@picsart/ai-sdk 5.11.0 → 5.13.0

package/index.d.ts

@@ -68,6 +68,9 @@ type ModelInputById = {
         prompt?: string;
         imageUrls: [string, ...string[]];
         audioUrl: string;
+        resolution?: "720p" | "1080p";
+        turboMode?: boolean;
+        seed?: number;
     };
     "bytedance-video-enhance": {
         videoUrl: string;
@@ -612,6 +615,12 @@ type ModelInputById = {
         resolution?: "720p" | "1080p" | "4k";
         renderingSpeed?: "std" | "pro";
         generateAudio?: boolean;
+        startFrame?: string;
+        endFrame?: string;
+        imageUrls?: string[];
+        videoUrl?: string;
+        referType?: "feature" | "base";
+        keepOriginalSound?: "yes" | "no";
         multiShot?: boolean;
         shotType?: "customize";
         multiPrompt?: Array<{
@@ -619,15 +628,6 @@ type ModelInputById = {
             prompt: string;
             duration: string;
         }>;
-        omniImageList?: Array<{
-            image_url: string;
-            type?: "first_frame" | "end_frame";
-        }>;
-        omniVideoList?: Array<{
-            video_url: string;
-            refer_type: "feature" | "base";
-            keep_original_sound: "yes" | "no";
-        }>;
         elementList?: Array<{
             element_id: string;
         }>;

package/index.js

@@ -1318,6 +1318,20 @@ var klingKeepOriginalSound = {
     }
   }
 };
+var klingOmniReferType = {
+  referType: {
+    label: "Reference Video Mode",
+    descriptor: {
+      kind: "enum",
+      valueType: "string",
+      options: [
+        { id: "feature", label: "Feature Reference" },
+        { id: "base", label: "Base Edit" }
+      ],
+      default: "feature"
+    }
+  }
+};
 var klingV3AdvancedParams = {
   multiShot: {
     label: "Multi-Shot Mode",
@@ -1405,54 +1419,6 @@ var klingOmniAdvancedParams = {
       }
     }
   },
-  omniImageList: {
-    label: "Reference Images",
-    descriptor: {
-      kind: "object",
-      array: { max: 10 },
-      fields: {
-        image_url: { kind: "text" },
-        type: {
-          kind: "enum",
-          required: false,
-          valueType: "string",
-          options: [
-            { id: "first_frame", label: "First Frame" },
-            { id: "end_frame", label: "End Frame" }
-          ],
-          default: "first_frame"
-        }
-      }
-    }
-  },
-  omniVideoList: {
-    label: "Reference Video",
-    descriptor: {
-      kind: "object",
-      array: { max: 1 },
-      fields: {
-        video_url: { kind: "text" },
-        // refer_type / keep_original_sound stay required — upstream
-        // ReferenceVideo marks both required. Descriptor defaults are
-        // informational only until upstream relaxes the wire contract.
-        refer_type: {
-          kind: "enum",
-          valueType: "string",
-          options: [
-            { id: "feature", label: "Feature Reference" },
-            { id: "base", label: "Base Edit" }
-          ],
-          default: "feature"
-        },
-        keep_original_sound: {
-          kind: "enum",
-          valueType: "string",
-          options: [{ id: "yes", label: "Yes" }, { id: "no", label: "No" }],
-          default: "yes"
-        }
-      }
-    }
-  },
   elementList: {
     label: "Element References",
     descriptor: {
@@ -1564,7 +1530,11 @@ var { MODELS } = defineModels("kling", [
     badge: ["popular", "premium"],
     description: "Long-form video up to 15s with native audio and start/end frame control.",
     constraints: [
-      { when: { renderingSpeed: { is: "std" } }, then: { endFrame: { disabled: true, reason: "End frame requires Pro or 4K mode." } } }
+      { when: { renderingSpeed: { is: "std" } }, then: { endFrame: { disabled: true, reason: "End frame requires Pro or 4K mode." } } },
+      // Backend: `voice_list` requires `sound=on`. Two rules because the
+      // `is` operator does not match an unset value (see core/constraints.ts).
