@koda-sl/baker-cli 0.299.0-dev.932cec6ab → 0.301.0-dev.50ab1c583

package/dist/{chunk-YMHERSTK.js → chunk-SSXER76B.js}

@@ -1211,6 +1211,39 @@ var GEMINI_OMNI_IMAGE_MIMES = ["image/png", "image/jpeg", "image/webp"];
 var GROK_IMAGINE_DURATIONS = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
 var WAN_27_DURATIONS = [2, 3, 4, 5, 6, 7, 8, 9, 10];
 var HAPPYHORSE_DURATIONS = [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15];
+var WAN_30_DURATIONS = [
+  2,
+  3,
+  4,
+  5,
+  6,
+  7,
+  8,
+  9,
+  10,
+  11,
+  12,
+  13,
+  14,
+  15,
+  16,
+  17,
+  18,
+  19,
+  20,
+  21,
+  22,
+  23,
+  24,
+  25,
+  26,
+  27,
+  28,
+  29,
+  30
+];
+var GROK_IMAGINE_15_DURATIONS = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15];
+var RUNWAY_45_DURATIONS = [2, 3, 4, 5, 6, 7, 8, 9, 10];
 var ELEVENLABS_OUTPUT_FORMATS = [
   "mp3_22050_32",
   "mp3_44100_32",
@@ -1248,7 +1281,10 @@ var VIDEO_GENERATE_MODELS = [
   "google/veo-3.1-lite",
   "kwaivgi/kling-v3.0-pro",
   "x-ai/grok-imagine-video",
+  "x-ai/grok-imagine-video-1.5",
   "alibaba/wan-2.7",
+  "alibaba/wan-3.0",
+  "runway/gen-4.5",
   "alibaba/happyhorse-1.1"
 ];
 var DEFAULT_VIDEO_GENERATE_MODEL = "google/gemini-omni-1.1-flash";
@@ -1828,6 +1864,67 @@ var MODEL_REGISTRY = {
         generate_audio: { kind: "boolean" }
       }
     },
+    "x-ai/grok-imagine-video-1.5": {
+      // NOT a replacement for the entry above, and deliberately absent from
+      // `AVATAR_VIDEO_MODELS` and `REFERENCE_TO_VIDEO_MODELS`: OpenRouter describes the
+      // older Grok as "text-, image-, and reference-conditioned" and 1.5 as taking "an
+      // optional starting image". Reference conditioning is the whole reason Grok carries
+      // avatars — it is what holds a face from the identity sheet — so swapping them
+      // would quietly cost every avatar clip its likeness.
+      //
+      // What it adds over the older one: 1080p, and 15s rather than 10.
+      label: "Grok Imagine Video 1.5",
+      inputs: [],
+      optional_inputs: [{ kind: "image", mimes: OPENROUTER_IMAGE_MIMES }],
+      required: ["prompt"],
+      params: {
+        prompt: { kind: "string" },
+        aspect_ratio: { kind: "string", enum: ["16:9", "9:16", "1:1", "4:3", "3:4", "3:2", "2:3"] },
+        resolution: { kind: "string", enum: ["480p", "720p", "1080p"] },
+        duration: { kind: "number", enum: GROK_IMAGINE_15_DURATIONS }
+      }
+    },
+    "alibaba/wan-3.0": {
+      // 2.7's successor on every axis that matters: 2–30s against 2–10, 1080p and 480p
+      // as well as 720p, and it keeps the native audio that made 2.7 the landscape
+      // talking-head option.
+      //
+      // The ratios are what it ADVERTISES. 2.7 advertised five and delivered 16:9 for
+      // all of them — measured, twice, which is why its enum here is a single value. That
+      // was a different model and this one is not gated on its predecessor's lie, but
+      // the first vertical clip anyone renders here is worth looking at before a campaign
+      // is built on it.
+      //
+      // No `last_frame`: 2.7 takes one and 3.0 does not, so a canvas that ends a shot on
+      // a chosen frame stays on 2.7.
+      label: "Alibaba Wan 3.0",
+      inputs: [],
+      optional_inputs: [{ kind: "image", mimes: OPENROUTER_IMAGE_MIMES }],
+      required: ["prompt"],
+      params: {
+        prompt: { kind: "string" },
+        aspect_ratio: { kind: "string", enum: ["16:9", "9:16", "1:1", "4:3", "3:4"] },
+        resolution: { kind: "string", enum: ["480p", "720p", "1080p"] },
+        duration: { kind: "number", enum: WAN_30_DURATIONS },
+        seed: { kind: "number" },
+        generate_audio: { kind: "boolean" }
+      }
+    },
+    "runway/gen-4.5": {
+      // One resolution, two shapes, and silent — `generate_audio: false` in its listing,
+      // so a talking head belongs on Wan or Omni. First frame only, no reference.
+      label: "Runway Gen-4.5",
+      inputs: [],
+      optional_inputs: [{ kind: "image", mimes: OPENROUTER_IMAGE_MIMES }],
+      required: ["prompt"],
+      params: {
+        prompt: { kind: "string" },
+        aspect_ratio: { kind: "string", enum: ["16:9", "9:16"] },
+        resolution: { kind: "string", enum: ["720p"] },
+        duration: { kind: "number", enum: RUNWAY_45_DURATIONS },
+        seed: { kind: "number" }
+      }
+    },
     "alibaba/happyhorse-1.1": {
       // Third reference-capable option: strong identity, correct ratios, no audio.
       // First-frame only (no `last_frame`), so it cannot close a beat on a
@@ -2317,6 +2414,32 @@ var USD_PER_SECOND = {
   },
   // Flat $0.10/s across resolutions and modes, audio included.
   "alibaba/wan-2.7": { silent: 0.1, audio: 0.1 },
+  // 3.0 prices per resolution where 2.7 was flat, and its rate still includes audio —
+  // there is no separate audio SKU, so a talking clip costs what a silent one costs.
+  // `duration_seconds_480p|720p|1080p` on GET /api/v1/videos/models.
+  "alibaba/wan-3.0": {
+    silent: 0.1,
+    audio: 0.1,
+    byResolution: {
+      "480p": { silent: 0.05, audio: 0.05 },
+      "720p": { silent: 0.1, audio: 0.1 },
+      "1080p": { silent: 0.2, audio: 0.2 }
+    }
+  },
+  // `cents_per_video_output_second_*` — 8c, 14c and 25c. Dearer than the older Grok at
+  // every tier, and it reaches 1080p, which that one does not. No audio, so both columns
+  // are the same rate rather than one being wrong.
+  "x-ai/grok-imagine-video-1.5": {
+    silent: 0.14,
+    audio: 0.14,
+    byResolution: {
+      "480p": { silent: 0.08, audio: 0.08 },
+      "720p": { silent: 0.14, audio: 0.14 },
+      "1080p": { silent: 0.25, audio: 0.25 }
+    }
+  },
+  // One flat rate, one resolution: `cents_per_second_output: 12`. Silent.
+  "runway/gen-4.5": { silent: 0.12, audio: 0.12 },
   "alibaba/happyhorse-1.1": {
     silent: 0.1278,
     audio: 0.1278,
@@ -9404,4 +9527,4 @@ export {
   defaultRegistry,
   createEngineFromEnv
 };
-//# sourceMappingURL=chunk-YMHERSTK.js.map
+//# sourceMappingURL=chunk-SSXER76B.js.map