@shop-os/modelkit 0.7.10 → 0.8.0

package/dist/index.cjs

@@ -4844,6 +4844,8 @@ function range(min, max) {
 }
 var KLING_PROMPT = { referenceSyntax: "numbered", maxChars: 2500 };
 var H3_PROMPT = { referenceSyntax: "numbered", maxChars: 5e4 };
+var KLING_V4_PROMPT = { referenceSyntax: "numbered", maxChars: 2500 };
+var KLING_V4_AUDIO = "always";
 var NO_REFERENCES = { images: 0, videos: 0, audios: 0 };
 function klingMotionControl(tier) {
   const label = tier === "pro" ? "Pro" : "Standard";
@@ -4944,6 +4946,79 @@ var VIDEO_FAMILY_MODELS = [
     },
     capabilities: { acceptsSourceVideo: true, videoEdit: true }
   },
+  {
+    // https://fal.ai/models/kling/v4/reference-to-video/api (+ /text-to-video,
+    // /image-to-video): one row; references (images or videos) go to reference-to-video,
+    // else a start frame to image-to-video, else text-to-video. Reference images: 10, the
+    // start frame included (fal: "counts toward the image reference limit"), 300 px min.
+    // Reference videos: 5, 3-30.05 s each and 30 s combined, 700 px min, 23.9-60.05 fps.
+    // The prompt names references as @Image1, @Video1 (no space; @Element1 for elements,
+    // which the kit does not send). No sound switch on any endpoint.
+    id: "kling-4.0",
+    aliases: ["kling-4", "kling-v4"],
+    label: "Kling 4.0",
+    type: "video",
+    adapter: "fal",
+    providerModelId: "kling/v4",
+    lifecycle: "active",
+    exposure: ["cowork", "space_builder"],
+    constraints: {
+      aspectRatios: ["adaptive", "16:9", "9:16", "1:1", "21:9"],
+      resolutions: ["720p", "1080p"],
+      durations: { min: 3, max: 30 },
+      durationPresets: [5, 10, 15, 30],
+      references: {
+        ...NO_REFERENCES,
+        images: 10,
+        videos: 5,
+        videoSeconds: { min: 3, max: 30.05, total: 30 }
+      },
+      supports: ["lastFrame"],
+      prompt: KLING_V4_PROMPT,
+      audio: KLING_V4_AUDIO
+    },
+    defaults: { durationSeconds: 5, resolution: "1080p", aspectRatio: "16:9" },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "kuaishou",
+      logo: KLING_LOGO,
+      description: "Kuaishou Kling 4.0: text-to-video, start/end frames, or reference-to-video from up to 10 images and 5 videos (30 seconds combined), 3-30 seconds at 720p or 1080p. Prompts name references @Image1, @Video1."
+    },
+    capabilities: { lastFrameRequiresFirstFrame: true }
+  },
+  {
+    // https://fal.ai/models/kling/v4/video-to-video/edit/api: source 3-30.05 s (700 px
+    // min, 23.9-60.05 fps), up to 10 reference images. duration 3-30, or left out for the
+    // source's length. The prompt names the source @Video1 and images @Image1, @Image2.
+    id: "kling-4.0-edit",
+    aliases: ["kling-4-edit", "kling-v4-edit"],
+    label: "Kling 4.0 Edit",
+    type: "video",
+    adapter: "fal",
+    providerModelId: "kling/v4/video-to-video/edit",
+    lifecycle: "active",
+    exposure: ["cowork", "space_builder"],
+    constraints: {
+      resolutions: ["720p", "1080p"],
+      durations: { min: 3, max: 30 },
+      references: { ...NO_REFERENCES, images: 10 },
+      requires: ["sourceVideo"],
+      sourceVideoSeconds: { min: 3, max: 30.05 },
+      supports: [],
+      prompt: KLING_V4_PROMPT,
+      audio: KLING_V4_AUDIO
+    },
+    defaults: { resolution: "1080p" },
+    usage: { ...VIDEO_USAGE },
+    pricingKind: "video_per_second",
+    display: {
+      provider: "kuaishou",
+      logo: KLING_LOGO,
+      description: "Kuaishou Kling 4.0 video edit: change a 3-30 second clip with a prompt and up to 10 reference images, at 720p or 1080p, as long as the source or a set 3-30 seconds. Prompts name the source @Video1 and images @Image1."
+    },
+    capabilities: { acceptsSourceVideo: true, videoEdit: true }
+  },
   {
     // https://fal.ai/models/minimax/h3-max-turbo/text-to-video/api (+ /image-to-video).
     // image-to-video takes a start frame, an end frame, or both, and no aspect_ratio.