+      { when: { generateAudio: { is: false } }, then: { voiceList: { disabled: true, reason: "Voice references require generated audio." } } },
+      { when: { generateAudio: { exists: false } }, then: { voiceList: { disabled: true, reason: "Voice references require generated audio." } } }
     ]
   },
   // ── Video: Kling V3 Turbo (resolution-tiered T2V + I2V) ───────────
@@ -1600,7 +1570,12 @@ var { MODELS } = defineModels("kling", [
     mode: "video",
     inputType: "t2v",
     description: "Flexible generation across creative styles using V3 Omni architecture, with optional 4K output.",
-    features: [feat("4K", "resolution"), feat("15 sec", "duration")],
+    features: [feat("Image + Video Input", "input"), feat("4K", "resolution"), feat("15 sec", "duration")],
+    // The omni task accepts reference media on the SAME workflow (no
+    // editWorkflow): `image_list` carries the optional first/end frames plus
+    // plain reference images, `video_list` carries a single reference clip.
+    // They are declared here as real file slots so `hasFileInput()` sees them
+    // and the app renders upload targets; payloads.ts assembles the arrays.
     paramConfig: {
       ...params.prompt({ maxLength: 2500 }),
       ...params.aspectRatio(["16:9", "9:16", "1:1"]),
@@ -1608,8 +1583,24 @@ var { MODELS } = defineModels("kling", [
       ...params.resolution(["720p", "1080p", "4k"], "720p"),
       ...params.renderingSpeed([{ id: "std", label: "Standard" }, { id: "pro", label: "Pro" }], "std"),
       ...params.generateAudio(false),
+      ...params.startFrame("First Frame"),
+      ...params.endFrame("End Frame"),
+      // Backend states no explicit cap for omni `image_list`; 10 mirrors the
+      // omni-image contract. The worker stays the authoritative gate.
+      ...params.imageInput(10, "Reference Images"),
+      ...params.videoInput("Reference Video", "reference", false),
+      ...klingOmniReferType,
+      ...klingKeepOriginalSound,
       ...klingOmniAdvancedParams
-    }
+    },
+    constraints: [
+      // KlingMode: `4k` is incompatible with video_list; the worker also drops
+      // generated sound whenever a reference clip is supplied.
+      { when: { videoUrl: { exists: true } }, then: {
+        resolution: { allowed: ["720p", "1080p"], reason: "4K output is unavailable with a reference video." },
+        generateAudio: { disabled: true, reason: "Kling disables generated sound when a reference video is supplied." }
+      } }
+    ]
   },
   {
     id: "kling-video-o1",
@@ -2014,8 +2005,6 @@ var buildKlingV3Payload = (defaultMode = "std") => (input) => {
     ...hasEndFrame ? { image_tail: input.endFrame } : {},
     ...input.negativePrompt ? { negative_prompt: input.negativePrompt } : {},
     ...hasSound ? { sound: "on" } : {},
-    // WorkflowTypes 1.0.5 still types mode as std/pro, but the backend accepts
-    // the catalog's 4k mode for Kling V3.
