@nodaro/shared 3.56.0 → 3.58.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/{chunk-PFXPAWPQ.cjs → chunk-4UUX62S3.cjs} +12 -12
- package/dist/{chunk-PFXPAWPQ.cjs.map → chunk-4UUX62S3.cjs.map} +1 -1
- package/dist/{chunk-CGZLENU4.cjs → chunk-4ZZ6SVIQ.cjs} +90 -6
- package/dist/chunk-4ZZ6SVIQ.cjs.map +1 -0
- package/dist/{chunk-33P5FI2U.cjs → chunk-5PJC4ZWO.cjs} +6 -6
- package/dist/{chunk-33P5FI2U.cjs.map → chunk-5PJC4ZWO.cjs.map} +1 -1
- package/dist/{chunk-TSKYEREJ.cjs → chunk-5UVBFY2G.cjs} +11 -11
- package/dist/{chunk-TSKYEREJ.cjs.map → chunk-5UVBFY2G.cjs.map} +1 -1
- package/dist/{chunk-2AR3OXAA.js → chunk-6CYS7VRH.js} +89 -7
- package/dist/chunk-6CYS7VRH.js.map +1 -0
- package/dist/{chunk-4O5STV7N.js → chunk-6TFWFYYI.js} +3 -3
- package/dist/{chunk-4O5STV7N.js.map → chunk-6TFWFYYI.js.map} +1 -1
- package/dist/{chunk-DURFRWAW.js → chunk-A5ESI2NY.js} +8 -8
- package/dist/{chunk-DURFRWAW.js.map → chunk-A5ESI2NY.js.map} +1 -1
- package/dist/{chunk-62ZQQA7L.js → chunk-BSENM6HM.js} +3 -3
- package/dist/{chunk-62ZQQA7L.js.map → chunk-BSENM6HM.js.map} +1 -1
- package/dist/{chunk-RX7WII6I.js → chunk-CNSRCZYY.js} +5 -5
- package/dist/{chunk-RX7WII6I.js.map → chunk-CNSRCZYY.js.map} +1 -1
- package/dist/{chunk-2J6NTA5I.js → chunk-DC2QPACA.js} +5 -5
- package/dist/chunk-DC2QPACA.js.map +1 -0
- package/dist/{chunk-HWJE4FNS.cjs → chunk-DCP7JFSE.cjs} +5 -5
- package/dist/{chunk-HWJE4FNS.cjs.map → chunk-DCP7JFSE.cjs.map} +1 -1
- package/dist/{chunk-AZDVL7IM.js → chunk-HEUWAIXG.js} +6 -6
- package/dist/chunk-HEUWAIXG.js.map +1 -0
- package/dist/{chunk-JO3XSUO2.cjs → chunk-I7NG2D6X.cjs} +21 -21
- package/dist/{chunk-JO3XSUO2.cjs.map → chunk-I7NG2D6X.cjs.map} +1 -1
- package/dist/{chunk-FCQYMPUQ.cjs → chunk-IKOKRJDI.cjs} +4 -4
- package/dist/{chunk-FCQYMPUQ.cjs.map → chunk-IKOKRJDI.cjs.map} +1 -1
- package/dist/{chunk-ZYITB47P.js → chunk-IQBIH3AE.js} +3 -3
- package/dist/{chunk-ZYITB47P.js.map → chunk-IQBIH3AE.js.map} +1 -1
- package/dist/{chunk-55FQBILM.js → chunk-JUEFRVHZ.js} +4 -4
- package/dist/{chunk-55FQBILM.js.map → chunk-JUEFRVHZ.js.map} +1 -1
- package/dist/{chunk-4ONS6QRT.js → chunk-K6UMB37V.js} +3 -3
- package/dist/{chunk-4ONS6QRT.js.map → chunk-K6UMB37V.js.map} +1 -1
- package/dist/{chunk-IKP4GOWZ.cjs → chunk-KYHDKKUJ.cjs} +8 -8
- package/dist/{chunk-IKP4GOWZ.cjs.map → chunk-KYHDKKUJ.cjs.map} +1 -1
- package/dist/{chunk-7DGJDISO.js → chunk-L26G7B27.js} +3 -3
- package/dist/{chunk-7DGJDISO.js.map → chunk-L26G7B27.js.map} +1 -1
- package/dist/{chunk-JHTYULY2.js → chunk-M7SMSAOU.js} +4 -4
- package/dist/{chunk-JHTYULY2.js.map → chunk-M7SMSAOU.js.map} +1 -1
- package/dist/{chunk-6GGHM33Z.js → chunk-ODTZV3BW.js} +4 -4
- package/dist/{chunk-6GGHM33Z.js.map → chunk-ODTZV3BW.js.map} +1 -1
- package/dist/{chunk-MROKQP4U.js → chunk-OMUYZKRB.js} +3 -3
