Dev 1574 reference audio support

src/components/form/FileUploadField.tsx

@@ -202,6 +202,17 @@ export function FileUploadField({
             />
           </label>
 
+          {fileCount > 0 && !isImageAccept && (
+            <button
+              type="button"
+              onClick={() => onFileChange(name, null)}
+              className="flex items-center justify-center w-8 h-8 bg-[var(--surface-2)] border border-[var(--border)] rounded hover:border-red-500 hover:bg-red-500/10 transition-colors flex-shrink-0"
+              title="Remove file"
+            >
+              <X className="w-3.5 h-3.5 text-[var(--muted)]" />
+            </button>
+          )}
+
           {isImageAccept && (
             <button
               type="button"

src/lib/endpoint-registry.ts

@@ -268,6 +268,69 @@ export const ENDPOINTS: EndpointDefinition[] = [
     ],
   },
 
+  {
+    id: 'audio2video',
+    name: 'Audio to Video',
+    group: 'video-generation',
+    method: 'POST',
+    path: '/aud2video',
+    description: 'Generate video from audio',
+    contentType: 'multipart',
+    isAsync: true,
+    hasPriceCalc: true,
+    priceCalcPath: '/aud2video/price-calculation',
+    params: [
+      promptParam(),
+      negativePromptParam(),
+      {
+        name: 'audio',
+        label: 'Audio File',
+        type: 'file',
+        required: true,
+        accept: 'audio/*,.mp3,.wav,.ogg,.flac',
+        description: 'Audio to condition the video (mp3/wav/ogg/flac, max 20MB)',
+      },
+      {
+        name: 'first_frame_image',
+        label: 'First Frame Image',
+        type: 'file',
+        required: false,
+        accept: 'image/*',
+        description: 'Optional starting frame image',
+      },
+      {
+        name: 'last_frame_image',
+        label: 'Last Frame Image',
+        type: 'file',
+        required: false,
+        accept: 'image/*',
+        description: 'Optional ending frame image',
+      },
+      modelSelectParam(),
+      widthParam(),
+      heightParam(),
+      guidanceParam(),
+      stepsParam(),
+      {
+        name: 'frames',
+        label: 'Frames',
+        type: 'number',
+        required: false,
+        nullable: true,
+        description: 'Number of frames to generate',
+      },
+      {
+        name: 'fps',
+        label: 'FPS',
+        type: 'number',
+        required: false,
+        nullable: true,
+        description: 'Frames per second',
+      },
+      seedParam(),
+    ],
+  },
+
   {
     id: 'txt2video',
     name: 'Text to Video',
@@ -424,7 +487,7 @@ export const ENDPOINTS: EndpointDefinition[] = [
     method: 'POST',
     path: '/txt2music',
     description: 'Generate music from text description',
-    contentType: 'json',
+    contentType: 'multipart',
     isAsync: true,
     hasPriceCalc: true,
     priceCalcPath: '/txt2music/price-calculation',
@@ -539,6 +602,14 @@ export const ENDPOINTS: EndpointDefinition[] = [
           { value: 'flac', label: 'FLAC' },
         ],
       },
+      {
+        name: 'reference_audio',
+        label: 'Reference Audio',
+        type: 'file',
+        required: false,
+        accept: 'audio/*,.mp3,.wav,.flac,.ogg,.m4a',
+        description: 'Optional reference audio for style transfer (mp3/wav/flac/ogg/m4a, max 10MB)',
+      },
     ],
   },