     mode,
     ...input.multiShot != null ? { multi_shot: input.multiShot } : {},
     ...input.shotType ? { shot_type: input.shotType } : {},
@@ -2056,22 +2045,32 @@ var buildKlingV26Payload = (input) => {
 };
 var stringElementList = (list) => list?.length ? { element_list: list.map((e) => ({ element_id: String(e.element_id) })) } : {};
 var buildOmniV3 = (input) => {
-  const hasBaseEdit = input.omniVideoList?.some((v) => v.refer_type === "base");
-  const hasReferenceVideo = !!input.omniVideoList?.length;
+  const imageList = [
+    ...input.startFrame ? [{ image_url: input.startFrame, type: "first_frame" }] : [],
+    ...input.endFrame ? [{ image_url: input.endFrame, type: "end_frame" }] : [],
+    ...(input.imageUrls ?? []).map((image_url) => ({ image_url }))
+  ];
+  const videoList = input.videoUrl ? [{
+    video_url: input.videoUrl,
+    refer_type: input.referType ?? "feature",
+    keep_original_sound: input.keepOriginalSound ?? "yes"
+  }] : [];
+  const hasBaseEdit = videoList[0]?.refer_type === "base";
+  const hasReferenceVideo = videoList.length > 0;
   const fourK = input.resolution === "4k" && !hasReferenceVideo;
   const hasSound = !!input.generateAudio && !hasReferenceVideo;
   return {
     ...input.multiShot ? {} : { prompt: input.prompt },
     model_name: "kling-v3-omni",
-    ...hasBaseEdit || input.omniImageList?.[0]?.type === "first_frame" ? {} : { aspect_ratio: input.aspectRatio ?? "16:9" },
+    ...hasBaseEdit || input.startFrame ? {} : { aspect_ratio: input.aspectRatio ?? "16:9" },
     // String(n) is just `string`; wire expects literal union. Narrowing cast.
     ...hasBaseEdit ? {} : { duration: String(input.duration ?? 5) },
     ...fourK ? { mode: "4k" } : input.renderingSpeed ? { mode: input.renderingSpeed } : {},
     ...input.multiShot != null ? { multi_shot: input.multiShot } : {},
     ...input.shotType ? { shot_type: input.shotType } : {},
     ...input.multiPrompt ? { multi_prompt: input.multiPrompt } : {},
-    ...input.omniImageList?.length ? { image_list: input.omniImageList } : {},
-    ...input.omniVideoList?.length ? { video_list: input.omniVideoList } : {},
+    ...imageList.length ? { image_list: imageList } : {},
+    ...videoList.length ? { video_list: videoList } : {},
     ...stringElementList(input.elementList),
     ...hasSound ? { sound: "on" } : {}
   };
@@ -2554,6 +2553,7 @@ var BYTEDANCE_ENHANCE_RESOLUTION_OPTIONS = [
   "4k",
   "8k"
 ];
+var BYTEDANCE_OMNIHUMAN_RESOLUTION_OPTIONS = ["720p", "1080p"];
 var BYTEDANCE_ENHANCE_FPS_OPTIONS = [30, 60, 120];
 var BYTEDANCE_ENHANCE_SCENE_OPTIONS = [
   "common",
@@ -2566,11 +2566,6 @@ var buildBytedanceUpscalerPayload = (ctx) => ({
   video_url: ctx.videoUrl,
   target_resolution: "1080p"
 });
-var buildBytedanceOmnihumanPayload = (ctx) => ({
-  image_url: ctx.imageUrls?.[0],
-  audio_url: ctx.audioUrl,
-  ...ctx.prompt ? { prompt: ctx.prompt } : {}
-});
 var { MODELS: MODELS6 } = defineModels("bytedance", [
   {
     id: "bytedance-video-upscaler",
@@ -2581,6 +2576,7 @@ var { MODELS: MODELS6 } = defineModels("bytedance", [
     estimatedTime: 88,
     mode: "video",
     inputType: "v2v",
+    deprecated: true,
     description: "AI upscale video resolution \u2014 enhance existing footage to 1080p.",
     features: [feat("Video Input", "input"), feat("1080p", "resolution")],
     // Vendor rejects sources already at/above the 1080p target: "The input
@@ -2593,16 +2589,33 @@ var { MODELS: MODELS6 } = defineModels("bytedance", [
     name: "ByteDance OmniHuman",
     addedAt: "2026-02-06",
     workflow: "bytedance/omnihuman/v1.5",
-    buildPayload: buildBytedanceOmnihumanPayload,
-    estimatedTime: 179,
+    // Pricing key of the direct BytePlus Vision AI integration that now serves
+    // this workflow (it replaced the fal.ai proxy). The catalog `id` keeps its
+    // `v1.5` spelling for stability, so the backend key has to be spelled out.
+    // Priced per second of output video, narrowed by `resolution`.
