dsh-audiogen 0.3.3 → 0.3.5

package/lib/index.js

@@ -250,6 +250,97 @@ async function openAITTS(channel, request, signal) {
 async function elevenLabs(channel, request, signal) {
 	const base = endpointBase(channel.apiUrl);
 	const model = (request.upstream ?? request.model) || "eleven_multilingual_v2";
+	const headers = {
+		"xi-api-key": channel.apiKey.trim(),
+		authorization: `Bearer ${channel.apiKey.trim()}`,
+		"content-type": "application/json",
+		accept: "audio/mpeg, application/json"
+	};
+	if (request.mode === "voice_design") {
+		const endpoint = `${base}/text-to-voice/design`;
+		const previewText = request.previewText?.trim() ?? "";
+		const body = {
+			voice_description: request.prompt,
+			...previewText.length >= 100 ? { text: previewText } : { auto_generate_text: true }
+		};
+		const response = await fetchWithTimeout(endpoint, {
+			method: "POST",
+			redirect: "error",
+			headers,
+			body: JSON.stringify(body),
+			signal
+		}, UPSTREAM_TIMEOUT_MS);
+		if (!response.ok) {
+			const detail = await response.text().catch(() => "");
+			throw new AudioGenError(`ElevenLabs voice design API error (HTTP ${response.status})${detail === "" ? "" : `: ${detail.slice(0, 300)}`}`, "audio-api-error");
+		}
+		const previews = (await response.json()).previews ?? [];
+		if (previews.length === 0) throw new AudioGenError("ElevenLabs voice design returned no previews", "audio-empty-result");
+		const outputs = [];
+		for (const preview of previews) {
+			const encoded = preview.audio_base_64?.trim() ?? "";
+			if (encoded === "") continue;
+			const data = new Uint8Array(Buffer.from(encoded, "base64"));
+			outputs.push({
+				data,
+				mime: preview.media_type ?? "audio/mpeg",
+				...preview.generated_voice_id === void 0 || preview.generated_voice_id === "" ? {} : { voiceId: preview.generated_voice_id }
+			});
+		}
+		if (outputs.length === 0) throw new AudioGenError("ElevenLabs voice design returned no audio", "audio-empty-result");
+		return outputs;
+	}
+	if (request.mode === "music") {
+		const endpoint = `${base}/music`;
+		const body = {
+			model_id: (request.upstream ?? request.model) || "music_v1",
+			prompt: request.prompt,
+			...request.duration !== void 0 && Number.isFinite(request.duration) ? { music_length_ms: Math.round(Math.min(6e5, Math.max(3e3, request.duration * 1e3))) } : {},
+			...request.lyrics !== void 0 && request.lyrics.trim() !== "" ? { lyrics_text: request.lyrics.trim() } : {},
+			...request.isInstrumental !== void 0 ? { force_instrumental: request.isInstrumental } : {}
+		};
+		const response = await fetchWithTimeout(endpoint, {
+			method: "POST",
+			redirect: "follow",
+			headers,
+			body: JSON.stringify(body),
+			signal
+		}, UPSTREAM_TIMEOUT_MS);
+		if (!response.ok) {
+			const detail = await response.text().catch(() => "");
+			throw new AudioGenError(`ElevenLabs music API error (HTTP ${response.status})${detail === "" ? "" : `: ${detail.slice(0, 300)}`}`, "audio-api-error");
+		}
+		return normalizeAudioResponse(response, {
+			apiKey: channel.apiKey,
+			fallbackMime: "audio/mpeg"
+		});
+	}
+	if (request.mode === "sfx") {
+		const endpoint = `${base}/sound-generation`;
+		const sfxModel = (request.upstream ?? request.model) || "eleven_text_to_sound_v2";
+		const body = {
+			text: request.prompt,
+			model_id: sfxModel,
+			...request.duration !== void 0 && Number.isFinite(request.duration) ? { duration_seconds: Math.min(30, Math.max(.5, request.duration)) } : {},
+			...request.loop !== void 0 ? { loop: request.loop } : {},
+			...request.promptInfluence !== void 0 && Number.isFinite(request.promptInfluence) ? { prompt_influence: Math.min(1, Math.max(0, request.promptInfluence)) } : {}
+		};
+		const response = await fetchWithTimeout(endpoint, {
+			method: "POST",
+			redirect: "follow",
+			headers,
+			body: JSON.stringify(body),
+			signal
+		}, UPSTREAM_TIMEOUT_MS);
+		if (!response.ok) {
+			const detail = await response.text().catch(() => "");
+			throw new AudioGenError(`ElevenLabs sound effects API error (HTTP ${response.status})${detail === "" ? "" : `: ${detail.slice(0, 300)}`}`, "audio-api-error");
+		}
+		return normalizeAudioResponse(response, {
+			apiKey: channel.apiKey,
+			fallbackMime: "audio/mpeg"
+		});
+	}
 	const voiceId = (request.voice ?? request.model ?? model).trim();
 	const endpoint = `${base}/text-to-speech/${encodeURIComponent(voiceId)}`;
 	const body = {
@@ -266,11 +357,7 @@ async function elevenLabs(channel, request, signal) {
 	return normalizeAudioResponse(await fetchWithTimeout(endpoint, {
 		method: "POST",
 		redirect: "error",
-		headers: {
-			"xi-api-key": channel.apiKey.trim(),
-			"content-type": "application/json",
-			accept: "audio/mpeg, application/json"
-		},
+		headers,
 		body: JSON.stringify(body),
 		signal
 	}, UPSTREAM_TIMEOUT_MS), {
@@ -563,7 +650,7 @@ async function generateAudio(channel, request, signal) {
 	if (channel.apiUrl.trim() === "") throw new AudioGenError("channel API URL is not configured", "audio-no-endpoint");
 	if (channel.apiKey.trim() === "") throw new AudioGenError("channel API key is not configured", "audio-no-key");
 	if (request.prompt.trim() === "") throw new AudioGenError("audio prompt/text is required", "audio-empty-prompt");
-	if (request.mode === "voice_design" && !isMiniMax$1(channel)) throw new AudioGenError("音色设计当前仅支持 MiniMax 渠道", "voice-design-unsupported");
+	if (request.mode === "voice_design" && !isMiniMax$1(channel) && !isElevenLabs$1(channel)) throw new AudioGenError("音色设计当前仅支持 MiniMax（/v1/voice_design）与 ElevenLabs（/v1/text-to-voice/design）渠道", "voice-design-unsupported");
 	if (isElevenLabs$1(channel)) return elevenLabs(channel, request, signal);
 	if (isMiniMax$1(channel)) return minimax(channel, request, signal);
