feat: add missing audio models (music, TTS schemas, sound effects)

- Add text-to-music and text-to-sound-effects to fal.ai RELEVANT_CATEGORIES
  so music/SFX models (MiniMax Music, Lyria 2, Stable Audio, etc.) appear
- Map new fal categories to text-to-audio capability in mapFalCategory
- Add ElevenLabs Sound Effects v2 model to KIE_MODELS registry
- Add parameter schemas for all 4 ElevenLabs models (turbo-v2.5,
  multilingual-v2, text-to-dialogue-v3, sound-effect-v2) in getKieSchema
- Add model defaults for ElevenLabs TTS and SFX models
- Remap prompt->text field for ElevenLabs models (Kie API requirement)
- Add audio response detection in generateWithKie (capability, URL
  extension, and content-type checks) so audio output is correctly
  returned as type "audio" instead of being misclassified as "image"

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/app/api/generate/providers/kie.ts

@@ -133,6 +133,21 @@ export function getKieModelDefaults(modelId: string): Record<string, unknown> {
         aspect_ratio: "16:9",
       };
 
+    // ElevenLabs TTS models
+    case "elevenlabs/turbo-v2.5":
+    case "elevenlabs/multilingual-v2":
+    case "elevenlabs/text-to-dialogue-v3":
+      return {
+        output_format: "mp3_44100_128",
+      };
+
+    // ElevenLabs Sound Effects
+    case "elevenlabs/sound-effect-v2":
+      return {
+        output_format: "mp3_44100_128",
+        prompt_influence: 0.3,
+      };
+
     default:
       return {};
   }
@@ -600,6 +615,14 @@ export async function generateWithKie(
     };
   }
 
+  // ElevenLabs models use "text" instead of "prompt"
+  if (modelId.startsWith("elevenlabs/")) {
+    if (inputParams.prompt) {
+      inputParams.text = inputParams.prompt;
+      delete inputParams.prompt;
+    }
+  }
+
   // All remaining Kie models use the standard createTask endpoint
   const requestBody: Record<string, unknown> = {
     model: modelId,
@@ -692,6 +715,13 @@ export async function generateWithKie(
   const data = pollResult.data;
   let mediaUrl: string | null = null;
   let isVideo = false;
+  let isAudio = false;
+
+  // Detect audio from model capabilities
+  const isAudioModel = input.model.capabilities.some(c => c.includes("audio"));
+  if (isAudioModel) {
+    isAudio = true;
+  }
 
   console.log(`[API:${requestId}] Kie poll result data:`, JSON.stringify(data).substring(0, 500));
 
@@ -751,10 +781,15 @@ export async function generateWithKie(
   }
 
   // Detect video from URL if not already detected
-  if (!isVideo && (mediaUrl.includes('.mp4') || mediaUrl.includes('.webm') || mediaUrl.includes('video'))) {
+  if (!isVideo && !isAudio && (mediaUrl.includes('.mp4') || mediaUrl.includes('.webm') || mediaUrl.includes('video'))) {
     isVideo = true;
   }
 
+  // Detect audio from URL if not already detected
+  if (!isVideo && !isAudio && (mediaUrl.includes('.mp3') || mediaUrl.includes('.wav') || mediaUrl.includes('.ogg') || mediaUrl.includes('.flac'))) {
+    isAudio = true;
+  }
+
   // Validate URL before fetching
   const mediaUrlCheck = validateMediaUrl(mediaUrl);
   if (!mediaUrlCheck.valid) {
@@ -778,10 +813,13 @@ export async function generateWithKie(
     return { success: false, error: `Media too large: ${(mediaContentLength / (1024 * 1024)).toFixed(0)}MB > 500MB limit` };
   }
 
-  const contentType = mediaResponse.headers.get("content-type") || (isVideo ? "video/mp4" : "image/png");
+  const contentType = mediaResponse.headers.get("content-type") || (isVideo ? "video/mp4" : isAudio ? "audio/mpeg" : "image/png");
   if (contentType.startsWith("video/")) {
     isVideo = true;
   }
+  if (contentType.startsWith("audio/")) {
+    isAudio = true;
+  }
 
   const mediaArrayBuffer = await mediaResponse.arrayBuffer();
   if (mediaArrayBuffer.byteLength > MAX_MEDIA_SIZE) {
@@ -792,6 +830,20 @@ export async function generateWithKie(
 
   console.log(`[API:${requestId}] Output: ${contentType}, ${mediaSizeMB.toFixed(2)}MB`);
 
+  // For audio models, return base64 encoded audio
+  if (isAudio) {
+    const audioBase64 = Buffer.from(mediaArrayBuffer).toString("base64");
+    console.log(`[API:${requestId}] SUCCESS - Returning audio`);
+    return {
+      success: true,
+      outputs: [{
+        type: "audio",
+        data: `data:${contentType};base64,${audioBase64}`,
+        url: mediaUrl,
+      }],
+    };
+  }
+
   // For very large videos (>20MB), return URL only (data left empty for consumers)
   if (isVideo && mediaSizeMB > 20) {
     console.log(`[API:${requestId}] SUCCESS - Returning URL for large video`);

