Merge pull request #174 from modelstudioai/feat/update-def-video-model

Switch default video model to wan3.0-video

Author: XXPermanentXX

packages/cli/tests/stress/lib/fixtures.mjs

@@ -178,7 +178,7 @@ export async function ensurePrerequisites(ctx) {
       "video",
       "generate",
       "--model",
-      "happyhorse-1.1-t2v",
+      "wan3.0-video",
       "--prompt",
       "压测前置短视频：海浪与静态远景，无明显人物。",
       "--duration",

packages/cli/tests/stress/lib/suite-fixtures.mjs

@@ -130,7 +130,7 @@ export async function generateCombinedFixtures({ suiteRoot, cliPackage }) {
       "video",
       "generate",
       "--model",
-      "happyhorse-1.1-t2v",
+      "wan3.0-video",
       "--prompt",
       "压测前置短视频：海浪与静态远景，无明显人物。",
       "--duration",

packages/cli/tests/stress/targets/video-i2v.mjs

@@ -16,7 +16,7 @@ const motions = [
 
 export const runStress = defineStressTarget({
   canonical: "video-i2v",
-  defaultModel: "happyhorse-1.1-i2v",
+  defaultModel: "wan3.0-video",
   batchDirPrefix: "video-i2v-batch",
   helpText: "pnpm run test:stress -- video-i2v [--reuse-fixtures] -- --count 5 -c 2",
 

packages/cli/tests/stress/targets/video-ref.mjs

@@ -16,7 +16,7 @@ const prompts = [
 
 export const runStress = defineStressTarget({
   canonical: "video-ref",
-  defaultModel: "happyhorse-1.1-r2v",
+  defaultModel: "wan3.0-video",
   batchDirPrefix: "video-ref-batch",
   helpText: "pnpm run test:stress -- video-ref [--reuse-fixtures] -- --count 5 -c 2",
 

packages/cli/tests/stress/targets/video-t2v.mjs

@@ -45,7 +45,7 @@ const pick = (arr) => arr[Math.floor(Math.random() * arr.length)];
 
 export const runStress = defineStressTarget({
   canonical: "video-t2v",
-  defaultModel: "happyhorse-1.1-t2v",
+  defaultModel: "wan3.0-video",
   batchDirPrefix: "video-t2v-batch",
   helpText: `用法：pnpm run test:stress -- video-t2v -- --concurrency 1 --count 3
 详见 docs/agents/stress-batch-tests.md`,

packages/commands/src/commands/config/shared.ts

@@ -69,7 +69,9 @@ export const UI_BOOLEAN_KEYS = new Set<string>(["telemetry"]);
 export const UI_MODEL_DEFAULTS: Record<string, string> = {
   default_text_model: "qwen3.8-max",
   default_image_model: "qwen-image-3.0",
-  default_video_model: "happyhorse-1.1-t2v",
+  default_video_model: "wan3.0-video",
+  default_image_to_video_model: "wan3.0-video",
+  default_reference_to_video_model: "wan3.0-video",
   default_speech_model: "cosyvoice-v3-flash",
   default_omni_model: "qwen3.5-omni-plus",
 };
@@ -100,8 +102,18 @@ export const UI_MODEL_CATALOG: Record<string, ModelOption[]> = {
     { id: "wanx2.x", role: "image/generate · async series" },
   ],
   default_video_model: [
-    { id: "happyhorse-1.1-t2v", role: "video/generate default · text-to-video" },
+    { id: "wan3.0-video", role: "video/generate · t2v / i2v / r2v default" },
+    { id: "happyhorse-1.1-t2v", role: "video/generate · text-to-video" },
     { id: "happyhorse-1.1-i2v", role: "video/generate · image-to-video" },
+    { id: "happyhorse-1.1-r2v", role: "video/ref · reference-to-video" },
+  ],
+  default_image_to_video_model: [
+    { id: "wan3.0-video", role: "video/generate default · image-to-video" },
+    { id: "happyhorse-1.1-i2v", role: "video/generate · image-to-video" },
+  ],
+  default_reference_to_video_model: [
+    { id: "wan3.0-video", role: "video/ref default · reference-to-video" },
+    { id: "happyhorse-1.1-r2v", role: "video/ref · reference-to-video" },
   ],
   default_speech_model: [
     { id: "cosyvoice-v3-flash", role: "speech/synthesize (TTS) default" },