 	if (isStability$1(channel)) return stabilityAudio(channel, request, signal);
@@ -642,7 +729,7 @@ const AUDIO_PRESETS = [
 		name: "ElevenLabs",
 		apiUrl: "https://api.elevenlabs.io/v1",
 		site: "https://elevenlabsai.cn",
-		hint: "ElevenLabs 语音合成（TTS）；建议点击「获取可用模型」拉取音色与模型",
+		hint: "ElevenLabs 语音合成（TTS）与音乐生成（POST /v1/music，music_v2）；可点「获取可用模型」拉取音色与模型",
 		models: [
 			{
 				alias: "Rachel",
@@ -678,6 +765,21 @@ const AUDIO_PRESETS = [
 				alias: "eleven_flash_v2_5",
 				id: "eleven_flash_v2_5",
 				category: "tts"
+			},
+			{
+				alias: "music_v2",
+				id: "music_v2",
+				category: "music"
+			},
+			{
+				alias: "music_v1",
+				id: "music_v1",
+				category: "music"
+			},
+			{
+				alias: "eleven_text_to_sound_v2",
+				id: "eleven_text_to_sound_v2",
+				category: "sfx"
 			}
 		]
 	},
@@ -1078,6 +1180,8 @@ function parseGenerateRequest(body) {
 		...num(body.duration) !== void 0 ? { duration: num(body.duration) } : {},
 		...typeof body.lyrics === "string" && body.lyrics.trim() !== "" ? { lyrics: body.lyrics.trim() } : {},
 		...typeof body.isInstrumental === "boolean" ? { isInstrumental: body.isInstrumental } : {},
+		...typeof body.loop === "boolean" ? { loop: body.loop } : {},
+		...num(body.promptInfluence) !== void 0 ? { promptInfluence: num(body.promptInfluence) } : {},
 		...typeof body.format === "string" && body.format.trim() !== "" ? { format: body.format.trim() } : {},
 		...typeof body.channelId === "string" && body.channelId !== "" ? { channelId: body.channelId } : {},
 		...str(body.emotion) !== void 0 ? { emotion: str(body.emotion) } : {},
@@ -1136,22 +1240,24 @@ function resolveChannelRequest(request, view) {
 	const defaults = view.channels.find((candidate) => candidate.id === view.defaultChannelId) ?? view.channels[0];
 	const target = explicit ?? defaults;
 	const asked = request.model.trim();
+	if (request.mode === "voice_design") {
+		if (target === void 0) return {
+			ok: false,
+			code: "no-channels",
+			message: "尚未配置任何渠道"
+		};
+		return {
+			ok: true,
+			request: {
+				...request,
+				model: "",
+				upstream: void 0,
+				channelId: target.id,
+				channel: target.name
+			}
+		};
+	}
 	if (asked === "") {
-		if (request.mode === "voice_design") {
-			if (target === void 0) 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,
@@ -1579,7 +1685,7 @@ function ensureConfigured(config) {
 function registerAgentAudioTools(ctx, resolve) {
 	return ctx.tools.register(defineTool({
 		name: "generate_audio",
-		description: "Generate audio with the configured audio provider. Supports text-to-speech, music generation, sound effects and MiniMax voice design. The tool call waits for the upstream result and returns same-origin audio URLs; pass those URLs to the user for playback or download. If multiple models are configured, first ask the user which one to use or pass model explicitly.",
+		description: "Generate audio with the configured audio provider. Supports text-to-speech, music generation, sound effects and voice design (MiniMax /v1/voice_design, ElevenLabs /v1/text-to-voice/design). The tool call waits for the upstream result and returns same-origin audio URLs; pass those URLs to the user for playback or download. If multiple models are configured, first ask the user which one to use or pass model explicitly.",
 		parameters: {
 			prompt: {
 				type: "string",
@@ -1624,6 +1730,14 @@ function registerAgentAudioTools(ctx, resolve) {
 				type: "boolean",
 				description: "Generate purely instrumental music without vocals/lyrics (MiniMax is_instrumental). When true, lyrics may be omitted."
 			},
+			loop: {
+				type: "boolean",
+				description: "Create a seamlessly looping sound effect (ElevenLabs sound generation loop, only for eleven_text_to_sound_v2)."
+			},
+			prompt_influence: {
+				type: "number",
+				description: "Sound effect prompt influence 0-1 (ElevenLabs prompt_influence, default 0.3): higher follows the prompt more closely, lower is more variable."
+			},
 			format: {
 				type: "string",
 				description: "Output format such as mp3 or wav. MiniMax music supports mp3/wav/pcm."
@@ -1655,11 +1769,11 @@ function registerAgentAudioTools(ctx, resolve) {
 			},
 			sample_rate: {
 				type: "integer",
-				description: "MiniMax TTS sample rate: 16000/24000/32000/44100/48000, default 32000 (audio_setting.sample_rate)."
+				description: "MiniMax sample rate: music 16000/24000/32000/44100 (default 44100); tts default 32000 (audio_setting.sample_rate)."
 			},
 			bitrate: {
 				type: "integer",
-				description: "MiniMax TTS bitrate in bps: 64000-320000, default 128000 (audio_setting.bitrate)."
+				description: "MiniMax bitrate in bps: 32000/64000/128000/256000 (music default 256000, tts default 128000; audio_setting.bitrate)."
 			},
 			channel: {
 				type: "integer",
@@ -1764,6 +1878,8 @@ function registerAgentAudioTools(ctx, resolve) {
 				...typeof args.duration === "number" ? { duration: args.duration } : {},
 				...typeof args.lyrics === "string" && args.lyrics.trim() !== "" ? { lyrics: args.lyrics.trim() } : {},
 				...typeof args.is_instrumental === "boolean" ? { isInstrumental: args.is_instrumental } : {},
+				...typeof args.loop === "boolean" ? { loop: args.loop } : {},
+				...typeof args.prompt_influence === "number" && Number.isFinite(args.prompt_influence) ? { promptInfluence: args.prompt_influence } : {},
 				...typeof args.format === "string" && args.format.trim() !== "" ? { format: args.format.trim() } : {},
 				...typeof args.emotion === "string" && args.emotion.trim() !== "" ? { emotion: args.emotion.trim() } : {},
 				...typeof args.vol === "number" && Number.isFinite(args.vol) ? { vol: args.vol } : {},