- package/dist/{chunk-MROKQP4U.js.map → chunk-OMUYZKRB.js.map} +1 -1
- package/dist/{chunk-AKDJKMAL.cjs → chunk-PK4ZR2SK.cjs} +4 -4
- package/dist/{chunk-AKDJKMAL.cjs.map → chunk-PK4ZR2SK.cjs.map} +1 -1
- package/dist/chunk-QCW6J6BQ.cjs +27 -0
- package/dist/chunk-QCW6J6BQ.cjs.map +1 -0
- package/dist/{chunk-ZXDH3L3P.cjs → chunk-QPBBIRLF.cjs} +40 -39
- package/dist/{chunk-ZXDH3L3P.cjs.map → chunk-QPBBIRLF.cjs.map} +1 -1
- package/dist/{chunk-ILMHITHK.cjs → chunk-QVPIO2NU.cjs} +5 -5
- package/dist/{chunk-ILMHITHK.cjs.map → chunk-QVPIO2NU.cjs.map} +1 -1
- package/dist/{chunk-EWARI452.cjs → chunk-S5VBJJUT.cjs} +46 -46
- package/dist/chunk-S5VBJJUT.cjs.map +1 -0
- package/dist/{chunk-ST4WQ6ND.js → chunk-SEQ5L72K.js} +52 -11
- package/dist/chunk-SEQ5L72K.js.map +1 -0
- package/dist/{chunk-EAKO7PGC.js → chunk-VVGSIWNL.js} +5 -4
- package/dist/{chunk-EAKO7PGC.js.map → chunk-VVGSIWNL.js.map} +1 -1
- package/dist/{chunk-O7DXNX5J.cjs → chunk-VZXRDSWA.cjs} +74 -28
- package/dist/chunk-VZXRDSWA.cjs.map +1 -0
- package/dist/{chunk-WKCFTWLO.cjs → chunk-WBERPAXE.cjs} +6 -6
- package/dist/{chunk-WKCFTWLO.cjs.map → chunk-WBERPAXE.cjs.map} +1 -1
- package/dist/{chunk-H7VMJBNZ.cjs → chunk-WLFT5UVV.cjs} +8 -8
- package/dist/{chunk-H7VMJBNZ.cjs.map → chunk-WLFT5UVV.cjs.map} +1 -1
- package/dist/{chunk-4JODLC2N.js → chunk-ZMH56CLL.js} +3 -3
- package/dist/{chunk-4JODLC2N.js.map → chunk-ZMH56CLL.js.map} +1 -1
- package/dist/{chunk-CLKU3ZIU.cjs → chunk-ZX6RRM6R.cjs} +43 -43
- package/dist/{chunk-CLKU3ZIU.cjs.map → chunk-ZX6RRM6R.cjs.map} +1 -1
- package/dist/entries/apps.cjs +42 -42
- package/dist/entries/apps.js +7 -7
- package/dist/entries/credits.cjs +150 -150
- package/dist/entries/credits.d.cts +1 -1
- package/dist/entries/credits.d.ts +1 -1
- package/dist/entries/credits.js +7 -7
- package/dist/entries/llm-models.d.cts +1 -1
- package/dist/entries/llm-models.d.ts +1 -1
- package/dist/entries/model-catalog.cjs +76 -68
- package/dist/entries/model-catalog.d.cts +2 -2
- package/dist/entries/model-catalog.d.ts +2 -2
- package/dist/entries/model-catalog.js +4 -4
- package/dist/entries/models.cjs +303 -283
- package/dist/entries/models.d.cts +3 -3
- package/dist/entries/models.d.ts +3 -3
- package/dist/entries/models.js +4 -4
- package/dist/entries/nodes.cjs +206 -206
- package/dist/entries/nodes.d.cts +1 -1
- package/dist/entries/nodes.d.ts +1 -1
- package/dist/entries/nodes.js +8 -8
- package/dist/entries/pipeline.cjs +152 -152
- package/dist/entries/pipeline.d.cts +1 -1
- package/dist/entries/pipeline.d.ts +1 -1
- package/dist/entries/pipeline.js +3 -3
- package/dist/entries/ugc.cjs +71 -71
- package/dist/entries/ugc.js +4 -4
- package/dist/entries/video.cjs +117 -117
- package/dist/entries/video.js +4 -4
- package/dist/index.cjs +1093 -1065