packages/commands/src/commands/video/generate.ts

@@ -24,9 +24,8 @@ import { BOOL_FLAG_PROMPT_EXTEND_API_DEFAULT, BOOL_FLAG_WATERMARK } from "bailia
 
 export default defineCommand({
   description: {
-    "en-US":
-      "Generate a video from text or image (happyhorse-1.1-t2v / happyhorse-1.1-i2v / wan2.6-t2v)",
-    "zh-CN": "根据文本或图片生成视频（happyhorse-1.1-t2v / happyhorse-1.1-i2v / wan2.6-t2v）",
+    "en-US": "Generate a video from text or image (wan3.0-video / wan2.6-t2v / happyhorse-1.1-i2v)",
+    "zh-CN": "根据文本或图片生成视频（wan3.0-video / wan2.6-t2v / happyhorse-1.1-i2v）",
   },
   auth: "apiKey",
   usageArgs: "--prompt <text> [--image <url>] [flags]",
@@ -35,8 +34,8 @@ export default defineCommand({
       type: "string",
       valueHint: "<model>",
       description: {
-        "en-US": "Model ID (default: happyhorse-1.1-t2v, or happyhorse-1.1-i2v with --image)",
-        "zh-CN": "模型 ID（默认：happyhorse-1.1-t2v；使用 --image 时为 happyhorse-1.1-i2v）",
+        "en-US": "Model ID (default: wan3.0-video)",
+        "zh-CN": "模型 ID（默认：wan3.0-video）",
       },
     },
     prompt: {
@@ -119,6 +118,16 @@ export default defineCommand({
         "zh-CN": "完成后将视频保存到文件",
       },
     },
+    file: {
+      type: "string",
+      valueHint: "<url-or-path>",
+      description: {
+        "en-US":
+          "Reference file URL or local path for file-to-video (wan3.0-video only; mutually exclusive with --image/--last-frame)",
+        "zh-CN":
+          "参考文件 URL 或本地路径，用于文件生视频（仅 wan3.0-video；与 --image/--last-frame 互斥）",
+      },
+    },
     ...ASYNC_FLAG,
     ...CONCURRENT_FLAG,
     pollInterval: {
@@ -159,12 +168,20 @@ export default defineCommand({
     const model =
       flags.model ||
       (flags.image
-        ? settings.defaultImageToVideoModel || "happyhorse-1.1-i2v"
-        : settings.defaultVideoModel || "happyhorse-1.1-t2v");
+        ? settings.defaultImageToVideoModel || "wan3.0-video"
+        : settings.defaultVideoModel || "wan3.0-video");
     const format = detectOutputFormat(settings.output);
 
     const imageUrl = flags.image;
     const lastFrameUrl = flags.lastFrame as string | undefined;
+    const fileUrl = flags.file as string | undefined;
+
+    if (fileUrl && (imageUrl || lastFrameUrl)) {
+      throw new BailianError(
+        "--file is mutually exclusive with --image/--last-frame.",
+        ExitCode.USAGE,
+      );
+    }
 
     // Auto-upload local image file for i2v
     let resolvedImageUrl: string | undefined;
@@ -178,6 +195,18 @@ export default defineCommand({
 
     // kf2v mode: both --image and --last-frame provided.
     const isKf2v = Boolean(resolvedImageUrl && resolvedLastFrameUrl);
+    // 万相 3.x（All-in-One）首尾帧走 media[]；旧 kf2v 仍走 image2video 平铺字段。
+    // 与 video/ref.ts 的 useReferenceAudio 保持同一判定：忽略大小写、覆盖 wan3.x 系列。
+    const isWan30 = /^wan3\./i.test(model);
+
+    if (fileUrl && !isWan30) {
+      throw new BailianError("--file is only supported by wan3.0-video.", ExitCode.USAGE);
+    }
+
+    let resolvedFileUrl: string | undefined;
+    if (fileUrl) {
+      resolvedFileUrl = await ctx.client.uploadFile(fileUrl, model);
+    }
 
