 +24,7 @@ export const CreateVoicePackInputSchema = object({
   provider: pipe(string(), nonEmpty('provider is required'), maxLength(100)),
   model: pipe(string(), nonEmpty('model is required'), maxLength(200)),
   voiceId: pipe(string(), nonEmpty('voiceId is required'), maxLength(200)),
+  upstreamVoiceId: pipe(string(), nonEmpty('upstreamVoiceId is required'), maxLength(200)),
   ttsModelId: pipe(string(), nonEmpty('ttsModelId is required'), maxLength(200)),
   params: optional(VoicePackParamsSchema, {}),
   costMultiplier: VoicePackCostMultiplierSchema,
@@ -36,6 +37,7 @@ export const UpdateVoicePackInputSchema = object({
   provider: optional(pipe(string(), nonEmpty('provider must not be empty'), maxLength(100))),
   model: optional(pipe(string(), nonEmpty('model must not be empty'), maxLength(200))),
   voiceId: optional(pipe(string(), nonEmpty('voiceId must not be empty'), maxLength(200))),
+  upstreamVoiceId: optional(pipe(string(), nonEmpty('upstreamVoiceId must not be empty'), maxLength(200))),
   ttsModelId: optional(pipe(string(), nonEmpty('ttsModelId must not be empty'), maxLength(200))),
   params: optional(VoicePackParamsSchema),
   costMultiplier: optional(VoicePackCostMultiplierSchema),
@@ -74,6 +76,7 @@ export function createVoicePackService(db: Database) {
         provider: input.provider,
         model: input.model,
         voiceId: input.voiceId,
+        upstreamVoiceId: input.upstreamVoiceId,
         ttsModelId: input.ttsModelId,
         params: input.params,
         costMultiplier: input.costMultiplier,

apps/ui-admin/src/modules/api.ts

@@ -180,6 +180,7 @@ export interface VoicePack {
   provider: string
   model: string
   voiceId: string
+  upstreamVoiceId: string
   ttsModelId: string
   params: VoicePackParams
   costMultiplier: number
@@ -194,6 +195,7 @@ export interface VoicePackPayload {
   provider: string
   model: string
   voiceId: string
+  upstreamVoiceId: string
   ttsModelId: string
   params?: VoicePackParams
   costMultiplier: number

apps/ui-admin/src/pages/VoicePackFormPage.vue

@@ -39,6 +39,7 @@ const form = reactive({
   provider: '',
   model: '',
   voiceId: '',
+  upstreamVoiceId: '',
   ttsModelId: '',
   paramsJson: DEFAULT_PARAMS,
   costMultiplier: 1,
@@ -113,7 +114,9 @@ const formError = computed(() => {
   if (!form.ttsModelId.trim())
     return 'TTS model ID is required'
   if (!form.voiceId.trim())
-    return 'Voice ID is required'
+    return 'Voice alias is required'
+  if (!form.upstreamVoiceId.trim())
+    return 'Upstream voice ID is required'
   if (!Number.isFinite(Number(form.costMultiplier)) || Number(form.costMultiplier) < 0)
     return 'Cost multiplier must be a non-negative number'
   return null
@@ -125,7 +128,7 @@ onMounted(async () => {
     fillSelectedPack()
   else
     resetForm()
-  await loadVoices(form.ttsModelId, { autoPick: !form.voiceId.trim() })
+  await loadVoices(form.ttsModelId, { autoPick: !form.upstreamVoiceId.trim() })
   modelChangeVoiceLoadingEnabled.value = true
 })
 