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "dsh-audiogen",
   "description": "AI audio generation plugin for the dsh web GUI: multi-vendor TTS/music/sound-effect channels (OpenAI-compatible, ElevenLabs, MiniMax, Stability AI and custom), per-channel model/voice catalogs, Agent tool and a sidebar AI 音频 panel.",
-  "version": "0.3.3",
+  "version": "0.3.5",
   "type": "module",
   "main": "lib/index.js",
   "exports": {

package/skills/design/SKILL.md

@@ -7,7 +7,13 @@
 - 帮助用户把音色或音效需求细化成可复用的描述，供后续 TTS/music/sfx 生成。
 - 输出不建议直接生成音频，而是给出音色参数建议和可用的模型/音色清单。
 
+## 面板/工具的音色设计（voice_design）
+- 渠道支持：MiniMax（POST /v1/voice_design，prompt + preview_text）与 ElevenLabs（POST /v1/text-to-voice/design）。
+- MiniMax 参数：`prompt`（音色描述）、`preview_text`（试听文本，任意长度）。
+- ElevenLabs 参数：`voice_description`（音色描述）；试听文本 `preview_text` 需 100-1000 字符，过短时自动 `auto_generate_text`；响应含 `previews[].audio_base_64` 与 `generated_voice_id`（可后续在 TTS 中复用该 voice_id）。
+- 面板音色设计模式顶部可切换「厂商 / 渠道」。
+
 ## 流程
 1. 询问目标风格、音域、情绪、适用场景。
 2. 生成结构化的音色描述（如温暖复古合成器、未来感 UI 提示音）。
-3. 在后续生成中复用该描述，必要时通过 `generate_audio` 试听。
+3. 在后续生成中复用该描述，必要时通过 `generate_audio`（mode=voice_design）试听。

