@koda-sl/baker-cli 0.185.2 → 0.186.0

package/README.md

@@ -2890,7 +2890,7 @@ The generation lane. `baker images` **sources** what already exists; `baker stud
 | `baker studio skills` | Craft directions accepted by `generate --skill <id>` |
 | `baker studio improve <prompt>` | Sharpen a rough brief — the same rewrite the client's prompt-bar wand performs. Returns the rewrite and the original; generates nothing |
 
-**Every per-model limit is enforced before anything is billed.** Prompt character caps (Kling 2,500 · Seedance 4,000 · Recraft 10,000 · gpt-image 32,000), reference-image counts (Recraft 1 · Gemini 14 · gpt-image 16 · Seedance 9 in reference mode), the uint32 seed range, RGB channel bounds, and `width:height` shapes are all checked against the chosen model up front — the request is refused with the number to fix rather than opening a batch that fails a take at a time. A knob the model does not have is refused too; only the *shape* snaps to the nearest it renders. `baker studio models` reads all of this from the same contract table the backend validates against, so it cannot drift.
+**Every per-model limit is enforced before anything is billed.** Prompt character caps (Kling 2,500 · Seedance 2.0 4,000 · Recraft 10,000 · gpt-image 32,000 · Seedance 2.5 uncapped), reference counts (Recraft 1 · Gemini 14 · gpt-image 16 · Seedance 2.0 9 · Seedance 2.5 30 images plus 10 source clips), the uint32 seed range, RGB channel bounds, and `width:height` shapes are all checked against the chosen model up front — the request is refused with the number to fix rather than opening a batch that fails a take at a time. A knob the model does not have is refused too; only the *shape* snaps to the nearest it renders. `baker studio models` reads all of this from the same contract table the backend validates against, so it cannot drift.
 
 Two of those checks run in the CLI itself, before the request is even sent. The **reference-image count** is one: `--reference` / `--image` accept file paths and URLs, which are uploaded to the media library first, so checking the count only at the backend means paying to ingest and describe every image in a set that was always going to be refused. The **first-frame rule** is the other — a clip started from one exact frame takes exactly one image whatever the model's reference slot allows, and reference mode is refused outright on Veo and Kling, which cannot read references at all.
 
@@ -2980,19 +2980,40 @@ baker studio animate "handheld drift right, steam rising from the cup" \
 baker studio animate "product rotates once on a turntable" --image j57abc…,j57def… --from references
 ```
 
-The image fixes the look, so the prompt describes **movement**. Models: `bytedance/seedance-2.0` (default), `google/veo-3.1`, `google/veo-3.1-fast`, `kwaivgi/kling-v3.0-pro`.
+The image fixes the look, so the prompt describes **movement**. Models: `google/gemini-omni-flash` (default), `bytedance/seedance-2.0`, `bytedance/seedance-2.5`, `google/veo-3.1`, `google/veo-3.1-fast`, `kwaivgi/kling-v3.0-pro`.
+
+**`bytedance/seedance-2.5` is the long-form, multimodal tier**, and the only model that does either of the two things below. Reach for it when the beat cannot be five seconds, or when the source is footage rather than a still.
+
+```bash
+baker studio animate "she keeps walking, camera stays with her, then she stops and looks up" \
+  --image j57abc… --from references --from-clip j57batch…:0 \
+  --model bytedance/seedance-2.5 --duration 20
+```
+
+- **4–30 seconds** in one render; everything else in the roster stops at 15, and Veo at 8.
+- **`--from-clip batchId:slot`** builds on takes already rendered, and **`--from-video <videoId>`** on anything in the video library (client footage, synced ads). Together up to 10 — this is how a clip is extended, edited, or has its motion carried into a new shot. Both need `--from references`. A library video's first use spends ~1 minute encoding a reusable copy and is refused with "try again shortly"; every use after is instant.
+- **No soundtrack input.** The model reads MP3/WAV reference audio; the only audio Baker stores is Mux's M4A, which it refuses. Nothing to pass it today.