@@ -11251,6 +11326,8 @@ var MOTION_PRO = "fal-ai/kling-video/v3/pro/motion-control";
 var MOTION_STANDARD = "fal-ai/kling-video/v3/standard/motion-control";
 var O3_REFERENCE = "fal-ai/kling-video/o3/pro/reference-to-video";
 var O3_EDIT = "fal-ai/kling-video/o3/pro/video-to-video/edit";
+var V4_FAMILY = "kling/v4";
+var V4_EDIT = "kling/v4/video-to-video/edit";
 function motionControl(endpoint, req, ctx) {
   const [character] = req.referenceImages ?? [];
   const [motion] = req.referenceVideos ?? [];
@@ -11310,6 +11387,61 @@ var klingV3Family = {
     return motionControl(id, req, ctx);
   }
 };
+function v4Generate(req, ctx) {
+  const images = urls(req.referenceImages);
+  const videos = urls(req.referenceVideos);
+  const duration = wholeSeconds(req, ctx);
+  const resolution = resolutionOf(req, ctx);
+  const input = { prompt: req.prompt.trim(), duration, resolution };
+  let endpoint;
+  if (images.length + videos.length > 0) {
+    endpoint = `${V4_FAMILY}/reference-to-video`;
+    if (images.length > 0) input["image_urls"] = images;
+    if (videos.length > 0) input["video_urls"] = videos;
+    if (req.firstFrame) input["start_image_url"] = refToUrl(req.firstFrame);
+    if (req.aspectRatio !== void 0) {
+      input["aspect_ratio"] = req.aspectRatio === "adaptive" ? "auto" : req.aspectRatio;
+    }
+  } else if (req.firstFrame) {
+    endpoint = `${V4_FAMILY}/image-to-video`;
+    input["start_image_url"] = refToUrl(req.firstFrame);
+    if (req.lastFrame) input["end_image_url"] = refToUrl(req.lastFrame);
+  } else {
+    endpoint = `${V4_FAMILY}/text-to-video`;
+    const ratio = concreteRatio(req);
+    if (ratio !== void 0) input["aspect_ratio"] = ratio;
+  }
+  return { endpoint, input, resolution, seconds: requestedSeconds(duration) };
+}
+function v4Edit(req, ctx) {
+  if (req.sourceVideo === void 0) throw refusal(ctx, "needs a source video", "missing_input");
+  const resolution = resolutionOf(req, ctx);
+  const input = {
+    prompt: req.prompt.trim(),
+    video_url: refToUrl(req.sourceVideo),
+    resolution
+  };
+  const images = urls(req.referenceImages);
+  if (images.length > 0) input["image_urls"] = images;
+  if (req.durationSeconds === void 0) {
+    return {
+      endpoint: V4_EDIT,
+      input,
+      resolution,
+      seconds: inputLengthSeconds(req, req.sourceVideo)
+    };
+  }
+  const duration = wholeSeconds(req, ctx);
+  input["duration"] = duration;
+  return { endpoint: V4_EDIT, input, resolution, seconds: requestedSeconds(duration) };
+}
+var KLING_V4_VIDEO_IDS = /* @__PURE__ */ new Set([V4_FAMILY, V4_EDIT]);
+var klingV4Family = {
+  ids: KLING_V4_VIDEO_IDS,
+  plan(req, ctx) {
+    return ctx.model.providerModelId.toLowerCase() === V4_EDIT ? v4Edit(req, ctx) : v4Generate(req, ctx);
+  }
+};
 
 // src/replicate-video-kinds.ts
 function replicateBaseModelId(providerModelId) {
@@ -11380,6 +11512,8 @@ var MOTION_PRO2 = "fal-ai/kling-video/v3/pro/motion-control";
 var MOTION_STANDARD2 = "fal-ai/kling-video/v3/standard/motion-control";
 var O3_REFERENCE2 = "fal-ai/kling-video/o3/pro/reference-to-video";
 var O3_EDIT2 = "fal-ai/kling-video/o3/pro/video-to-video/edit";
+var KLING_V4 = "kling/v4";
+var KLING_V4_EDIT = "kling/v4/video-to-video/edit";
 var H3_TURBO = "minimax/h3-max-turbo";
 var H3_RECAST = "minimax/h3-max/recast";
 var H3_RELIGHT = "minimax/h3-max/relight";
@@ -11388,6 +11522,7 @@ var HAPPY_HORSE_EDIT = "alibaba/happy-horse/video-edit";
 var VIDU_Q4 = "fal-ai/vidu/q4";
 var OMNIHUMAN_V2 = "bytedance/omnihuman/v2";
 var KLING_O3_IMAGE_CAP = 7;
+var KLING_V4_IMAGE_CAP = 10;
 var TURBO_AUDIO_MIN_SECONDS = 2;
 var OMNIHUMAN_AUDIO_SECONDS = { min: 4, max: 300 };
 function missing(message) {
@@ -11443,6 +11578,23 @@ function o3Reference2(req) {
   }
   return needsPrompt(req);
 }
+function hasReferences(req) {