     const watermark = resolveWatermark(flags.watermark);
     const promptExtend = resolveBooleanFlag(flags.promptExtend, undefined, "prompt-extend");
@@ -187,16 +216,26 @@ export default defineCommand({
       input: {
         prompt: prompt,
         negative_prompt: flags.negativePrompt || undefined,
-        // kf2v: first+last frame flat fields via image2video endpoint.
+        // wan3.0 kf2v: media[first_frame, last_frame] via video-generation endpoint.
+        // legacy kf2v: first_frame_url/last_frame_url via image2video endpoint.
         // wan2.1~2.6 i2v: flat img_url via video-generation endpoint.
         // wan2.7+ / happyhorse i2v: media[] via video-generation endpoint.
-        ...(isKf2v
-          ? { first_frame_url: resolvedImageUrl, last_frame_url: resolvedLastFrameUrl }
-          : resolvedImageUrl
-            ? /wan[x]?2\.[1-6]/i.test(model)
-              ? { img_url: resolvedImageUrl }
-              : { media: [{ type: "first_frame" as const, url: resolvedImageUrl }] }
-            : {}),
+        ...(resolvedFileUrl
+          ? { media: [{ type: "file" as const, url: resolvedFileUrl }] }
+          : isKf2v
+            ? isWan30
+              ? {
+                  media: [
+                    { type: "first_frame" as const, url: resolvedImageUrl! },
+                    { type: "last_frame" as const, url: resolvedLastFrameUrl! },
+                  ],
+                }
+              : { first_frame_url: resolvedImageUrl, last_frame_url: resolvedLastFrameUrl }
+            : resolvedImageUrl
+              ? /wan[x]?2\.[1-6]/i.test(model)
+                ? { img_url: resolvedImageUrl }
+                : { media: [{ type: "first_frame" as const, url: resolvedImageUrl }] }
+              : {}),
       },
       parameters: {
         resolution: flags.resolution || undefined,
@@ -210,13 +249,29 @@ export default defineCommand({
 
     if (settings.dryRun) {
       let previewBody = body;
-      if (isKf2v) {
+      if (resolvedFileUrl) {
+        previewBody = {
+          ...body,
+          input: {
+            ...body.input,
+            media: [{ type: "file" as const, url: redactDataUri(resolvedFileUrl) }],
+          },
+        };
+      } else if (isKf2v) {
+        const redactedFirst = redactDataUri(resolvedImageUrl ?? "");
+        const redactedLast = redactDataUri(resolvedLastFrameUrl ?? "");
         previewBody = {
           ...body,
           input: {
             ...body.input,
-            first_frame_url: redactDataUri(resolvedImageUrl ?? ""),
-            last_frame_url: redactDataUri(resolvedLastFrameUrl ?? ""),
+            ...(isWan30
+              ? {
+                  media: [
+                    { type: "first_frame" as const, url: redactedFirst },
+                    { type: "last_frame" as const, url: redactedLast },
+                  ],
+                }
+              : { first_frame_url: redactedFirst, last_frame_url: redactedLast }),
           },
         };
       } else if (resolvedImageUrl) {
@@ -243,7 +298,7 @@ export default defineCommand({
       settings,
       () =>
         ctx.client.requestJson<DashScopeAsyncResponse>({
-          path: isKf2v ? image2videoPath() : videoGeneratePath(),
+          path: isKf2v && !isWan30 ? image2videoPath() : videoGeneratePath(),
           method: "POST",
           body,
           async: true,