+- **The output shape follows the source clip** — `--aspect-ratio` does not apply when building on one, and building on one **costs about double** the per-second rate (measured; the estimate accounts for it).
+- **`--from text`** renders with no image at all. Every model does text-to-video; this is the only mode where `--image` may be omitted.
+- **Up to 30 reference images**, against 9 on Seedance 2.0.
+- **480p / 720p only** — 1080p and 4K are real Seedance tiers but not on this route.
+- **Sound costs nothing extra** here, unlike every other model that offers it.
+- **No watermark** unless `--watermark` asks for one; it cannot be removed after the render.
+- **No people from an uploaded picture.** A privacy pre-check refuses any recognisable face in the input image — invented characters included, and images from ByteDance's own image model included. Clips that hit it are re-rendered automatically on Gemini Omni (reference mode) or Veo (frame mode).
 
 | Flag | Description |
 |---|---|
-| `--image` | **Required.** Comma-separated library image ids / local paths / image URLs; the first is the opening frame |
-| `--from` | `frame` (default — animate the first image) \| `references` (use the images as visual reference). Mutually exclusive at every provider |
-| `--model` | Video model id (default `bytedance/seedance-2.0`) |
+| `--image` | Comma-separated library image ids / local paths / image URLs; the first is the opening frame. **Required unless `--from text`** |
+| `--from` | `frame` (default — animate the first image) \| `references` (use the assets as visual guidance) \| `text` (no image; render the prompt). The first two are mutually exclusive at every provider |
+| `--from-clip` | Studio takes to build on, `batchId:slot` comma-separated (Seedance 2.5, needs `--from references`) |
+| `--from-video` | Library video ids to build on — same 10-clip budget as `--from-clip` |
+| `--model` | Video model id (default `google/gemini-omni-flash`) |
 | `--aspect-ratio` | Default `9:16` |
 | `--duration` | Seconds; per-model set (default 5, snapped onto what the model renders) |
 | `--quality` | Resolution, e.g. `720p` (default) \| `1080p` |
 | `--count` | Takes, 1–3 (default 1). A clip costs ~45x an image |
 | `--sound` | Let the model generate audio. Off by default (billed on top) |
-| `--negative` / `--seed` / `--people` / `--enhance-prompt` / `--conditioning-scale` / `--cfg-scale` | Per-model knobs; refused when the chosen model doesn't advertise them |
+| `--negative` / `--seed` / `--people` / `--enhance-prompt` / `--conditioning-scale` / `--cfg-scale` / `--watermark` | Per-model knobs; refused when the chosen model doesn't advertise them |
 | `--wait` / `--max-wait` | Wait for the takes (default on, 600000 ms) |
 | `--full` | Include settings, references, chat attribution, `createdAt` |
 

package/dist/{chunk-3TXZKSAR.js → chunk-OWDTSPY4.js}

@@ -1124,7 +1124,8 @@ var GPT_IMAGE_PROMPT_MAX = 32e3;
 var RECRAFT_PROMPT_MAX = 1e4;
 var SEEDANCE_PROMPT_MAX = 4e3;
 var KLING_PROMPT_MAX = 2500;
-var VEO_SEED_MAX = 4294967295;
+var UINT32_MAX = 4294967295;
+var VEO_SEED_MAX = UINT32_MAX;
 var GPT_IMAGE_MAX_REFERENCES = 16;
 var GEMINI_IMAGE_MAX_REFERENCES = 14;
