From ffa6b2efdf9bcc526c3dc0d5999629eb44eac1fc Mon Sep 17 00:00:00 2001
From: kang <wankang2050@gmail.com>
Date: Tue, 19 May 2026 09:07:59 +0800
Subject: [PATCH] fix: align model provider configuration

---
 .env.local.example           |  4 +--
 RULES.md                     |  4 +--
 src/app/page.tsx             | 13 +++++----
 src/components/PackPanel.tsx |  6 ++--
 src/lib/providers.ts         |  2 +-
 src/lib/types.ts             |  7 +++++
 src/lib/videoProviders.ts    | 54 +++++++++++++++++++++++++++---------
 7 files changed, 64 insertions(+), 26 deletions(-)

diff --git a/.env.local.example b/.env.local.example
index f226301..6baac68 100644
--- a/.env.local.example
+++ b/.env.local.example
@@ -1,10 +1,10 @@
 # GPT 最高规格 API。没填 OPENAI_API_KEY 时图片/素材包生成走 mock 占位图。
 OPENAI_API_KEY=
 GPT_TEXT_MODEL=gpt-5.5
-GPT_IMAGE_MODEL=image-2
+GPT_IMAGE_MODEL=gpt-image-2
 GPT_API_BASE=https://api.openai.com/v1
 
 # 视频生成固定走 Seedance。未配置 Key 时 /api/video/generate 返回 503。
 SEEDANCE_API_KEY=
-SEEDANCE_MODEL=seedance-1-0-pro
+SEEDANCE_MODEL=doubao-seedance-2-0-260128
 SEEDANCE_API_BASE=https://ark.cn-beijing.volces.com/api/v3
diff --git a/RULES.md b/RULES.md
index 6154b01..cfee3db 100644
--- a/RULES.md
+++ b/RULES.md
@@ -23,10 +23,10 @@
 ## 环境变量
 - `OPENAI_API_KEY` — GPT API Key；文本/结构化/图片生成统一走 GPT 最高规格配置
 - `GPT_TEXT_MODEL` — 默认 `gpt-5.5`，用于角色设定等结构化输出
-- `GPT_IMAGE_MODEL` — 默认 `image-2`，用于意向图和三类素材包图片生成
+- `GPT_IMAGE_MODEL` — 默认 `gpt-image-2`，用于意向图和三类素材包图片生成
 - `GPT_API_BASE` — 默认 `https://api.openai.com/v1`
 - `SEEDANCE_API_KEY` — Seedance 视频生成 Key；未配置时视频接口返回 503
-- `SEEDANCE_MODEL` — 默认 `seedance-1-0-pro`
+- `SEEDANCE_MODEL` — 默认 `doubao-seedance-2-0-260128`
 - `SEEDANCE_API_BASE` — 默认 `https://ark.cn-beijing.volces.com/api/v3`
 - 配置位置：`.env.local`（gitignored），参考 `.env.local.example`
 - 图片生成未配置 GPT Key 时回退 mock（SVG 占位图），视频生成不 mock，必须配置 Seedance Key
diff --git a/src/app/page.tsx b/src/app/page.tsx
index aa252c7..98f2515 100644
--- a/src/app/page.tsx
+++ b/src/app/page.tsx
@@ -15,6 +15,7 @@ import type {
   PackKind,
   VideoGenerationResponse,
 } from '@/lib/types';
+import type { VIDEO_TEMPLATES } from '@/lib/templates';
 
 export default function Home() {
   const [sessions, setSessions] = useState<GenSession[]>([]);
@@ -146,22 +147,24 @@ export default function Home() {
     }
   }
 