@@ -194,8 +197,11 @@ async function loadVoices(model: string, options: { autoPick: boolean }) {
     const result = await adminApi.speechVoices(model.trim())
     voices.value = result.voices
     recommendedVoices.value = result.recommended
-    if (options.autoPick && !form.voiceId.trim())
+    if (options.autoPick && !form.upstreamVoiceId.trim()) {
-      form.voiceId = firstRecommendedVoiceId(result.recommended) ?? result.voices[0]?.id ?? ''
+      form.upstreamVoiceId = firstRecommendedVoiceId(result.recommended) ?? result.voices[0]?.id ?? ''
+      if (!form.voiceId.trim())
+        form.voiceId = form.upstreamVoiceId
+    }
   }
   catch (error) {
     voices.value = []
@@ -223,6 +229,7 @@ function fillForm(pack: VoicePack) {
   form.provider = pack.provider
   form.model = pack.model
   form.voiceId = pack.voiceId
+  form.upstreamVoiceId = pack.upstreamVoiceId
   form.ttsModelId = pack.ttsModelId
   form.paramsJson = JSON.stringify(pack.params ?? {}, null, 2)
   form.costMultiplier = pack.costMultiplier
@@ -238,6 +245,7 @@ function resetForm() {
   form.provider = modelParts.provider
   form.model = modelParts.model
   form.voiceId = ''
+  form.upstreamVoiceId = ''
   form.ttsModelId = modelId
   form.paramsJson = DEFAULT_PARAMS
   form.costMultiplier = 1
@@ -279,6 +287,7 @@ function payload(): VoicePackPayload {
     provider: form.provider.trim(),
     model: form.model.trim(),
     voiceId: form.voiceId.trim(),
+    upstreamVoiceId: form.upstreamVoiceId.trim(),
     ttsModelId: form.ttsModelId.trim(),
     params: parseParams(),
     costMultiplier: Number(form.costMultiplier),
@@ -340,7 +349,7 @@ async function testVoicePack() {
     const body = {
       model: form.ttsModelId.trim(),
       input: text,
-      voice: form.voiceId.trim(),
+      voice: form.upstreamVoiceId.trim(),
       speed: normalizeRateOption(params.rate),
       extra_body: voicePackExtraBody(params),
     }
@@ -494,19 +503,27 @@ function normalizeRateOption(value: string | number | boolean | null | undefined
               :placeholder="ttsModelPlaceholder"
               required
             />
-            <DatalistField
+            <FieldInput
               v-model="form.voiceId"
-              :description="loadingVoices ? 'Loading voices for the selected model...' : 'Voice catalog from /api/v1/audio/voices.'"
+              description="Product-facing Voice Pack voice alias shown to clients and analytics."
               input-class="font-mono text-xs"
-              label="Voice ID"
+              label="Voice alias"
-              list-id="voice-pack-voices"
+              placeholder="narrator-cn"
-              :options="voiceOptions"
-              :placeholder="voicePlaceholder"
               required
             />
           </div>
 
           <div :class="['grid', 'gap-4', 'md:grid-cols-2']">
+            <DatalistField
+              v-model="form.upstreamVoiceId"
+              :description="loadingVoices ? 'Loading voices for the selected model...' : 'Voice catalog from /api/v1/audio/voices.'"
+              input-class="font-mono text-xs"
+              label="Upstream voice ID"
+              list-id="voice-pack-voices"
+              :options="voiceOptions"
+              :placeholder="voicePlaceholder"
+              required
+            />
             <DatalistField
               v-model="form.provider"
               description="Derived from the model ID when possible; editable for custom routing metadata."

apps/ui-admin/src/pages/VoicePacksPage.vue

@@ -83,7 +83,8 @@ function editPack(pack: VoicePack) {
         <tr>
           <th>Name</th>
           <th>Routing</th>
-          <th>Voice</th>
+          <th>Voice Alias</th>
+          <th>Upstream Voice</th>
           <th>Cost</th>
           <th>Status</th>
           <th>Updated</th>
@@ -118,6 +119,9 @@ function editPack(pack: VoicePack) {
           <td :class="['text-xs', 'font-mono']">
             {{ pack.voiceId }}
           </td>
+          <td :class="['text-xs', 'font-mono']">
+            {{ pack.upstreamVoiceId }}
+          </td>
           <td>{{ formatMultiplier(pack.costMultiplier) }}</td>
           <td>
             <span :class="['badge', pack.enabled ? 'badge-green' : 'badge-amber']">

packages/stage-pages/src/pages/settings/modules/speech.vue

@@ -75,6 +75,8 @@ const errorMessage = ref('')
 const selectedSpeechSource = ref<string | null>(null)
 
 const VOICE_PACK_SOURCE_ID = 'voice-pack'
+const VOICE_PACK_REQUEST_MODEL_ID = 'auto'
+const VOICE_PACK_ANALYTICS_MODEL_ID = 'voice_pack'
 const STREAMING_MODEL_OPTION_PREFIX = 'streaming:'
 
 const isOfficialSpeechProvider = computed(() => activeSpeechProvider.value === OFFICIAL_SPEECH_PROVIDER_ID)
@@ -198,6 +200,15 @@ function createVoicePackVoice(voicePack: VoicePackSnapshot): VoiceInfo {
   }
 }
 