 var RECRAFT_MAX_REFERENCES = 1;
@@ -1136,9 +1137,42 @@ var GEMINI_LITE_IMAGE_SIZE = ["1K"];
 var OPENROUTER_IMAGE_QUALITY = ["auto", "low", "medium", "high"];
 var RECRAFT_IMAGE_AR = ["1:1", "4:3", "3:4", "16:9", "9:16"];
 var SEEDANCE_DURATIONS = [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15];
+var SEEDANCE_25_DURATIONS = [
+  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 SEEDANCE_25_ASPECT_RATIOS = ["1:1", "3:4", "9:16", "4:3", "16:9", "21:9"];
+var SEEDANCE_25_RESOLUTIONS = ["480p", "720p"];
+var SEEDANCE_25_MAX_IMAGE_REFERENCES = 30;
+var SEEDANCE_25_MAX_VIDEO_REFERENCES = 10;
+var SEEDANCE_25_MAX_AUDIO_REFERENCES = 10;
 var KLING_DURATIONS = [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15];
 var GEMINI_OMNI_DURATIONS = [3, 4, 5, 6, 7, 8, 9, 10];
-var GEMINI_OMNI_DEFAULT_SECONDS = 10;
 var GEMINI_OMNI_ASPECT_RATIOS = ["16:9", "9:16"];
 var GEMINI_OMNI_IMAGE_MIMES = ["image/png", "image/jpeg", "image/webp"];
 var ELEVENLABS_OUTPUT_FORMATS = [
@@ -1152,6 +1186,8 @@ var ELEVENLABS_OUTPUT_FORMATS = [
 var ELEVENLABS_MAX_TEXT_CHARS = 45454;
 var ELEVENLABS_MAX_MUSIC_LENGTH_MS = 454545;
 var OPENROUTER_IMAGE_MIMES = ["image/png", "image/jpeg", "image/webp", "image/gif"];
+var OPENROUTER_VIDEO_REFERENCE_MIMES = ["video/mp4", "video/quicktime", "video/webm"];
+var OPENROUTER_AUDIO_REFERENCE_MIMES = ["audio/mpeg", "audio/mp3", "audio/wav"];
 var REPLICATE_IMAGE_MIMES = ["image/png", "image/jpeg", "image/webp"];
 var REPLICATE_VIDEO_MIMES = ["video/mp4", "video/webm", "video/quicktime"];
 var DECONSTRUCT_VIDEO_MIMES = ["video/mp4", "video/webm", "video/quicktime"];
@@ -1167,21 +1203,30 @@ var IMAGE_GENERATE_MODELS = [
 var DEFAULT_IMAGE_GENERATE_MODEL = "google/gemini-3.1-flash-image-preview";
 var VIDEO_GENERATE_MODELS = [
   "bytedance/seedance-2.0",
+  "bytedance/seedance-2.5",
   "google/gemini-omni-flash",
   "google/veo-3.1",
   "google/veo-3.1-fast",
   "kwaivgi/kling-v3.0-pro"
 ];
 var DEFAULT_VIDEO_GENERATE_MODEL = "google/gemini-omni-flash";
-var REFERENCE_TO_VIDEO_MODELS = ["bytedance/seedance-2.0", "google/gemini-omni-flash"];
+var REFERENCE_TO_VIDEO_MODELS = [
+  "bytedance/seedance-2.0",
+  "bytedance/seedance-2.5",
+  "google/gemini-omni-flash"
+];
 function supportsReferenceToVideo(model) {
   return REFERENCE_TO_VIDEO_MODELS.includes(model);
 }
 function isGeminiOmniModel(model) {
   return model.startsWith("google/gemini-omni");
 }
+function isSeedance25Model(model) {
+  return model === "bytedance/seedance-2.5";
+}
 var LAST_FRAME_TO_VIDEO_MODELS = [
   "bytedance/seedance-2.0",
+  "bytedance/seedance-2.5",
   "google/veo-3.1",
   "google/veo-3.1-fast",
   "kwaivgi/kling-v3.0-pro"
@@ -1488,6 +1533,49 @@ var MODEL_REGISTRY = {
         generate_audio: { kind: "boolean" }
       }
     },
+    "bytedance/seedance-2.5": {
+      // Seedance 2.5 — the LONG-FORM, MULTIMODAL tier, and the only curated video
+      // model reached through OpenRouter that is not frame-only.
+      //
+      // Routed through OpenRouter's video gateway (provider `seed`), NOT through
+      // Replicate like 2.0 — the gateway carries the multimodal reference surface
+      // 2.5 exists for, and Replicate's 2.0 wrapper only has `reference_images`.