src/app/api/models/[modelId]/route.ts

@@ -609,6 +609,42 @@ function getKieSchema(modelId: string): ExtractedSchema {
         { name: "image_urls", type: "image", required: true, label: "Image", isArray: true },
       ],
     },
+    // ============ Audio/TTS models ============
+    "elevenlabs/turbo-v2.5": {
+      parameters: [
+        { name: "voice_id", type: "string", description: "Voice ID to use for synthesis" },
+        { name: "stability", type: "number", description: "Voice stability (0-1)", default: 0.5, minimum: 0, maximum: 1 },
+        { name: "similarity_boost", type: "number", description: "Similarity boost (0-1)", default: 0.75, minimum: 0, maximum: 1 },
+        { name: "output_format", type: "string", description: "Audio output format", enum: ["mp3_44100_128", "mp3_44100_192", "pcm_16000", "pcm_22050", "pcm_24000", "pcm_44100"], default: "mp3_44100_128" },
+      ],
+      inputs: [{ name: "prompt", type: "text", required: true, label: "Text" }],
+    },
+    "elevenlabs/multilingual-v2": {
+      parameters: [
+        { name: "voice_id", type: "string", description: "Voice ID to use for synthesis" },
+        { name: "stability", type: "number", description: "Voice stability (0-1)", default: 0.5, minimum: 0, maximum: 1 },
+        { name: "similarity_boost", type: "number", description: "Similarity boost (0-1)", default: 0.75, minimum: 0, maximum: 1 },
+        { name: "output_format", type: "string", description: "Audio output format", enum: ["mp3_44100_128", "mp3_44100_192", "pcm_16000", "pcm_22050", "pcm_24000", "pcm_44100"], default: "mp3_44100_128" },
+      ],
+      inputs: [{ name: "prompt", type: "text", required: true, label: "Text" }],
+    },
+    "elevenlabs/text-to-dialogue-v3": {
+      parameters: [
+        { name: "stability", type: "number", description: "Voice stability (0-1)", default: 0.5, minimum: 0, maximum: 1 },
+        { name: "similarity_boost", type: "number", description: "Similarity boost (0-1)", default: 0.75, minimum: 0, maximum: 1 },
+        { name: "output_format", type: "string", description: "Audio output format", enum: ["mp3_44100_128", "mp3_44100_192", "pcm_16000", "pcm_22050", "pcm_24000", "pcm_44100"], default: "mp3_44100_128" },
+      ],
+      inputs: [{ name: "prompt", type: "text", required: true, label: "Text / Dialogue Script" }],
+    },
+    "elevenlabs/sound-effect-v2": {
+      parameters: [
+        { name: "duration_seconds", type: "number", description: "Duration in seconds (0.5-22)", minimum: 0.5, maximum: 22 },
+        { name: "loop", type: "boolean", description: "Enable smooth looping", default: false },
+        { name: "prompt_influence", type: "number", description: "How closely to follow the prompt (0-1)", default: 0.3, minimum: 0, maximum: 1 },
+        { name: "output_format", type: "string", description: "Audio output format", enum: ["mp3_44100_128", "mp3_44100_192", "pcm_16000", "pcm_22050", "pcm_24000", "pcm_44100"], default: "mp3_44100_128" },
+      ],
+      inputs: [{ name: "prompt", type: "text", required: true, label: "Sound Description" }],
+    },
     // ============ Video models ============
     "grok-imagine/text-to-video": {
       parameters: [

src/app/api/models/route.ts

@@ -52,6 +52,8 @@ const RELEVANT_CATEGORIES = [
   "text-to-3d",
   "image-to-3d",
   "text-to-speech",
+  "text-to-music",
+  "text-to-sound-effects",
 ];
 
 // Kie.ai models (hardcoded - no discovery API available)
@@ -310,7 +312,7 @@ const KIE_MODELS: ProviderModel[] = [
     coverImage: undefined,
     pageUrl: "https://docs.kie.ai/veo3-api/quickstart",
   },
-  // ============ Audio/TTS Models (3) ============
+  // ============ Audio/TTS Models (4) ============
   {
     id: "elevenlabs/turbo-v2.5",
     name: "ElevenLabs Turbo v2.5",
@@ -341,6 +343,16 @@ const KIE_MODELS: ProviderModel[] = [
     pricing: { type: "per-run", amount: 0.06, currency: "USD" },
     pageUrl: "https://kie.ai/elevenlabs/text-to-dialogue-v3",
   },
+  {
+    id: "elevenlabs/sound-effect-v2",
+    name: "ElevenLabs Sound Effects v2",
+    description: "Generate sound effects from text descriptions. Supports looping, 0.5-22 second duration, and multiple output formats via Kie.ai.",
+    provider: "kie",
+    capabilities: ["text-to-audio"],
+    coverImage: undefined,
+    pricing: { type: "per-run", amount: 0.02, currency: "USD" },
+    pageUrl: "https://kie.ai/elevenlabs-sound-effect",
+  },
 ];
 
 // Gemini image models (hardcoded - these don't come from an external API)
@@ -759,6 +771,12 @@ function mapFalCategory(category: string): ModelCapability | null {
   if (category === "text-to-speech") {
     return "text-to-audio";
   }
+  if (category === "text-to-music") {
+    return "text-to-audio";
+  }
+  if (category === "text-to-sound-effects") {
+    return "text-to-audio";
+  }
   if (RELEVANT_CATEGORIES.includes(category)) {
     return category as ModelCapability;
   }