+function formatVoicePackCostMultiplier(costMultiplier: number) {
+  return `Flux cost: ${costMultiplier}x`
+}
+
+function voicePackDescription(description: string | null | undefined, costMultiplier: number) {
+  const cost = formatVoicePackCostMultiplier(costMultiplier)
+  return description ? `${description} · ${cost}` : cost
+}
+
 function voicePackVoiceId(packId: string) {
   return `voice-pack:${packId}`
 }
@@ -211,7 +222,7 @@ const displayedVoiceOptions = computed(() => {
     const options = voicePacks.value.map(pack => ({
       id: voicePackVoiceId(pack.id),
       name: pack.name,
-      description: pack.description ?? undefined,
+      description: voicePackDescription(pack.description, pack.costMultiplier),
       previewURL: '',
       customizable: false,
     }))
@@ -221,7 +232,7 @@ const displayedVoiceOptions = computed(() => {
       options.unshift({
         id: frozenVoiceId,
         name: voicePack.name,
-        description: voicePack.name,
+        description: voicePackDescription(voicePack.name, voicePack.costMultiplier),
         previewURL: '',
         customizable: false,
       })
@@ -265,6 +276,12 @@ const currentSpeechVoiceId = computed(() => {
   return activeSpeechVoiceId.value || ''
 })
 
+const currentVoicePackCostMultiplier = computed(() => {
+  if (isVoicePackSourceSelected.value && boundVoicePack.value)
+    return formatVoicePackCostMultiplier(boundVoicePack.value.costMultiplier)
+  return ''
+})
+
 function syncBoundVoicePackSelection() {
   const voicePack = boundVoicePack.value
   if (!voicePack)
@@ -272,7 +289,7 @@ function syncBoundVoicePackSelection() {
 
   selectedSpeechSource.value = VOICE_PACK_SOURCE_ID
   activeSpeechProvider.value = OFFICIAL_SPEECH_PROVIDER_ID
-  activeSpeechModel.value = voicePack.ttsModelId
+  activeSpeechModel.value = VOICE_PACK_REQUEST_MODEL_ID
   activeSpeechVoiceId.value = voicePack.voiceId
   activeSpeechVoice.value = createVoicePackVoice(voicePack)
   return true
@@ -283,7 +300,7 @@ function syncBoundVoicePackSelection() {
  */
 function currentTtsModelId() {
   if (isVoicePackSourceSelected.value && boundVoicePack.value)
-    return boundVoicePack.value.ttsModelId
+    return VOICE_PACK_ANALYTICS_MODEL_ID
   return activeSpeechModel.value || 'unknown'
 }
 
@@ -396,7 +413,7 @@ function selectSpeechSource(sourceId: string) {
     activeSpeechProvider.value = OFFICIAL_SPEECH_PROVIDER_ID
     const voicePack = boundVoicePack.value
     if (voicePack) {
-      activeSpeechModel.value = voicePack.ttsModelId
+      activeSpeechModel.value = VOICE_PACK_REQUEST_MODEL_ID
       activeSpeechVoiceId.value = voicePack.voiceId
       activeSpeechVoice.value = createVoicePackVoice(voicePack)
       return
@@ -481,12 +498,12 @@ async function bindVoicePack(pack: (typeof voicePacks.value)[number]) {
 
   selectedSpeechSource.value = VOICE_PACK_SOURCE_ID
   activeSpeechProvider.value = OFFICIAL_SPEECH_PROVIDER_ID
-  activeSpeechModel.value = pack.ttsModelId
+  activeSpeechModel.value = VOICE_PACK_REQUEST_MODEL_ID
   activeSpeechVoiceId.value = pack.voiceId
   activeSpeechVoice.value = {
     id: pack.voiceId,
     name: pack.name,
-    description: pack.description ?? pack.name,
+    description: voicePackDescription(pack.description ?? pack.name, pack.costMultiplier),
     previewURL: '',
     languages: [{ code: 'en', title: 'English' }],
     provider: activeSpeechProvider.value,
@@ -495,14 +512,14 @@ async function bindVoicePack(pack: (typeof voicePacks.value)[number]) {
 
   trackVoicePackBound({
     tts_provider_id: activeSpeechProvider.value || 'unknown',
-    tts_model_id: pack.ttsModelId,
+    tts_model_id: VOICE_PACK_ANALYTICS_MODEL_ID,
     voice_id: pack.voiceId,
     voice_pack_id: pack.id,
     source: 'settings',
   })
   trackVoiceSelected({
     tts_provider_id: activeSpeechProvider.value || 'unknown',
-    tts_model_id: pack.ttsModelId,
+    tts_model_id: VOICE_PACK_ANALYTICS_MODEL_ID,
     voice_id: pack.voiceId,
     voice_type: 'voice_pack',
     voice_pack_id: pack.id,
@@ -603,7 +620,7 @@ async function generateTestSpeech() {
 
