dsh-audiogen 0.2.0 → 0.3.2

package/src/protocol.ts

@@ -8,7 +8,7 @@
 export const AUDIOGEN_SETTINGS_NAMESPACE = 'dsh-audiogen'
 
 /** Published package version shared by the host updater and the client UI. */
-export const PLUGIN_VERSION = '0.2.0'
+export const PLUGIN_VERSION = '0.3.2'
 
 /** Same-origin route family (loopback-only, mirroring dsh-imagegen). */
 export const SETTINGS_API = {
@@ -45,7 +45,7 @@ export const HISTORY_API = {
 export const HISTORY_MAX = 50
 
 /** Audio generation modes. */
-export type AudioMode = 'tts' | 'music' | 'sfx'
+export type AudioMode = 'tts' | 'music' | 'sfx' | 'voice_design'
 
 /** The capability category of an audio model/voice. */
 export type AudioModelCategory =
@@ -94,6 +94,8 @@ export interface PresetProviderView {
   name: string
   apiUrl: string
   hint: string
+  /** Official vendor website, shown as a link in the channel editor. */
+  site?: string
   models: ModelMapping[]
 }
 
@@ -106,6 +108,8 @@ export interface GenerateAudioRequest {
   prompt: string
   /** Optional voice alias for TTS. */
   voice?: string
+  /** Optional preview text for voice-design APIs. */
+  previewText?: string
   /** Optional speaking rate / speed multiplier. */
   speed?: number
   /** Requested duration in seconds (music/sfx). */
@@ -118,6 +122,38 @@ export interface GenerateAudioRequest {
   channel?: string
   /** Upstream model id actually sent (host-filled from alias mapping). */
   upstream?: string
+
+  // ---- MiniMax TTS 专属字段（其他厂商渠道忽略）——————
+  /** MiniMax 音色情绪，如 happy/sad/angry/nervous/fearful/bored；默认按音色自身。 */
+  emotion?: string
+  /** MiniMax 音量，范围 0-10，默认 1。 */
+  vol?: number
+  /** MiniMax 音调偏移（半音），范围 -12~12，默认 0。 */
+  pitch?: number
+  /** MiniMax 文本归一化处理开关（默认 true）。 */
+  textNormalization?: boolean
+  /** MiniMax 数学公式朗读开关（默认 false）。 */
+  latexRead?: boolean
+  /** MiniMax 发音词典 tone 条目，元素形如 "处理/(chu3)(li3)" 或 "危险/dangerous"。 */
+  pronunciationTone?: string[]
+  /** MiniMax 采样率：16000/24000/32000/44100/48000，默认 32000。 */
+  sampleRate?: number
+  /** MiniMax 码率（bps）：64000-320000，默认 128000。 */
+  bitrate?: number
+  /** MiniMax 声道数：1 或 2，默认 1。 */
+  audioChannel?: number
+  /** MiniMax 强制 CBR 编码（avoid VBR），默认 false。 */
+  forceCbr?: boolean
+  /** MiniMax 字幕开关：true 时响应携带字幕内容/文件。 */
+  subtitleEnable?: boolean
+  /** MiniMax AIGC 水印开关。 */
+  aigcWatermark?: boolean
+  /** MiniMax 语言增强（language_boost），如"中英混读"，按模型支持情况。 */
+  languageBoost?: string
+  /** MiniMax 变声参数（voice_modify，speech-2.8 等支持）。 */
+  voiceModify?: { pitch?: number; intensity?: number; timbre?: number; soundEffects?: string }
+  /** MiniMax 双音色混合权重（timbre_weights）。 */
+  timbreWeights?: Array<{ voiceId: string; weight: number }>
 }
 
 /** One generated audio, normalized host-side to base64. */
@@ -134,6 +170,8 @@ export interface GeneratedAudio {
   url: string
   /** Stable audio id / file name. */
   id: string
+  /** Optional voice id returned by a voice-design API. */
+  voiceId?: string
 }
 