package/skills/music/SKILL.md

@@ -37,3 +37,16 @@
 ## 常见错误
 - `lyrics-required`：MiniMax 音乐生成需要歌词，或开启纯音乐。
 - `HTTP 400` 且含 `2013`：上游参数不合法，检查 lyrics / audio_setting 枚举。
+
+## ElevenLabs Music（POST /v1/music，模型 music_v1 / music_v2）
+
+| 字段 | 工具/面板参数 | 说明 |
+| --- | --- | --- |
+| model_id | model（music_v2 等） | 官方枚举：music_v1 / music_v2，默认 music_v1 |
+| prompt | prompt | 音乐/歌词主题描述（不能与 composition_plan 同用；引擎用 prompt） |
+| music_length_ms | duration（秒，自动×1000） | 3000ms - 600000ms（3s-600s），超出自动收敛区间 |
+| lyrics_text | lyrics（歌词） | 选填歌词文本 |
+| force_instrumental | is_instrumental（纯音乐） | true 保证无演唱（无词） |
+| seed / generation_mode / finetune_* | — | 高级字段，暂未透出 |
+
+> 响应为音频字节流（audio/*，常为 mp3）。请求同时携带 `xi-api-key` 与 `Authorization: Bearer`，以兼容 New API 类网关（官方站任一头即可）。

package/skills/sfx/SKILL.md

@@ -2,15 +2,29 @@
 
 ## 触发
 - `/audio:sfx <描述>`
-- 用户说“生成一个音效 / 提示音 / 環境音”
+- 用户说“生成一个音效 / 提示音 / 环境音”
 
 ## 参数
 - prompt: 必填，音效描述
-- model: 可选，已配置模型
-- duration: 可选，秒数
+- model: 可选，已配置模型（ElevenLabs：eleven_text_to_sound_v2）
+- duration: 可选，秒数（ElevenLabs 0.5-30，留空则自动）
+- loop: 可选，ElevenLabs 无缝循环音效（仅 eleven_text_to_sound_v2）
+- prompt_influence: 可选，ElevenLabs 提示词影响度 0-1（默认 0.3，越高越贴近提示词、越少随机）
 - format: 可选
 
 ## 流程
 1. 确认已配置音效生成渠道。
 2. 调用 `generate_audio`，mode=sfx。
 3. 将音频 URL 返回。
+
+## ElevenLabs Sound Generation（POST /v1/sound-generation）
+
+| 字段 | 工具/面板参数 | 说明 |
+| --- | --- | --- |
+| text | prompt | 必填，转换为音效的文本/描述 |
+| model_id | model | 官方枚举：eleven_text_to_sound_v2（默认） |
+| duration_seconds | duration | 0.5-30 秒；留空由提示词推算最优时长 |
+| loop | loop | 是否生成平滑循环音效（仅 eleven_text_to_sound_v2） |
+| prompt_influence | prompt_influence | 0-1，默认 0.3；越高越贴提示词，越低越多样 |
+
+> 响应为 audio/mpeg 二进制。请求同时携带 `xi-api-key` 与 `Authorization: Bearer`，兼容 New API 类网关。