-  async function handleGenerateVideo(image: GenImage, promptTemplate: string) {
+  async function handleGenerateVideo(image: GenImage, template: typeof VIDEO_TEMPLATES[number]) {
     if (!current || videoLoading) return;
     setVideoLoading(true);
     try {
       const character = current.characterSpec
         ? `${current.characterSpec.name}，${current.characterSpec.oneLiner}`
         : current.prompt;
+      const prompt = template.promptTemplate.replace('{character}', character);
       const r = await fetch('/api/video/generate', {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({
-          prompt: promptTemplate.replace('{character}', character),
+          prompt,
           imageUrl: image.url,
-          duration: 6,
-          ratio: '16:9',
-          resolution: '1080p',
+          duration: template.duration,
+          ratio: template.ratio,
+          generateAudio: true,
+          watermark: false,
         }),
       });
       if (!r.ok) {
diff --git a/src/components/PackPanel.tsx b/src/components/PackPanel.tsx
index b8ae248..a4c49ce 100644
--- a/src/components/PackPanel.tsx
+++ b/src/components/PackPanel.tsx
@@ -266,7 +266,7 @@ function TextTemplateSection() {
 function VideoSection({ videoLoading, primaryImage, onGenerateVideo }: {
   videoLoading: boolean;
   primaryImage: GenImage;
-  onGenerateVideo: (image: GenImage, promptTemplate: string) => void;
+  onGenerateVideo: (image: GenImage, template: typeof VIDEO_TEMPLATES[number]) => void;
 }) {
   const [open, setOpen] = useState(false);
   const [showPromptId, setShowPromptId] = useState<string | null>(null);
@@ -331,7 +331,7 @@ function VideoSection({ videoLoading, primaryImage, onGenerateVideo }: {
                   )}
                 </div>
                 <button
-                  onClick={() => onGenerateVideo(primaryImage, template.promptTemplate)}
+                  onClick={() => onGenerateVideo(primaryImage, template)}
                   disabled={videoLoading}
                   className="self-center btn btn-primary text-[11px] px-3 py-1.5 disabled:opacity-40"
                 >
@@ -401,7 +401,7 @@ export default function PackPanel({
   onGenerate: (image: GenImage, kind: PackKind) => void;
   onGenerateAll: (image: GenImage) => void;
   onLockCharacter: (image: GenImage) => void;
-  onGenerateVideo: (image: GenImage, prompt: string) => void;
+  onGenerateVideo: (image: GenImage, template: typeof VIDEO_TEMPLATES[number]) => void;
 }) {
   const [activeNav, setActiveNav] = useState('pack-patent');
   const selectedImages = session.images.filter(image => image.status === 'selected');
diff --git a/src/lib/providers.ts b/src/lib/providers.ts
index b9b6e1f..0512ee8 100644
--- a/src/lib/providers.ts
+++ b/src/lib/providers.ts
@@ -3,7 +3,7 @@ import type { GenImage } from './types';
 export type Provider = 'mock' | 'gpt';
 
 export const GPT_TEXT_MODEL = process.env.GPT_TEXT_MODEL || 'gpt-5.5';
-export const GPT_IMAGE_MODEL = process.env.GPT_IMAGE_MODEL || 'image-2';
+export const GPT_IMAGE_MODEL = process.env.GPT_IMAGE_MODEL || 'gpt-image-2';
 const GPT_API_BASE = process.env.GPT_API_BASE || 'https://api.openai.com/v1';
 
 export function detectProvider(): Provider {
diff --git a/src/lib/types.ts b/src/lib/types.ts
index 23772fd..131ef1d 100644
--- a/src/lib/types.ts
+++ b/src/lib/types.ts
@@ -179,9 +179,16 @@ export type LockCharacterResponse = {
 export type VideoGenerationRequest = {
   prompt: string;
   imageUrl?: string;
+  references?: Array<{
+    type: 'image_url' | 'video_url' | 'audio_url';
+    url: string;
+    role?: 'reference_image' | 'reference_video' | 'reference_audio';
+  }>;
   duration?: number;
   ratio?: '16:9' | '9:16' | '1:1' | '4:3' | '3:4';
   resolution?: '720p' | '1080p';
+  generateAudio?: boolean;
+  watermark?: boolean;
 };
 
 export type VideoGenerationResponse = {
diff --git a/src/lib/videoProviders.ts b/src/lib/videoProviders.ts
index f727c2f..ed9324e 100644
--- a/src/lib/videoProviders.ts
+++ b/src/lib/videoProviders.ts
@@ -1,10 +1,10 @@
 import type { VideoGenerationRequest, VideoGenerationResponse } from './types';
 