+    modelId: "bytedance-omnihuman-1.5",
+    // The vendor quotes a real-time factor of 23 (720p) / 27 (1080p), so wall
+    // time is driven by the driving audio's length: ~230s for a 10s clip at
+    // 720p. Measured 92-131s for a 1.84s clip — mostly queue and fixed
+    // overhead. This is a representative mid-length figure, not a ceiling.
+    estimatedTime: 250,
     mode: "video",
     inputType: "i2v",
     description: "Animate a portrait with realistic body movement driven by audio.",
-    features: [feat("Image Input", "input"), feat("Audio Input", "audio")],
+    features: [
+      feat("Image Input", "input"),
+      feat("Audio Input", "audio"),
+      feat("1080p", "resolution")
+    ],
     paramConfig: {
       ...params.prompt({ required: false }),
       ...params.imageInput(1, "Portrait Image", true),
-      ...params.audioInput("Audio Track", true)
+      ...params.audioInput("Audio Track", true),
+      ...params.resolution([...BYTEDANCE_OMNIHUMAN_RESOLUTION_OPTIONS], "1080p"),
+      ...p.boolean("turboMode", false, "Turbo Mode"),
+      // -1 (the vendor default) means "pick a random seed"; any positive value
+      // reproduces the same result for identical inputs.
+      ...p.range("seed", -1, 2147483647, -1)
     }
   },
   {
@@ -2616,9 +2629,6 @@ var { MODELS: MODELS6 } = defineModels("bytedance", [
     estimatedTime: 300,
     mode: "video",
     inputType: "v2v",
-    // The pa-bytedance-pluggable-worker MR adding this task is still open, so
-    // the workflow is not deployed. Flip this off once the worker ships.
-    disabled: true,
     description: "Denoise, color-correct and super-resolve existing footage up to 8K, with frame-rate conversion.",
     features: [feat("Video Required", "input"), feat("Up to 8K", "resolution"), feat("Enhance", "quality")],
     paramConfig: {
@@ -2661,8 +2671,17 @@ var buildBytedanceVideoEnhancePayload = (input) => {
     ...input.scene && toolVersion === DEFAULT_TOOL_VERSION ? { scene: input.scene } : {}
   };
 };
+var buildBytedanceOmniHumanPayload = (input) => ({
+  image_url: input.imageUrls[0],
+  audio_url: input.audioUrl,
+  resolution: input.resolution ?? "1080p",
+  ...input.prompt ? { prompt: input.prompt } : {},
+  ...input.turboMode ? { turbo_mode: true } : {},
+  ...input.seed != null && input.seed !== -1 ? { seed: input.seed } : {}
+});
 registerPayloads(MODELS6, {
-  "bytedance-video-enhance": buildBytedanceVideoEnhancePayload
+  "bytedance-video-enhance": buildBytedanceVideoEnhancePayload,
+  "bytedance-omnihuman-v1.5": buildBytedanceOmniHumanPayload
 });
 
 // src/vendors/catalog/videography.ts
@@ -5382,6 +5401,7 @@ var fluxResolutions = ["1K", "2K", "4K"];
 var buildFluxV2Payload = (modelId) => (ctx) => ({
   prompt: ctx.prompt,
   model: modelId,
+  count: ctx.count ?? 1,
   imageUrls: ctx.imageUrls ?? [],
   resolution: ctx.resolution ?? "1K",
   aspectRatio: ctx.aspectRatio ?? "1:1",
@@ -5411,17 +5431,18 @@ var buildFluxKontextPayload = (modelId) => (ctx) => {
   return {
     prompt: ctx.prompt,
     model: modelId,
+    count: ctx.count ?? 1,
     imageUrls: ctx.imageUrls ?? [],
     ...aspectRatio ? { aspectRatio } : {}
   };
 };
 var fluxV2Base = {
-  workflow: "flux-v2",
+  workflow: "bfl/v1/flux-2",
   mode: "image",
   inputType: "t2i"
 };
 var fluxKontextBase = {
-  workflow: "flux-kontext",
+  workflow: "bfl/v1/flux-kontext",
   mode: "image",
   inputType: "t2i"
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "5.11.0",
+  "version": "5.13.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",