From e19d0a09c871ca1430dcb54a6e4bf23bd92051ce Mon Sep 17 00:00:00 2001
From: Dawid Wenderski <dawidope@gmail.com>
Date: Mon, 9 Mar 2026 12:48:55 +0100
Subject: [PATCH 1/4] feat: add reference_audio support to txt2music endpoint

Switch contentType to multipart and add optional reference audio file
upload field for style transfer.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 src/lib/endpoint-registry.ts | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/src/lib/endpoint-registry.ts b/src/lib/endpoint-registry.ts
index e13f6c5..8ecffd0 100644
--- a/src/lib/endpoint-registry.ts
+++ b/src/lib/endpoint-registry.ts
@@ -424,7 +424,7 @@ 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 +539,14 @@ 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)',
+      },
     ],
   },
 

From 5881fc24f2267d3efbc831776ca531262858dbf9 Mon Sep 17 00:00:00 2001
From: Dawid Wenderski <dawidope@gmail.com>
Date: Mon, 9 Mar 2026 12:49:32 +0100
Subject: [PATCH 2/4] fix: add clear button for non-image file uploads
 (audio/video)

Image files had remove buttons on preview thumbnails, but non-image
files (audio, video) had no way to clear the selection after upload.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 src/components/form/FileUploadField.tsx | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/src/components/form/FileUploadField.tsx b/src/components/form/FileUploadField.tsx
index 192c7cc..792b801 100644
--- a/src/components/form/FileUploadField.tsx
+++ b/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"

From abd16383a2402974654a274776535eecdd2592ce Mon Sep 17 00:00:00 2001
From: Dawid Wenderski <dawidope@gmail.com>
Date: Mon, 9 Mar 2026 12:57:17 +0100
Subject: [PATCH 3/4] feat: add aud2video (audio to video) endpoint

New endpoint for generating video conditioned on audio input.
Similar to img2video but with required audio file and optional
first/last frame images.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 src/lib/endpoint-registry.ts | 63 ++++++++++++++++++++++++++++++++++++
 1 file changed, 63 insertions(+)

diff --git a/src/lib/endpoint-registry.ts b/src/lib/endpoint-registry.ts
index 8ecffd0..9ea908d 100644
--- a/src/lib/endpoint-registry.ts
+++ b/src/lib/endpoint-registry.ts
@@ -268,6 +268,69 @@ export const ENDPOINTS: EndpointDefinition[] = [
     ],
   },
 
+  {
+    id: 'aud2video',
+    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',

From f5566bef9e6c1d0878aee33ad1cf5cea877b8886 Mon Sep 17 00:00:00 2001
From: Dawid Wenderski <dawidope@gmail.com>
Date: Mon, 9 Mar 2026 13:08:22 +0100
Subject: [PATCH 4/4] fix: match aud2video endpoint id to model inference_type
 'audio2video'

Model filtering uses endpoint.id === inference_type. The API models
report 'audio2video' but endpoint had id 'aud2video'.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 src/lib/endpoint-registry.ts | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/lib/endpoint-registry.ts b/src/lib/endpoint-registry.ts
index 9ea908d..04de839 100644
--- a/src/lib/endpoint-registry.ts
+++ b/src/lib/endpoint-registry.ts
@@ -269,7 +269,7 @@ export const ENDPOINTS: EndpointDefinition[] = [
   },
 
   {
-    id: 'aud2video',
+    id: 'audio2video',
     name: 'Audio to Video',
     group: 'video-generation',
     method: 'POST',