   const voicePack = boundVoicePack.value
   if (voicePack) {
-    model = voicePack.ttsModelId
+    model = VOICE_PACK_REQUEST_MODEL_ID
     if (!voice || voice.id !== voicePack.voiceId)
       voice = createVoicePackVoice(voicePack)
   }
@@ -928,7 +945,10 @@ function handleDeleteProvider(providerId: string) {
             </h2>
             <div class="flex flex-col items-start gap-1 text-neutral-400 md:flex-row md:items-center md:justify-between dark:text-neutral-500">
               <span>Customize how your AI assistant speaks</span>
-              <span v-if="currentSpeechVoiceId" class="text-sm text-neutral-400 font-medium dark:text-neutral-400">Current voice: {{ currentSpeechVoiceId }}</span>
+              <span v-if="currentSpeechVoiceId" class="text-sm text-neutral-400 font-medium dark:text-neutral-400">
+                Current voice: {{ currentSpeechVoiceId }}
+                <span v-if="currentVoicePackCostMultiplier">· {{ currentVoicePackCostMultiplier }}</span>
+              </span>
             </div>
           </div>
 
@@ -964,7 +984,6 @@ function handleDeleteProvider(providerId: string) {
             <VoiceCardManySelect
               v-model:search-query="voiceSearchQuery"
               v-model:voice-id="displayedSpeechVoiceId"
-              :show-visualizer="false"
               :voices="displayedVoiceOptions"
               :searchable="true"
               :search-placeholder="t('settings.pages.modules.speech.sections.section.provider-voice-selection.search_voices_placeholder')"

packages/stage-ui/src/components/scenes/Stage.vue

@@ -436,7 +436,7 @@ const speechPipeline = createSpeechPipeline<AudioBuffer>({
 
     const voicePack = voicePackForSpeechProvider(activeSpeechProvider.value, activeCard.value?.extensions.airi.modules.speech.voicePack)
     if (voicePack) {
-      model = voicePack.ttsModelId
+      model = 'auto'
       if (!voice || voice.id !== voicePack.voiceId)
         voice = createVoicePackVoice(voicePack)
     }

packages/stage-ui/src/stores/modules/airi-card.test.ts

@@ -117,10 +117,7 @@ describe('airi-card store', () => {
     const pack = {
       id: 'vp-1',
       name: 'Neuro Sama',
-      provider: 'volcengine',
-      model: 'seed-tts-2.0',
       voiceId: 'voice-neuro',
-      ttsModelId: 'volcengine/neuro-pool',
       params: { pitch: '+20%', volume: '+5%' },
       costMultiplier: 1.5,
     }
@@ -130,15 +127,12 @@ describe('airi-card store', () => {
     expect(bound).toBe(true)
     expect(cardStore.activeCard?.extensions.airi.modules.speech).toMatchObject({
       provider: OFFICIAL_SPEECH_PROVIDER_ID,
-      model: 'volcengine/neuro-pool',
+      model: 'auto',
       voice_id: 'voice-neuro',
       voicePack: {
         packId: 'vp-1',
         name: 'Neuro Sama',
-        provider: 'volcengine',
-        model: 'seed-tts-2.0',
         voiceId: 'voice-neuro',
-        ttsModelId: 'volcengine/neuro-pool',
         params: { pitch: '+20%', volume: '+5%' },
         costMultiplier: 1.5,
       },
@@ -157,12 +151,9 @@ describe('airi-card store', () => {
     cardStore.bindVoicePackToActiveCard({
       id: 'vp-1',
       name: 'Frozen',
-      provider: 'volcengine',
-      model: 'seed-tts-2.0',
       voiceId: 'voice-a',
-      ttsModelId: 'volcengine/pool-a',
       params,
-      costMultiplier: 1,
+      costMultiplier: 2,
     })
 
     params.pitch = '-10%'