- package/dist/index.d.cts +2 -2
- package/dist/index.d.ts +2 -2
- package/dist/index.js +17 -17
- package/dist/{model-catalog-vQ3NWn24.d.cts → model-catalog-CCsA1isC.d.cts} +58 -13
- package/dist/{model-catalog-vQ3NWn24.d.ts → model-catalog-CCsA1isC.d.ts} +58 -13
- package/dist/{model-constants-QZVHZX25.d.ts → model-constants-BtXkUsX3.d.cts} +36 -9
- package/dist/{model-constants-QZVHZX25.d.cts → model-constants-BtXkUsX3.d.ts} +36 -9
- package/package.json +1 -1
- package/src/__tests__/auto-aspect-source.test.ts +93 -0
- package/src/__tests__/gvp-supported-providers.test.ts +2 -0
- package/src/__tests__/normalize-model-input.test.ts +62 -8
- package/src/__tests__/resolutions-by-aspect-ratio.test.ts +134 -0
- package/src/__tests__/seedance-25-catalog-modes.test.ts +35 -0
- package/src/__tests__/video-frame-slot.test.ts +5 -0
- package/src/credit-identifiers.ts +3 -1
- package/src/entries/model-catalog.ts +2 -0
- package/src/entries/models.ts +5 -0
- package/src/model-catalog.ts +161 -20
- package/src/model-constants.ts +79 -7
- package/src/node-default-mappings.ts +1 -0
- package/src/video-ui-defaults.ts +2 -1
- package/dist/chunk-2AR3OXAA.js.map +0 -1
- package/dist/chunk-2J6NTA5I.js.map +0 -1
- package/dist/chunk-AZDVL7IM.js.map +0 -1
- package/dist/chunk-CGZLENU4.cjs.map +0 -1
- package/dist/chunk-EWARI452.cjs.map +0 -1
- package/dist/chunk-O7DXNX5J.cjs.map +0 -1
- package/dist/chunk-ST4WQ6ND.js.map +0 -1
- package/dist/chunk-XPUGMYAJ.cjs +0 -27
- package/dist/chunk-XPUGMYAJ.cjs.map +0 -1
|
@@ -295,10 +295,34 @@ interface ModelCatalogEntry {
|
|
|
295
295
|
* Declared only where the provider documents its own form: Grok Imagine's
|
|
296
296
|
* `@image(n)` (docs.kie.ai/market/grok-imagine/image-to-video) and
|
|
297
297
|
* HappyHorse's `[Image N]` (docs.kie.ai/market/happyhorse-1-1/reference-to-video).
|
|
298
|
+
* Gemini Omni Flash 1.2 documents none and is written plain `image N`, the
|
|
299
|
+
* form of KIE's own example; beside references its frames only guide the
|
|
300
|
+
* clip, in either wording.
|
|
298
301
|
*/
|
|
299
|
-
referenceSyntax?: "@image(n)" | "[Image N]";
|
|
302
|
+
referenceSyntax?: "@image(n)" | "[Image N]" | "image N";
|
|
300
303
|
aspectRatios?: readonly string[];
|
|
304
|
+
/**
|
|
305
|
+
* The model lists "auto", but its own auto does not keep the photo's shape
|
|
306
|
+
* (GPT Image 2.6 renders a portrait square). "auto" with a photo is then
|
|
307
|
+
* resolved as on a model without a native auto: to the listed ratio nearest
|
|
308
|
+
* the photo (`autoAspectNeedsSourceImage`, `normalizeModelInput`). With no
|
|
309
|
+
* photo, or one whose size cannot be read, the model's own "auto" is sent,
|
|
310
|
+
* as before. "auto" stays in `aspectRatios`, so every menu still offers it.