 /** Successful generate outcome. */
@@ -153,6 +191,8 @@ export interface HistoryAudioRef {
   mime: string
   /** Duration in seconds when known. */
   duration?: number
+  /** Optional generated voice id. */
+  voiceId?: string
 }
 
 /** A saved generation as the browser consumes it. */
@@ -163,6 +203,7 @@ export interface HistoryEntry {
   model: string
   prompt: string
   voice?: string
+  voiceId?: string
   speed?: number
   duration?: number
   format?: string
@@ -179,6 +220,7 @@ export interface HistoryEntryInput {
   model: string
   prompt: string
   voice?: string
+  voiceId?: string
   speed?: number
   duration?: number
   format?: string

package/src/routes.ts

@@ -90,18 +90,56 @@ function messageOf(error: unknown): string {
 }
 
 function parseGenerateRequest(body: Record<string, unknown>): GenerateAudioRequest | undefined {
-  const mode = body.mode === 'music' ? 'music' : body.mode === 'sfx' ? 'sfx' : 'tts'
+  const mode = body.mode === 'music' ? 'music' : body.mode === 'sfx' ? 'sfx' : body.mode === 'voice_design' ? 'voice_design' : 'tts'
   const prompt = typeof body.prompt === 'string' ? body.prompt.trim() : ''
   if (prompt === '') return undefined
+  const num = (value: unknown): number | undefined => typeof value === 'number' && Number.isFinite(value) ? value : undefined
+  const str = (value: unknown): string | undefined => typeof value === 'string' && value.trim() !== '' ? value.trim() : undefined
+  const flag = (value: unknown): boolean | undefined => typeof value === 'boolean' ? value : undefined
+  const tone = Array.isArray(body.pronunciationTone)
+    ? body.pronunciationTone.filter((item): item is string => typeof item === 'string' && item.trim() !== '').map(item => item.trim())
+    : undefined
+  const voiceModifyRaw = body.voiceModify
+  const voiceModify = typeof voiceModifyRaw === 'object' && voiceModifyRaw !== null
+    ? {
+      ...(num((voiceModifyRaw as Record<string, unknown>).pitch) !== undefined ? { pitch: num((voiceModifyRaw as Record<string, unknown>).pitch)! } : {}),
+      ...(num((voiceModifyRaw as Record<string, unknown>).intensity) !== undefined ? { intensity: num((voiceModifyRaw as Record<string, unknown>).intensity)! } : {}),
+      ...(num((voiceModifyRaw as Record<string, unknown>).timbre) !== undefined ? { timbre: num((voiceModifyRaw as Record<string, unknown>).timbre)! } : {}),
+      ...(str((voiceModifyRaw as Record<string, unknown>).soundEffects) !== undefined ? { soundEffects: str((voiceModifyRaw as Record<string, unknown>).soundEffects)! } : {}),
+    }
+    : undefined
+  const timbreWeights = Array.isArray(body.timbreWeights)
+    ? body.timbreWeights
+      .filter((item): item is Record<string, unknown> => typeof item === 'object' && item !== null && typeof (item as Record<string, unknown>).voiceId === 'string' && typeof (item as Record<string, unknown>).weight === 'number')
+      .map(item => ({ voiceId: (item.voiceId as string).trim(), weight: item.weight as number }))
+      .filter(item => item.voiceId !== '')
+    : undefined
   return {
     mode,
     model: typeof body.model === 'string' ? body.model.trim() : '',
     prompt,
     ...(typeof body.voice === 'string' && body.voice.trim() !== '' ? { voice: body.voice.trim() } : {}),
-    ...(typeof body.speed === 'number' ? { speed: body.speed } : {}),
-    ...(typeof body.duration === 'number' ? { duration: body.duration } : {}),
+    ...(typeof body.previewText === 'string' && body.previewText.trim() !== '' ? { previewText: body.previewText.trim() } : {}),
+    ...(num(body.speed) !== undefined ? { speed: num(body.speed)! } : {}),
+    ...(num(body.duration) !== undefined ? { duration: num(body.duration)! } : {}),
     ...(typeof body.format === 'string' && body.format.trim() !== '' ? { format: body.format.trim() } : {}),
     ...(typeof body.channelId === 'string' && body.channelId !== '' ? { channelId: body.channelId } : {}),
+    // ---- MiniMax TTS 专属字段（其他厂商忽略） ----
+    ...(str(body.emotion) !== undefined ? { emotion: str(body.emotion)! } : {}),
+    ...(num(body.vol) !== undefined ? { vol: num(body.vol)! } : {}),
+    ...(num(body.pitch) !== undefined ? { pitch: num(body.pitch)! } : {}),
+    ...(flag(body.textNormalization) !== undefined ? { textNormalization: flag(body.textNormalization)! } : {}),
+    ...(flag(body.latexRead) !== undefined ? { latexRead: flag(body.latexRead)! } : {}),
+    ...(tone !== undefined && tone.length > 0 ? { pronunciationTone: tone } : {}),
+    ...(num(body.sampleRate) !== undefined ? { sampleRate: num(body.sampleRate)! } : {}),
+    ...(num(body.bitrate) !== undefined ? { bitrate: num(body.bitrate)! } : {}),
+    ...(num(body.audioChannel) !== undefined ? { audioChannel: num(body.audioChannel)! } : {}),
+    ...(flag(body.forceCbr) !== undefined ? { forceCbr: flag(body.forceCbr)! } : {}),
+    ...(flag(body.subtitleEnable) !== undefined ? { subtitleEnable: flag(body.subtitleEnable)! } : {}),
+    ...(flag(body.aigcWatermark) !== undefined ? { aigcWatermark: flag(body.aigcWatermark)! } : {}),
+    ...(str(body.languageBoost) !== undefined ? { languageBoost: str(body.languageBoost)! } : {}),
+    ...(voiceModify !== undefined ? { voiceModify } : {}),
+    ...(timbreWeights !== undefined && timbreWeights.length > 0 ? { timbreWeights } : {}),
   }
 }
 