package/src/agent-audio-tools.ts

@@ -100,7 +100,7 @@ function ensureConfigured(config: AgentAudioToolConfig): void {
 export function registerAgentAudioTools(ctx: Context, resolve: () => AgentAudioToolConfig): () => void {
   const disposer = ctx.tools.register(defineTool({
     name: 'generate_audio',
-    description: 'Generate audio with the configured audio provider. Supports text-to-speech, music generation, sound effects and MiniMax voice design. The tool call waits for the upstream result and returns same-origin audio URLs; pass those URLs to the user for playback or download. If multiple models are configured, first ask the user which one to use or pass model explicitly.',
+    description: 'Generate audio with the configured audio provider. Supports text-to-speech, music generation, sound effects and voice design (MiniMax /v1/voice_design, ElevenLabs /v1/text-to-voice/design). The tool call waits for the upstream result and returns same-origin audio URLs; pass those URLs to the user for playback or download. If multiple models are configured, first ask the user which one to use or pass model explicitly.',
     parameters: {
       prompt: { type: 'string', required: true, description: 'For tts, the text to speak. For music/sfx, a descriptive prompt.' },
       mode: { type: 'string', enum: ['tts', 'music', 'sfx', 'voice_design'], description: 'Generation mode. Defaults to tts.' },
@@ -111,6 +111,8 @@ export function registerAgentAudioTools(ctx: Context, resolve: () => AgentAudioT
       duration: { type: 'number', description: 'Requested duration in seconds for music/sfx.' },
       lyrics: { type: 'string', description: 'Lyrics for music generation (MiniMax music-3.0/music-cover). Required unless is_instrumental is true. Split verses with an empty line.' },
       is_instrumental: { type: 'boolean', description: 'Generate purely instrumental music without vocals/lyrics (MiniMax is_instrumental). When true, lyrics may be omitted.' },
+      loop: { type: 'boolean', description: 'Create a seamlessly looping sound effect (ElevenLabs sound generation loop, only for eleven_text_to_sound_v2).' },
+      prompt_influence: { type: 'number', description: 'Sound effect prompt influence 0-1 (ElevenLabs prompt_influence, default 0.3): higher follows the prompt more closely, lower is more variable.' },
       format: { type: 'string', description: 'Output format such as mp3 or wav. MiniMax music supports mp3/wav/pcm.' },
       // ---- MiniMax TTS only (ignored by other providers) ----
       emotion: { type: 'string', description: 'MiniMax TTS emotion, e.g. happy/sad/angry/nervous/fearful/bored (voice_setting.emotion).' },
@@ -119,8 +121,8 @@ export function registerAgentAudioTools(ctx: Context, resolve: () => AgentAudioT
       text_normalization: { type: 'boolean', description: 'MiniMax TTS text normalization switch (voice_setting.text_normalization).' },
       latex_read: { type: 'boolean', description: 'MiniMax TTS math formula reading switch (voice_setting.latex_read).' },
       pronunciation_tone: { type: 'array', items: { type: 'string' }, description: 'MiniMax TTS pronunciation dictionary tone entries, each "word/pronunciation", e.g. ["处理/(chu3)(li3)", "危险/dangerous"] (pronunciation_dict.tone).' },
-      sample_rate: { type: 'integer', description: 'MiniMax TTS sample rate: 16000/24000/32000/44100/48000, default 32000 (audio_setting.sample_rate).' },
-      bitrate: { type: 'integer', description: 'MiniMax TTS bitrate in bps: 64000-320000, default 128000 (audio_setting.bitrate).' },
+      sample_rate: { type: 'integer', description: 'MiniMax sample rate: music 16000/24000/32000/44100 (default 44100); tts default 32000 (audio_setting.sample_rate).' },
+      bitrate: { type: 'integer', description: 'MiniMax bitrate in bps: 32000/64000/128000/256000 (music default 256000, tts default 128000; audio_setting.bitrate).' },
       channel: { type: 'integer', description: 'MiniMax TTS audio channels: 1 or 2, default 1 (audio_setting.channel).' },
       force_cbr: { type: 'boolean', description: 'MiniMax TTS force CBR encoding (audio_setting.force_cbr).' },
       subtitle_enable: { type: 'boolean', description: 'MiniMax TTS subtitle output switch (subtitle_enable).' },
@@ -199,6 +201,8 @@ export function registerAgentAudioTools(ctx: Context, resolve: () => AgentAudioT
         ...(typeof args.duration === 'number' ? { duration: args.duration } : {}),
         ...(typeof args.lyrics === 'string' && args.lyrics.trim() !== '' ? { lyrics: args.lyrics.trim() } : {}),
         ...(typeof args.is_instrumental === 'boolean' ? { isInstrumental: args.is_instrumental } : {}),
+        ...(typeof args.loop === 'boolean' ? { loop: args.loop } : {}),
+        ...(typeof args.prompt_influence === 'number' && Number.isFinite(args.prompt_influence) ? { promptInfluence: args.prompt_influence } : {}),
         ...(typeof args.format === 'string' && args.format.trim() !== '' ? { format: args.format.trim() } : {}),
         // ---- MiniMax TTS 专属字段 ----
         ...(typeof args.emotion === 'string' && args.emotion.trim() !== '' ? { emotion: args.emotion.trim() } : {}),

