@picsart/ai-sdk 5.12.0 → 5.14.0

package/index.d.ts

@@ -615,6 +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<{
@@ -622,15 +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;
         }>;
@@ -1313,7 +1310,7 @@ type ModelInputById = {
         prompt: string;
         aspectRatio?: "16:9" | "9:16" | "1:1" | "4:3" | "3:4" | "21:9" | "adaptive";
         resolution?: "480p" | "720p" | "1080p";
-        duration?: 4 | 5 | 6 | 8 | 10 | 12 | 15 | 20 | 25 | 30;
+        duration?: number;
         generateAudio?: boolean;
         returnLastFrame?: boolean;
         outputFormat?: "mp4" | "mov";
@@ -1337,7 +1334,7 @@ type ModelInputById = {
         prompt: string;
         aspectRatio?: "adaptive";
         resolution?: "480p" | "720p" | "1080p";
-        duration?: 4 | 5 | 6 | 8 | 10 | 12 | 15 | 20 | 25 | 30;
+        duration?: number;
         generateAudio?: boolean;
         outputFormat?: "mp4" | "mov";
         videoUrls: [string, ...string[]];
@@ -1879,7 +1876,10 @@ interface ModelParamsAccessor {
     file(key: string): FileEntry | undefined;
     prompt(): TextEntry | undefined;
     aspectRatio(): EnumEntry | undefined;
-    duration(): EnumEntry | undefined;
+    /** Duration is an enum on models with a fixed option list and a range on
+     *  models whose vendor accepts every value in a span (kling-t2a,
+     *  seedance-2.5), so narrow on `.kind` before reading `.options` / `.min`. */
+    duration(): EnumEntry | RangeEntry | undefined;
     resolution(): EnumEntry | undefined;
     generateAudio(): BooleanEntry | undefined;
     startFrame(): FileEntry | undefined;

package/index.js

@@ -846,6 +846,17 @@ var p = {
       }
     };
   },
+  /** Continuous duration (seconds) — for models whose vendor accepts every
+   *  value in a span instead of a fixed option list. Reach for this only when
+   *  the span holds more than 10 possible values; 10 or fewer stays an enum
+   *  (`duration`), which shows the exact options instead of a slider. */
+  durationRange(min, max, def, step = 1) {
+    return {
+      duration: {
+        descriptor: { kind: "range", min, max, step, default: def }
+      }
+    };
+  },
   resolution(opts, def) {
     return {
       resolution: {
@@ -1216,6 +1227,7 @@ var params = {
   prompt: p.prompt,
   aspectRatio: p.aspectRatio,
   duration: p.duration,
+  durationRange: p.durationRange,
   count: p.count,
   resolution: p.resolution,
   negativePrompt: p.negativePrompt,
@@ -1318,6 +1330,20 @@ 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 +1431,6 @@ 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 +1542,11 @@ 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 +1582,12 @@ 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 +1595,24 @@ 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 +2017,6 @@ 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 +2057,32 @@ 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" } : {}
   };
@@ -3852,7 +3863,7 @@ var buildSeedance25VideoExtendPayload = (ctx) => ({
 });
 var SEEDANCE_AR = ["16:9", "9:16", "1:1", "4:3", "3:4", "21:9", "adaptive"];
 var SEEDANCE_V2_DURATIONS = [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15];
-var SEEDANCE_25_DURATIONS = [4, 5, 6, 8, 10, 12, 15, 20, 25, 30];
+var SEEDANCE_25_DURATION = { min: 4, max: 30 };
 var { MODELS: MODELS12 } = defineModels("seedance", [
   {
     id: "seedance-2.5",
@@ -3872,7 +3883,7 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
       ...params.prompt(),
       ...params.aspectRatio(SEEDANCE_AR),
       ...params.resolution(["480p", "720p", "1080p"], "1080p"),
-      ...params.duration(SEEDANCE_25_DURATIONS, 5),
+      ...params.durationRange(SEEDANCE_25_DURATION.min, SEEDANCE_25_DURATION.max, 5),
       ...params.generateAudio(),
       ...params.returnLastFrame(),
       ...p.enum("outputFormat", ["mp4", "mov"], "mp4", { label: "Format" }),
@@ -3929,7 +3940,7 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
       // duration stays user-selectable.
       ...params.aspectRatio(["adaptive"]),
       ...params.resolution(["480p", "720p", "1080p"], "1080p"),
-      ...params.duration(SEEDANCE_25_DURATIONS, 15),
+      ...params.durationRange(SEEDANCE_25_DURATION.min, SEEDANCE_25_DURATION.max, 15),
       ...params.generateAudio(),
       ...p.enum("outputFormat", ["mp4", "mov"], "mp4", { label: "Format" }),
       ...params.videoInputs(10, "Source Videos", true, void 0, SEEDANCE_25_MAX_VIDEO_BYTES)
@@ -10230,8 +10241,12 @@ var ModelParamsAccessorImpl = class {
   aspectRatio() {
     return this.narrow("aspectRatio", "enum");
   }
+  /** Enum on fixed-option models, range where the vendor accepts every value
+   *  in a span — callers narrow on `.kind`. */
   duration() {
-    return this.narrow("duration", "enum");
+    const entry = this.param("duration");
+    if (!entry || entry.kind !== "enum" && entry.kind !== "range") return void 0;
+    return entry;
   }
   resolution() {
     return this.narrow("resolution", "enum");
@@ -10309,7 +10324,8 @@ var ConstrainedParamsAccessor = class {
     return this.applyEnum("aspectRatio", this.inner.aspectRatio());
   }
   duration() {
-    return this.applyEnum("duration", this.inner.duration());
+    const entry = this.inner.duration();
+    return entry?.kind === "enum" ? this.applyEnum("duration", entry) : this.applyEntry("duration", entry);
   }
   resolution() {
     return this.applyEnum("resolution", this.inner.resolution());

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "5.12.0",
+  "version": "5.14.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",