-export const SEEDANCE_MODEL = process.env.SEEDANCE_MODEL || 'seedance-1-0-pro';
+export const SEEDANCE_MODEL = process.env.SEEDANCE_MODEL || 'doubao-seedance-2-0-260128';
 const SEEDANCE_API_BASE = process.env.SEEDANCE_API_BASE || 'https://ark.cn-beijing.volces.com/api/v3';
 
 function durationOrDefault(duration?: number): number {
-  return Math.min(Math.max(duration ?? 6, 3), 10);
+  return Math.min(Math.max(duration ?? 6, 3), 15);
 }
 
 function normalizeStatus(status?: string): VideoGenerationResponse['status'] {
@@ -14,15 +14,49 @@ function normalizeStatus(status?: string): VideoGenerationResponse['status'] {
   return 'submitted';
 }
 
+function publicUrlOrUndefined(url?: string): string | undefined {
+  if (!url) return undefined;
+  if (!/^https?:\/\//i.test(url)) return undefined;
+  if (/^https?:\/\/(localhost|127\.0\.0\.1|\[::1\])/i.test(url)) return undefined;
+  return url;
+}
+
+function buildContent(opts: VideoGenerationRequest): Array<Record<string, unknown>> {
+  const content: Array<Record<string, unknown>> = [{ type: 'text', text: opts.prompt.trim() }];
+  const refs = [...(opts.references ?? [])];
+  if (opts.imageUrl) refs.unshift({ type: 'image_url', url: opts.imageUrl, role: 'reference_image' });
+
+  for (const ref of refs) {
+    const url = publicUrlOrUndefined(ref.url);
+    if (!url) continue;
+    if (ref.type === 'image_url') {
+      content.push({ type: 'image_url', image_url: { url }, role: ref.role ?? 'reference_image' });
+    }
+    if (ref.type === 'video_url') {
+      content.push({ type: 'video_url', video_url: { url }, role: ref.role ?? 'reference_video' });
+    }
+    if (ref.type === 'audio_url') {
+      content.push({ type: 'audio_url', audio_url: { url }, role: ref.role ?? 'reference_audio' });
+    }
+  }
+  return content;
+}
+
 export async function generateSeedanceVideo(opts: VideoGenerationRequest): Promise<VideoGenerationResponse> {
   const key = process.env.SEEDANCE_API_KEY;
   if (!key) throw new Error('SEEDANCE_API_KEY missing');
   if (!opts.prompt?.trim()) throw new Error('prompt required');
 
-  const content: Array<Record<string, unknown>> = [{ type: 'text', text: opts.prompt.trim() }];
-  if (opts.imageUrl) {
-    content.push({ type: 'image_url', image_url: { url: opts.imageUrl } });
-  }
+  const content = buildContent(opts);
+  const body: Record<string, unknown> = {
+    model: SEEDANCE_MODEL,
+    content,
+    generate_audio: opts.generateAudio ?? true,
+    ratio: opts.ratio || '16:9',
+    duration: durationOrDefault(opts.duration),
+    watermark: opts.watermark ?? false,
+  };
+  if (opts.resolution) body.resolution = opts.resolution;
 
   const res = await fetch(`${SEEDANCE_API_BASE}/contents/generations/tasks`, {
     method: 'POST',
@@ -30,13 +64,7 @@ export async function generateSeedanceVideo(opts: VideoGenerationRequest): Promi
       'Content-Type': 'application/json',
       Authorization: `Bearer ${key}`,
     },
-    body: JSON.stringify({
-      model: SEEDANCE_MODEL,
-      content,
-      duration: durationOrDefault(opts.duration),
-      ratio: opts.ratio || '16:9',
-      resolution: opts.resolution || '1080p',
-    }),
+    body: JSON.stringify(body),
   });
 
   if (!res.ok) throw new Error(`Seedance ${res.status}: ${await res.text()}`);