@@ -141,6 +179,10 @@ function resolveChannelRequest(
   const target = explicit ?? defaults
   const asked = request.model.trim()
   if (asked === '') {
+    if (request.mode === 'voice_design') {
+      if (target === undefined) return { ok: false, code: 'no-channels', message: '尚未配置任何渠道' }
+      return { ok: true, request: { ...request, channelId: target.id, channel: target.name } }
+    }
     const alias = target?.models[0]?.alias ?? ''
     if (alias === '') {
       return { ok: false, code: 'no-models', message: `渠道「${target?.name ?? ''}」尚未配置模型/音色，请先在设置中添加` }
@@ -207,7 +249,9 @@ export function makeRoutes(deps: AudiogenRoutesDeps): WebRoute[] {
           ?? view.channels[0]
         const channel: AudioChannel = {
           id: stored?.id ?? 'preview',
-          preset: stored?.preset ?? '',
+          // The draft's own preset wins (even when empty = custom): discovery
+          // must follow the vendor being configured, not the default channel.
+          preset: typeof body?.preset === 'string' ? body.preset.trim() : (stored?.preset ?? ''),
           name: stored?.name ?? '',
           apiUrl: typeof body?.apiUrl === 'string' && body.apiUrl.trim() !== '' ? body.apiUrl.trim() : (stored?.apiUrl ?? ''),
           apiKey: typeof body?.apiKey === 'string' && body.apiKey.trim() !== '' ? body.apiKey.trim() : (stored?.apiKey ?? ''),
@@ -300,6 +344,7 @@ export function makeRoutes(deps: AudiogenRoutesDeps): WebRoute[] {
               mime: saved.mime,
               bytes: saved.bytes,
               url: `${AUDIO_API.file}/${encodeURIComponent(saved.file)}`,
+              ...(output.voiceId === undefined ? {} : { voiceId: output.voiceId }),
             })
           }
           let history