package/src/audio-engine.ts

@@ -242,6 +242,114 @@ async function openAITTS(channel: AudioChannel, request: GenerateAudioRequest, s
 async function elevenLabs(channel: AudioChannel, request: GenerateAudioRequest, signal?: AbortSignal): Promise<Array<{ data: Uint8Array; mime: string; voiceId?: string }>> {
   const base = endpointBase(channel.apiUrl)
   const model = (request.upstream ?? request.model) || 'eleven_multilingual_v2'
+  // 官方使用 xi-api-key；额外携带 Authorization Bearer 以兼容 New API 类网关。
+  const headers = {
+    'xi-api-key': channel.apiKey.trim(),
+    authorization: `Bearer ${channel.apiKey.trim()}`,
+    'content-type': 'application/json',
+    accept: 'audio/mpeg, application/json',
+  }
+
+  // ------------- ElevenLabs Voice Design（POST /v1/text-to-voice/design） -------------
+  // voice_description 必填；text 100-1000 字符，过短时用 auto_generate_text；
+  // 返回 previews[].audio_base_64 与 previews[].generated_voice_id。
+  if (request.mode === 'voice_design') {
+    const endpoint = `${base}/text-to-voice/design`
+    const previewText = request.previewText?.trim() ?? ''
+    const body: Record<string, unknown> = {
+      voice_description: request.prompt,
+      ...(previewText.length >= 100 ? { text: previewText } : { auto_generate_text: true }),
+    }
+    const response = await fetchWithTimeout(endpoint, {
+      method: 'POST',
+      redirect: 'error',
+      headers,
+      body: JSON.stringify(body),
+      signal,
+    }, UPSTREAM_TIMEOUT_MS)
+    if (!response.ok) {
+      const detail = await response.text().catch(() => '')
+      throw new AudioGenError(`ElevenLabs voice design API error (HTTP ${response.status})${detail === '' ? '' : `: ${detail.slice(0, 300)}`}`, 'audio-api-error')
+    }
+    const payload = await response.json() as {
+      previews?: Array<{ audio_base_64?: string; generated_voice_id?: string; media_type?: string }>
+    }
+    const previews = payload.previews ?? []
+    if (previews.length === 0) throw new AudioGenError('ElevenLabs voice design returned no previews', 'audio-empty-result')
+    const outputs: Array<{ data: Uint8Array; mime: string; voiceId?: string }> = []
+    for (const preview of previews) {
+      const encoded = preview.audio_base_64?.trim() ?? ''
+      if (encoded === '') continue
+      const data = new Uint8Array(Buffer.from(encoded, 'base64'))
+      outputs.push({
+        data,
+        mime: preview.media_type ?? 'audio/mpeg',
+        ...(preview.generated_voice_id === undefined || preview.generated_voice_id === '' ? {} : { voiceId: preview.generated_voice_id }),
+      })
+    }
+    if (outputs.length === 0) throw new AudioGenError('ElevenLabs voice design returned no audio', 'audio-empty-result')
+    return outputs
+  }
+
+  // ------------- ElevenLabs Music（POST /v1/music） -------------
+  // 模型：music_v1 / music_v2；prompt 与 composition_plan 二选一（引擎用 prompt）。
+  if (request.mode === 'music') {
+    const endpoint = `${base}/music`
+    const musicModel = (request.upstream ?? request.model) || 'music_v1'
+    const body: Record<string, unknown> = {
+      model_id: musicModel,
+      prompt: request.prompt,
+      ...(request.duration !== undefined && Number.isFinite(request.duration)
+        ? { music_length_ms: Math.round(Math.min(600_000, Math.max(3_000, request.duration * 1000))) }
+        : {}),
+      ...(request.lyrics !== undefined && request.lyrics.trim() !== '' ? { lyrics_text: request.lyrics.trim() } : {}),
+      ...(request.isInstrumental !== undefined ? { force_instrumental: request.isInstrumental } : {}),
+    }
+    const response = await fetchWithTimeout(endpoint, {
+      method: 'POST',
+      redirect: 'follow',
+      headers,
+      body: JSON.stringify(body),
+      signal,
+    }, UPSTREAM_TIMEOUT_MS)
+    if (!response.ok) {
+      const detail = await response.text().catch(() => '')
+      throw new AudioGenError(`ElevenLabs music API error (HTTP ${response.status})${detail === '' ? '' : `: ${detail.slice(0, 300)}`}`, 'audio-api-error')
+    }
+    return normalizeAudioResponse(response, { apiKey: channel.apiKey, fallbackMime: 'audio/mpeg' })
+  }
+
+  // ------------- ElevenLabs Sound Effects（POST /v1/sound-generation） -------------
+  // 官方模型：eleven_text_to_sound_v2；text 必填；duration_seconds 0.5-30；
+  // loop 仅该模型可用；prompt_influence 0-1（默认 0.3）。
+  if (request.mode === 'sfx') {
+    const endpoint = `${base}/sound-generation`
+    const sfxModel = (request.upstream ?? request.model) || 'eleven_text_to_sound_v2'
+    const body: Record<string, unknown> = {
+      text: request.prompt,
+      model_id: sfxModel,
+      ...(request.duration !== undefined && Number.isFinite(request.duration)
+        ? { duration_seconds: Math.min(30, Math.max(0.5, request.duration)) }
+        : {}),
+      ...(request.loop !== undefined ? { loop: request.loop } : {}),
+      ...(request.promptInfluence !== undefined && Number.isFinite(request.promptInfluence)
+        ? { prompt_influence: Math.min(1, Math.max(0, request.promptInfluence)) }
+        : {}),
+    }
+    const response = await fetchWithTimeout(endpoint, {
+      method: 'POST',
+      redirect: 'follow',
+      headers,
+      body: JSON.stringify(body),
+      signal,
+    }, UPSTREAM_TIMEOUT_MS)
+    if (!response.ok) {
+      const detail = await response.text().catch(() => '')
+      throw new AudioGenError(`ElevenLabs sound effects API error (HTTP ${response.status})${detail === '' ? '' : `: ${detail.slice(0, 300)}`}`, 'audio-api-error')
+    }
+    return normalizeAudioResponse(response, { apiKey: channel.apiKey, fallbackMime: 'audio/mpeg' })
+  }
+
   const voiceId = (request.voice ?? request.model ?? model).trim()
   const endpoint = `${base}/text-to-speech/${encodeURIComponent(voiceId)}`
   const body: Record<string, unknown> = {
@@ -258,11 +366,7 @@ async function elevenLabs(channel: AudioChannel, request: GenerateAudioRequest,
   const response = await fetchWithTimeout(endpoint, {
     method: 'POST',
     redirect: 'error',
-    headers: {
-      'xi-api-key': channel.apiKey.trim(),
-      'content-type': 'application/json',
-      accept: 'audio/mpeg, application/json',
-    },
+    headers,
     body: JSON.stringify(body),
     signal,
   }, UPSTREAM_TIMEOUT_MS)
@@ -591,8 +695,8 @@ export async function generateAudio(
   if (channel.apiUrl.trim() === '') throw new AudioGenError('channel API URL is not configured', 'audio-no-endpoint')
   if (channel.apiKey.trim() === '') throw new AudioGenError('channel API key is not configured', 'audio-no-key')
   if (request.prompt.trim() === '') throw new AudioGenError('audio prompt/text is required', 'audio-empty-prompt')
-  if (request.mode === 'voice_design' && !isMiniMax(channel)) {
-    throw new AudioGenError('音色设计当前仅支持 MiniMax 渠道', 'voice-design-unsupported')
+  if (request.mode === 'voice_design' && !isMiniMax(channel) && !isElevenLabs(channel)) {
+    throw new AudioGenError('音色设计当前仅支持 MiniMax（/v1/voice_design）与 ElevenLabs（/v1/text-to-voice/design）渠道', 'voice-design-unsupported')
   }
 