+      // The backend branches on the exact 2.0 id for the Replicate path, so this
+      // model must never be folded into a `startsWith("bytedance/seedance")`
+      // routing test: that is one character away from sending 2.5 to a wrapper
+      // that cannot express half of what it is being asked to do.
+      //
+      // Same upstream likeness filter as 2.0 — ByteDance classifies PIXELS, so an
+      // AI-generated face that reads photographic is refused exactly like a
+      // photograph (content_policy_blocked). The workflow's automatic Veo recovery
+      // therefore covers this model too.
+      //
+      // No published prompt cap on this route, so none is invented: OpenRouter's
+      // request schema declares no `maxLength` and the model is sold on long,
+      // multi-shot briefs. A ceiling we made up would refuse the exact prompts
+      // this tier exists to render.
+      label: "ByteDance Seedance 2.5",
+      inputs: [],
+      optional_inputs: [
+        { kind: "image", mimes: OPENROUTER_IMAGE_MIMES, max: SEEDANCE_25_MAX_IMAGE_REFERENCES },
+        { kind: "video", mimes: OPENROUTER_VIDEO_REFERENCE_MIMES, max: SEEDANCE_25_MAX_VIDEO_REFERENCES },
+        { kind: "audio", mimes: OPENROUTER_AUDIO_REFERENCE_MIMES, max: SEEDANCE_25_MAX_AUDIO_REFERENCES }
+      ],
+      required: ["prompt"],
+      params: {
+        prompt: { kind: "string" },
+        aspect_ratio: { kind: "string", enum: SEEDANCE_25_ASPECT_RATIOS },
+        resolution: { kind: "string", enum: SEEDANCE_25_RESOLUTIONS },
+        duration: { kind: "number", enum: SEEDANCE_25_DURATIONS },
+        seed: { kind: "number", integer: true, min: 0, max: UINT32_MAX },
+        generate_audio: { kind: "boolean" },
+        // A visible AI watermark on an ad creative is a defect, and this is the
+        // only curated model that lets us turn one off: `watermark` is named in
+        // the model's own `allowed_passthrough_parameters`, so it rides as a
+        // top-level body field exactly like Kling's `cfg_scale` does.
+        watermark: { kind: "boolean" }
+      }
+    },
     "google/gemini-omni-flash": {
       // Google Gemini Omni Flash — the DEFAULT video model. Natively multimodal:
       // it reasons over text + images and renders video WITH synchronized audio in
@@ -1763,10 +1851,15 @@ function promptMaxLength(kind, model) {
   const schema = MODEL_REGISTRY[kind]?.[model]?.params.prompt;
   return schema?.kind === "string" ? schema.maxLength : void 0;
 }
-function maxInputReferences(kind, model) {
+function inputSlot(kind, model, inputKind) {
   const spec = MODEL_REGISTRY[kind]?.[model];
-  const slot = spec?.optional_inputs?.find((input) => input.kind === "image") ?? spec?.inputs.find((input) => input.kind === "image");
-  return slot?.max;
+  return spec?.optional_inputs?.find((input) => input.kind === inputKind) ?? spec?.inputs.find((input) => input.kind === inputKind);
+}
+function maxInputSlot(kind, model, inputKind) {
+  return inputSlot(kind, model, inputKind)?.max;
+}
+function maxInputReferences(kind, model) {
+  return maxInputSlot(kind, model, "image");