packages/stage-ui/src/stores/modules/airi-card.ts

@@ -23,10 +23,7 @@ export type VoicePackParams = Record<string, string | number | boolean | null>
 export interface VoicePackBindingInput {
   id: string
   name: string
-  provider: string
-  model: string
   voiceId: string
-  ttsModelId: string
   params: VoicePackParams
   costMultiplier: number
 }
@@ -34,10 +31,7 @@ export interface VoicePackBindingInput {
 export interface VoicePackSnapshot {
   packId: string
   name: string
-  provider: string
-  model: string
   voiceId: string
-  ttsModelId: string
   params: VoicePackParams
   costMultiplier: number
 }
@@ -210,8 +204,7 @@ export const useAiriCardStore = defineStore('airi-card', () => {
     return updateActiveCardModules(({ modules }) => {
       const existingVoicePack = modules.speech.voicePack
       const shouldKeepVoicePack = speech.provider === OFFICIAL_SPEECH_PROVIDER_ID
-        && existingVoicePack?.ttsModelId === speech.model
+        && existingVoicePack?.voiceId === speech.voice_id
-        && existingVoicePack.voiceId === speech.voice_id
 
       return {
         speech: {
@@ -368,10 +361,7 @@ export const useAiriCardStore = defineStore('airi-card', () => {
     const voicePack: VoicePackSnapshot = {
       packId: pack.id,
       name: pack.name,
-      provider: pack.provider,
-      model: pack.model,
       voiceId: pack.voiceId,
-      ttsModelId: pack.ttsModelId,
       params: { ...pack.params },
       costMultiplier: pack.costMultiplier,
     }
@@ -379,7 +369,7 @@ export const useAiriCardStore = defineStore('airi-card', () => {
     const speech: AiriExtension['modules']['speech'] = {
       ...extension.modules.speech,
       provider: OFFICIAL_SPEECH_PROVIDER_ID,
-      model: pack.ttsModelId,
+      model: 'auto',
       voice_id: pack.voiceId,
       voicePack,
     }
@@ -399,7 +389,7 @@ export const useAiriCardStore = defineStore('airi-card', () => {
     })
 
     activeSpeechProvider.value = OFFICIAL_SPEECH_PROVIDER_ID
-    activeSpeechModel.value = pack.ttsModelId
+    activeSpeechModel.value = 'auto'
     activeSpeechVoiceId.value = pack.voiceId
 
     return true

packages/stage-ui/src/stores/modules/speech.test.ts

@@ -66,10 +66,7 @@ describe('speech store helpers', () => {
     const voicePack = {
       packId: 'vp-1',
       name: 'Frozen',
-      provider: 'volcengine',
-      model: 'seed-tts-2.0',
       voiceId: 'voice-a',
-      ttsModelId: 'volcengine/pool-a',
       params: {},
       costMultiplier: 1,
     }
@@ -198,9 +195,9 @@ describe('speech store helpers', () => {
 
   /**
    * @example
-   * speechStore.resolveVoicePackSpeechInput({ text, voice, voicePack: { packId: 'vp-1', costMultiplier: 1.5 } })
+   * speechStore.resolveVoicePackSpeechInput({ text, voice, voicePack: { packId: 'vp-1' } })
    */
-  it('passes Voice Pack snapshot billing metadata through adapter options', () => {
+  it('passes only Voice Pack identity through adapter options', () => {
     const speechStore = useSpeechStore()
     const voice = {
       id: 'voice-1',
@@ -215,7 +212,6 @@ describe('speech store helpers', () => {
       params: {},
       voicePack: {
         packId: 'vp-1',
-        costMultiplier: 1.5,
       },
       supportsAdapterProsody: true,
     })
@@ -223,7 +219,6 @@ describe('speech store helpers', () => {
     expect(request.providerConfig.extraBody).toEqual({
       voice_pack: {
         pack_id: 'vp-1',
-        cost_multiplier: 1.5,
       },
     })
   })

packages/stage-ui/src/stores/modules/speech.ts

@@ -29,7 +29,7 @@ interface VoicePackSpeechInputOptions {
   voice: VoiceInfo
   providerConfig?: Record<string, unknown>
   params?: VoicePackParams
-  voicePack?: Pick<VoicePackSnapshot, 'packId' | 'costMultiplier'>
+  voicePack?: Pick<VoicePackSnapshot, 'packId'>
   forceSSML?: boolean
   supportsSSML?: boolean
   supportsAdapterProsody?: boolean
@@ -517,7 +517,6 @@ export const useSpeechStore = defineStore('speech', () => {
         ...(providerConfig.extraBody as Record<string, unknown> | undefined),
         voice_pack: {
           pack_id: options.voicePack.packId,
-          cost_multiplier: options.voicePack.costMultiplier,
           ...(needsProsody && options.supportsAdapterProsody
             ? { pitch, volume }
             : {}),