   if (isElevenLabs(channel)) return elevenLabs(channel, request, signal)

package/src/audio-presets.ts

@@ -53,7 +53,7 @@ export const AUDIO_PRESETS: AudioPresetProvider[] = [
     name: 'ElevenLabs',
     apiUrl: 'https://api.elevenlabs.io/v1',
     site: 'https://elevenlabsai.cn',
-    hint: 'ElevenLabs 语音合成（TTS）；建议点击「获取可用模型」拉取音色与模型',
+    hint: 'ElevenLabs 语音合成（TTS）与音乐生成（POST /v1/music，music_v2）；可点「获取可用模型」拉取音色与模型',
     models: [
       { alias: 'Rachel', id: '21m00Tcm4TlvDq8ikWAM', category: 'tts' },
       { alias: 'Adam', id: 'pNInz6obpgDQGcFmaJgB', category: 'tts' },
@@ -62,6 +62,11 @@ export const AUDIO_PRESETS: AudioPresetProvider[] = [
       { alias: 'eleven_multilingual_v2', id: 'eleven_multilingual_v2', category: 'tts' },
       { alias: 'eleven_turbo_v2_5', id: 'eleven_turbo_v2_5', category: 'tts' },
       { alias: 'eleven_flash_v2_5', id: 'eleven_flash_v2_5', category: 'tts' },
+      // ElevenLabs Music（POST /v1/music）
+      { alias: 'music_v2', id: 'music_v2', category: 'music' },
+      { alias: 'music_v1', id: 'music_v1', category: 'music' },
+      // ElevenLabs Sound Effects（POST /v1/sound-generation）
+      { alias: 'eleven_text_to_sound_v2', id: 'eleven_text_to_sound_v2', category: 'sfx' },
     ],
   },
   {

package/src/client/AudioGenPanel.tsx

@@ -62,6 +62,9 @@ export function AudioGenPanel(props: { api: AudiogenApi; scope: AudiogenScope })
   const [duration, setDuration] = useState('')
   const [lyrics, setLyrics] = useState('')
   const [instrumental, setInstrumental] = useState(false)
+  // ElevenLabs 音效参数
+  const [loop, setLoop] = useState(false)
+  const [promptInfluence, setPromptInfluence] = useState('')
   const [format, setFormat] = useState('mp3')
   // MiniMax TTS 高级参数（其他厂商忽略）
   const [emotion, setEmotion] = useState('')
@@ -76,12 +79,21 @@ export function AudioGenPanel(props: { api: AudiogenApi; scope: AudiogenScope })
   const [error, setError] = useState<string | null>(null)
   const [outputs, setOutputs] = useState<GeneratedAudio[]>([])
   const { entries, reload, clear } = useHistory()
+  // 音色设计模式的厂商/渠道选择（默认渠道）
+  const [designChannelId, setDesignChannelId] = useState('')
 
   const isMiniMaxChannel = useMemo(() => {
     const target = channels.find(candidate => candidate.id === modelOptions.defaultChannelId) ?? channels[0]
     return target !== undefined && (target.preset === 'minimax' || /minimax/i.test(target.apiUrl))
   }, [channels, modelOptions.defaultChannelId])
 