+  return count(req.referenceImages) + count(req.referenceVideos) > 0;
+}
+function klingV4(req) {
+  if (hasReferences(req)) {
+    if (req.lastFrame !== void 0) {
+      return unsupported2("takes an end frame only without reference images or videos");
+    }
+    const images = count(req.referenceImages) + (req.firstFrame === void 0 ? 0 : 1);
+    if (images > KLING_V4_IMAGE_CAP) {
+      return unsupported2(
+        `takes up to ${KLING_V4_IMAGE_CAP} images in all, the start frame included; this request has ${images}`
+      );
+    }
+  }
+  return needsPrompt(req);
+}
 function h3Turbo(req) {
   if (count(req.referenceImages) + count(req.referenceVideos) > 0) {
     return unsupported2("takes no reference images or videos");
@@ -11521,6 +11673,8 @@ var RULES = /* @__PURE__ */ new Map([
   [O3_REFERENCE2, o3Reference2],
   [KLING_O3_OMNI, o3Reference2],
   [O3_EDIT2, (req) => needsSource(req) ?? needsPrompt(req)],
+  [KLING_V4, klingV4],
+  [KLING_V4_EDIT, (req) => needsSource(req) ?? needsPrompt(req)],
   [H3_TURBO, h3Turbo],
   [MINIMAX_H3_INSERT, h3Insert],
   [H3_RECAST, h3Recast],
@@ -11570,6 +11724,19 @@ function applyFalVideoRules(model, req, callerRatio, warnings) {
     warnings.push(viduSilentWarning(model));
     req.generateAudio = false;
   }
+  if (id === KLING_V4 && req.firstFrame !== void 0 && !hasReferences(req)) {
+    if (req.aspectRatio !== void 0) {
+      if (callerRatio !== void 0) {
+        warnings.push({
+          code: "ratio_dropped",
+          field: "aspectRatio",
+          from: req.aspectRatio,
+          message: `${model.id} follows the start frame's aspect ratio`
+        });
+      }
+      delete req.aspectRatio;
+    }
+  }
   const endFrameOnly = req.lastFrame !== void 0 && req.firstFrame === void 0;
   if (id === H3_TURBO && endFrameOnly && req.aspectRatio !== void 0) {
     if (callerRatio !== void 0 && callerRatio !== "adaptive") {
@@ -11746,6 +11913,7 @@ var viduFamily = { ids: VIDU_MODEL_IDS, plan: plan3 };
 var FAMILIES2 = [
   klingV3Family,
   klingOmniFamily,
+  klingV4Family,
   minimaxH3MaxFamily,
   happyHorseFamily,
   viduFamily,
@@ -12032,8 +12200,8 @@ async function collectFalImages(payload, ctx, req, label) {
   };
 }
 async function generateSpecImage(req, ctx, deps, spec) {
-  const hasReferences = (req.references ?? []).length > 0;
-  const endpoint = hasReferences && spec.spec.editEndpoint !== void 0 ? spec.spec.editEndpoint : ctx.model.providerModelId;
+  const hasReferences2 = (req.references ?? []).length > 0;
+  const endpoint = hasReferences2 && spec.spec.editEndpoint !== void 0 ? spec.spec.editEndpoint : ctx.model.providerModelId;
   const payload = await runJob(ctx, deps, endpoint, spec.input);
   return collectFalImages(payload, ctx, req, `fal ${endpoint}`);
 }
@@ -14378,8 +14546,8 @@ function imageBody(req, ctx, outputFormat) {
 function videoBody(req, ctx, mode) {
   const prompt = checkPrompt(req, ctx);
   const model = lumaModel(ctx);
-  const hasReferences = (req.referenceImages ?? []).length + (req.referenceVideos ?? []).length + (req.referenceAudios ?? []).length > 0;
-  if (hasReferences) {
+  const hasReferences2 = (req.referenceImages ?? []).length + (req.referenceVideos ?? []).length + (req.referenceAudios ?? []).length > 0;
+  if (hasReferences2) {
     refuseOrDrop(
       ctx,
       "referenceImages",
@@ -17201,7 +17369,7 @@ function keysFromEnv(env, prefix = "") {
   set("deepseek", pick2("DEEPSEEK_API_KEY"));
   set("moonshot", pick2("MOONSHOT_API_KEY"));
   set("minimax", pick2("MINIMAX_API_KEY"));
-  set("ark", pick2("ARK_API_KEY", "BYTEPLUS_ARK_API_KEY", "BYTEDANCE_API_KEY"));
+  set("ark", pick2("BYTEDANCE_API_KEY"));
   set("fal", pick2("FAL_KEY", "FAL_API_KEY"));
   set("replicate", pick2("REPLICATE_API_TOKEN", "REPLICATE_API_KEY"));
   set("prodia", pick2("PRODIA_API_KEY", "PRODIA_TOKEN"));