 }
 function supportedAspectRatios(kind, model) {
   const schema = MODEL_REGISTRY[kind]?.[model]?.params.aspect_ratio;
@@ -1956,6 +2049,14 @@ var DEFAULT_VIDEO_DURATION_S = 5;
 function isSeedanceModel(model) {
   return model.startsWith("bytedance/seedance");
 }
+var SEEDANCE_25_USD_PER_SECOND = {
+  "480p": 0.1075,
+  "720p": 0.2326
+};
+var SEEDANCE_25_FALLBACK_USD_PER_SECOND = 0.2326;
+function seedance25UsdPerSecond(resolution) {
+  return SEEDANCE_25_USD_PER_SECOND[resolution ?? DEFAULT_VIDEO_RESOLUTION] ?? SEEDANCE_25_FALLBACK_USD_PER_SECOND;
+}
 var FALLBACK_USD_PER_SECOND = 0.5;
 function seedanceUsdPerSecond(resolution) {
   return SEEDANCE_USD_PER_SECOND[resolution ?? DEFAULT_VIDEO_RESOLUTION] ?? FALLBACK_USD_PER_SECOND;
@@ -1981,11 +2082,33 @@ var USD_PER_SECOND = {
 };
 var GEMINI_OMNI_USD_PER_1M_VIDEO_TOKENS = 17.5;
 var GEMINI_OMNI_VIDEO_TOKENS_PER_SECOND = 5792;
+var GEMINI_OMNI_WINDOW_SECONDS = 10;
+var GEMINI_OMNI_WINDOW_MARGIN = 1.05;
 var GEMINI_OMNI_USD_PER_SECOND = GEMINI_OMNI_VIDEO_TOKENS_PER_SECOND * GEMINI_OMNI_USD_PER_1M_VIDEO_TOKENS / 1e6;
-function estimateVideoCostUsd({ model, duration, resolution, generateAudio }) {
+var VIDEO_SOURCE_COST_MULTIPLIER = 2.5;
+var FAL_SEEDANCE_USD_PER_1K_TOKENS = 0.0214;
+var SEEDANCE_SOURCE_CLIP_DISCOUNT = 0.6;
+var SEEDANCE_DEAREST_FRAME = {
+  "480p": { width: 992, height: 432 },
+  "720p": { width: 1112, height: 834 }
+};
+function falSeedanceUsdPerSecond(resolution, buildsOnSourceClip = false) {
+  const frame = SEEDANCE_DEAREST_FRAME[resolution ?? DEFAULT_VIDEO_RESOLUTION] ?? SEEDANCE_DEAREST_FRAME["720p"];
+  const tokensPerSecond = (frame?.width ?? 1112) * (frame?.height ?? 834) * 24 / 1024;
+  const rate = tokensPerSecond * FAL_SEEDANCE_USD_PER_1K_TOKENS / 1e3;
+  return buildsOnSourceClip ? rate * SEEDANCE_SOURCE_CLIP_DISCOUNT : rate;
+}
+function estimateVideoCostUsd(input) {
+  const { model, duration, resolution, generateAudio, buildsOnSourceClip } = input;
   const seconds = duration ?? DEFAULT_VIDEO_DURATION_S;
+  const sourceClipPremium = buildsOnSourceClip ? VIDEO_SOURCE_COST_MULTIPLIER : 1;
   if (isGeminiOmniModel(model)) {
-    return (duration ?? GEMINI_OMNI_DEFAULT_SECONDS) * GEMINI_OMNI_USD_PER_SECOND;
+    return GEMINI_OMNI_WINDOW_SECONDS * GEMINI_OMNI_USD_PER_SECOND * GEMINI_OMNI_WINDOW_MARGIN;
+  }
+  if (isSeedance25Model(model)) {
+    const viaOpenRouter = seedance25UsdPerSecond(resolution) * seconds * sourceClipPremium;
+    if (!input.mayRouteToFal) return viaOpenRouter;
+    return Math.max(viaOpenRouter, falSeedanceUsdPerSecond(resolution, buildsOnSourceClip) * seconds);
   }
   if (isSeedanceModel(model)) return seedanceUsdPerSecond(resolution) * seconds;
   const published = USD_PER_SECOND[model];
@@ -8214,6 +8337,7 @@ export {
   DEFAULT_CLIP_DURATION_S,
   nearestClipDuration,
   promptMaxLength,
+  maxInputSlot,
   maxInputReferences,
   nearestSupportedAspectRatio,
   nearestSupportedImageSize,
@@ -8244,4 +8368,4 @@ export {
   defaultRegistry,
   createEngineFromEnv
 };
-//# sourceMappingURL=chunk-3TXZKSAR.js.map
+//# sourceMappingURL=chunk-OWDTSPY4.js.map