+  useEffect(() => {
+    if (channels.length === 0) return
+    if (designChannelId === '' || !channels.some(candidate => candidate.id === designChannelId)) {
+      setDesignChannelId(modelOptions.defaultChannelId ?? channels[0]!.id)
+    }
+  }, [channels, modelOptions.defaultChannelId, designChannelId])
+
   const visibleModels = useMemo(() => {
     if (mode === 'voice_design') return []
     return modelOptions.models
@@ -105,14 +117,17 @@ export function AudioGenPanel(props: { api: AudiogenApi; scope: AudiogenScope })
     try {
       const response = await api.generate({
         mode,
-        model: (model || visibleModels[0]) ?? '',
+        model: mode === 'voice_design' ? '' : (model || visibleModels[0]) ?? '',
         prompt: prompt.trim(),
+        ...(mode === 'voice_design' && designChannelId !== '' ? { channelId: designChannelId } : {}),
         ...(previewText.trim() !== '' ? { previewText: previewText.trim() } : {}),
         ...(voice.trim() !== '' ? { voice: voice.trim() } : {}),
         ...(speed.trim() !== '' ? { speed: Number(speed) } : {}),
         ...(duration.trim() !== '' ? { duration: Number(duration) } : {}),
         ...(lyrics.trim() !== '' ? { lyrics: lyrics.trim() } : {}),
         ...(instrumental ? { isInstrumental: true } : {}),
+        ...(loop ? { loop: true } : {}),
+        ...(promptInfluence.trim() !== '' ? { promptInfluence: Number(promptInfluence) } : {}),
         ...(format.trim() !== '' ? { format: format.trim() } : {}),
         ...(emotion.trim() !== '' ? { emotion: emotion.trim() } : {}),
         ...(vol.trim() !== '' ? { vol: Number(vol) } : {}),
@@ -173,10 +188,24 @@ export function AudioGenPanel(props: { api: AudiogenApi; scope: AudiogenScope })
           </label>
 
           {mode === 'voice_design' ? (
-            <label className={css.label}>
-              <span>试听文本</span>
-              <input className={css.input} value={previewText} onChange={event => setPreviewText(event.target.value)} placeholder="你好，这是新设计的音色试听。" />
-            </label>
+            <>
+              <label className={css.label}>
+                <span>厂商 / 渠道</span>
+                <select className={css.select} value={designChannelId} onChange={event => setDesignChannelId(event.target.value)}>
+                  {channels.length === 0 ? <option value="">（尚未配置渠道）</option> : null}
+                  {channels.map(candidate => (
+                    <option key={candidate.id} value={candidate.id}>
+                      {candidate.name}（{candidate.preset === 'minimax' ? 'MiniMax' : candidate.preset === 'elevenlabs' ? 'ElevenLabs' : candidate.preset || '自定义'}）
+                    </option>
+                  ))}
+                </select>
+              </label>
+              <p className={css.hint}>MiniMax：/v1/voice_design；ElevenLabs：/v1/text-to-voice/design（试听文本 100-1000 字符，过短将自动生成）</p>
+              <label className={css.label}>
+                <span>试听文本</span>
+                <input className={css.input} value={previewText} onChange={event => setPreviewText(event.target.value)} placeholder="你好，这是新设计的音色试听。" />
+              </label>
+            </>
           ) : null}
 
           {needModel ? (
@@ -256,6 +285,19 @@ export function AudioGenPanel(props: { api: AudiogenApi; scope: AudiogenScope })
             </label>
           ) : null}
 
+          {mode === 'sfx' ? (
+            <>
+              <label className={css.checkbox}>
+                <input type="checkbox" checked={loop} onChange={event => setLoop(event.target.checked)} />
+                <span>循环音效 loop（无缝循环，需 eleven_text_to_sound_v2）</span>
+              </label>
+              <label className={css.label}>
+                <span>提示词影响度 prompt_influence (0-1)</span>
+                <input className={css.input} type="number" step="0.1" min="0" max="1" value={promptInfluence} onChange={event => setPromptInfluence(event.target.value)} placeholder="0.3" />
+              </label>
+            </>
+          ) : null}
+
           {mode === 'music' ? (
             <>
               <label className={css.label}>
@@ -266,6 +308,28 @@ export function AudioGenPanel(props: { api: AudiogenApi; scope: AudiogenScope })
                 <input type="checkbox" checked={instrumental} onChange={event => setInstrumental(event.target.checked)} />
                 <span>纯音乐（无歌词/人声）is_instrumental</span>
               </label>
+              <div className={css.row}>
+                <label className={css.label}>
+                  <span>采样率</span>
+                  <select className={css.select} value={sampleRate} onChange={event => setSampleRate(event.target.value)}>
+                    <option value="">默认（44100）</option>
+                    <option value="16000">16000</option>
+                    <option value="24000">24000</option>
+                    <option value="32000">32000</option>
+                    <option value="44100">44100</option>
+                  </select>
+                </label>
+                <label className={css.label}>
+                  <span>码率 bps</span>
+                  <select className={css.select} value={bitrate} onChange={event => setBitrate(event.target.value)}>
+                    <option value="">默认（256000）</option>
+                    <option value="32000">32000</option>
+                    <option value="64000">64000</option>
+                    <option value="128000">128000</option>
+                    <option value="256000">256000</option>
+                  </select>
+                </label>
+              </div>
             </>
           ) : null}