packages/commands/src/commands/video/ref.ts

@@ -24,8 +24,9 @@ import { BOOL_FLAG_PROMPT_EXTEND_API_DEFAULT, BOOL_FLAG_WATERMARK } from "bailia
 export default defineCommand({
   description: {
     "en-US":
-      "Reference-to-video generation (happyhorse-1.1-r2v / wan2.6-r2v): multi-subject, multi-shot with voice",
-    "zh-CN": "参考生视频（happyhorse-1.1-r2v / wan2.6-r2v）：支持多主体、多镜头和语音",
+      "Reference-to-video generation (wan3.0-video / happyhorse-1.1-r2v / wan2.6-r2v): multi-subject, multi-shot with voice",
+    "zh-CN":
+      "参考生视频（wan3.0-video / happyhorse-1.1-r2v / wan2.6-r2v）：支持多主体、多镜头和语音",
   },
   auth: "apiKey",
   usageArgs: "--prompt <text> --image <url>... [--ref-video <url>...] [flags]",
@@ -34,8 +35,8 @@ export default defineCommand({
       type: "string",
       valueHint: "<model>",
       description: {
-        "en-US": "Model ID (default: happyhorse-1.1-r2v)",
-        "zh-CN": "模型 ID（默认：happyhorse-1.1-r2v）",
+        "en-US": "Model ID (default: wan3.0-video)",
+        "zh-CN": "模型 ID（默认：wan3.0-video）",
       },
     },
     prompt: {
@@ -67,16 +68,20 @@ export default defineCommand({
       type: "array",
       valueHint: "<url>",
       description: {
-        "en-US": "Voice URL for corresponding image (pairs by position)",
-        "zh-CN": "对应图片的语音 URL（按位置配对）",
+        "en-US":
+          "Voice URL for corresponding image (pairs by position). On wan3.0-video emitted as reference_audio (refer to Audio 1, Audio 2 in prompt)",
+        "zh-CN":
+          "对应图片的语音 URL（按位置配对）。在 wan3.0-video 上作为 reference_audio 条目发送（prompt 中用「音频1」「音频2」引用）",
       },
     },
     videoVoice: {
       type: "array",
       valueHint: "<url>",
       description: {
-        "en-US": "Voice URL for corresponding ref-video (pairs by position)",
-        "zh-CN": "对应参考视频的语音 URL（按位置配对）",
+        "en-US":
+          "Voice URL for corresponding ref-video (pairs by position). On wan3.0-video emitted as reference_audio (refer to Audio 1, Audio 2 in prompt)",
+        "zh-CN":
+          "对应参考视频的语音 URL（按位置配对）。在 wan3.0-video 上作为 reference_audio 条目发送（prompt 中用「音频1」「音频2」引用）",
       },
     },
     resolution: {
@@ -182,11 +187,16 @@ export default defineCommand({
     const imageVoices = flags.imageVoice || [];
     const videoVoices = flags.videoVoice || [];
 
-    const model = flags.model || settings.defaultReferenceToVideoModel || "happyhorse-1.1-r2v";
+    const model = flags.model || settings.defaultReferenceToVideoModel || "wan3.0-video";
     const format = detectOutputFormat(settings.output);
 
+    // wan3.0-video emits voice as standalone reference_audio entries (referenced as Audio 1, Audio 2 in prompt);
+    // other models (wan2.7-r2v / happyhorse-1.1-r2v, etc.) keep the legacy reference_voice field attached to each asset.
+    const useReferenceAudio = /^wan3\./i.test(model);
+
     // --- Resolve file URLs (auto-upload local files) ---
     const media: DashScopeVideoRefRequest["input"]["media"] = [];
+    const audioEntries: Array<{ type: "reference_audio"; url: string }> = [];
 
     // Add reference images
     for (let imageIndex = 0; imageIndex < images.length; imageIndex++) {
@@ -199,7 +209,11 @@ export default defineCommand({
       // Pair voice by position
       if (imageVoices[imageIndex]) {
         const resolvedVoice = await ctx.client.uploadFile(imageVoices[imageIndex]!, model);
-        entry.reference_voice = resolvedVoice;
+        if (useReferenceAudio) {
+          audioEntries.push({ type: "reference_audio", url: resolvedVoice });
+        } else {
+          entry.reference_voice = resolvedVoice;
+        }
       }
 
       media.push(entry);
@@ -216,12 +230,23 @@ export default defineCommand({
       // Pair voice by position
       if (videoVoices[videoIndex]) {
         const resolvedVoice = await ctx.client.uploadFile(videoVoices[videoIndex]!, model);
-        entry.reference_voice = resolvedVoice;
+        if (useReferenceAudio) {
+          audioEntries.push({ type: "reference_audio", url: resolvedVoice });
+        } else {
+          entry.reference_voice = resolvedVoice;
+        }
       }
 
       media.push(entry);
     }
 
+    // wan3.0: append reference voices as standalone media entries, ordered image-voice then video-voice (Audio 1, Audio 2, ...)
+    if (useReferenceAudio) {
+      for (const audioEntry of audioEntries) {
+        media.push(audioEntry);
+      }
+    }
+
     // --- Build request body ---
     const promptExtend = resolveBooleanFlag(flags.promptExtend, undefined, "prompt-extend");
     const watermark = resolveWatermark(flags.watermark);

packages/commands/tests/e2e/topic-routes.ts

@@ -62,6 +62,7 @@ export const VIDEO_ROUTES: E2eRouteExports = {
   "video ref": "videoRef",
   "video task get": "videoTaskGet",
   "video download": "videoDownload",
+  "speech synthesize": "speechSynthesize",
 };
 
 export const VISION_ROUTES: E2eRouteExports = {