|
|
311
|
+
*/
|
|
312
|
+
autoAspectFromSourceImage?: true;
|
|
301
313
|
resolutions?: readonly string[];
|
|
314
|
+
/**
|
|
315
|
+
* The aspect ratios at which this model renders fewer sizes than
|
|
316
|
+
* `resolutions`: each one names the sizes it renders, in the catalog's own
|
|
317
|
+
* spelling. A ratio not listed renders every entry of `resolutions`.
|
|
318
|
+
* `normalizeModelInput` moves a size such a ratio does not render to the
|
|
319
|
+
* largest one it does at or below it, and says so, before the run is priced;
|
|
320
|
+
* a menu reads the sizes on offer with `resolutionsAtAspectRatio`.
|
|
321
|
+
*/
|
|
322
|
+
resolutionsByAspectRatio?: Readonly<Record<string, readonly string[]>>;
|
|
323
|
+
/** The resolution a caller who names none lands on when it is not the
|
|
324
|
+
* cheapest entry (see `defaultResolutionFor`). */
|
|
325
|
+
defaultResolution?: string;
|
|
302
326
|
/**
|
|
303
327
|
* What this model actually RENDERS for a resolution outside `resolutions`.
|
|
304
328
|
*
|
|
@@ -489,8 +513,10 @@ declare function validateModelInput(modelId: string, input: {
|
|
|
489
513
|
* The resolution a Flux 2 model should land on when its stored value is absent
|
|
490
514
|
* or invalid. Flux 2 exposes ascending megapixel options ("0.5 MP"…"4 MP"), so
|
|
491
515
|
* snapping to `options[0]` would silently downgrade every node to the cheapest
|
|
492
|
-
* tier; each variant has a sensible mid default instead.
|
|
493
|
-
*
|
|
516
|
+
* tier; each variant has a sensible mid default instead. A model can also
|
|
517
|
+
* declare one (`defaultResolution`, Gemini Omni Flash 1.2: its cheapest tier is
|
|
518
|
+
* a 360p preview). Returns undefined for every other model (they snap to
|
|
519
|
+
* `options[0]` normally).
|
|
494
520
|
*/
|
|
495
521
|
declare function defaultResolutionFor(modelId: string): string | undefined;
|
|
496
522
|
/** One correction `normalizeModelInput` made, for disclosure to the user. */
|
|
@@ -520,8 +546,9 @@ interface SourceImageSize {
|
|
|
520
546
|
interface ModelInputContext {
|
|
521
547
|
/**
|
|
522
548
|
* The image an image-to-image or edit request transforms. Only read to
|
|
523
|
-
* resolve "auto" on a model without a native auto
|
|
524
|
-
*
|
|
549
|
+
* resolve "auto" on a model without a native auto, or whose own auto does
|
|
550
|
+
* not keep the photo's shape (`autoAspectFromSourceImage`); absent or
|
|
551
|
+
* unusable means the size is unknown and "auto" snaps the way it always has.
|
|
525
552
|
*/
|
|
526
553
|
sourceImage?: SourceImageSize;
|
|
527
554
|
}
|
|
@@ -538,11 +565,28 @@ declare function isAutoAspectToken(value: unknown): boolean;
|
|
|
538
565
|
declare function nearestCatalogAspectRatio(sourceWidth: number, sourceHeight: number, ratios: readonly string[]): string | undefined;
|
|
539
566
|
/**
|
|
540
567
|
* True when `aspectRatio` is "auto" and `modelId`'s catalog ratio list does not
|
|
541
|
-
* carry "auto"
|
|
542
|
-
*
|
|
543
|
-
*
|
|
568
|
+
* carry "auto", or carries it but the model's own auto does not keep the
|
|
569
|
+
* photo's shape (`autoAspectFromSourceImage`) — the cases where the source
|
|
570
|
+
* image's size decides the ratio. A caller that has to fetch the size asks
|
|
571
|
+
* this first, so a request that will not use it never pays for the read.
|
|
544
572
|
*/
|
|
545
573
|
declare function autoAspectNeedsSourceImage(modelId: string | undefined, aspectRatio: unknown): boolean;
|
|
574
|
+
/**
|
|
575
|
+
* The sizes `modelId` renders at `aspectRatio`, in the catalog's order: its
|
|
576
|
+
* `resolutions`, narrowed where `resolutionsByAspectRatio` lists that ratio.
|
|
577
|
+
* `undefined` for an unknown model or one with no size setting. What a menu
|
|
578
|
+
* offers once a ratio is chosen.
|
|
579
|
+
*/
|
|
580
|
+
declare function resolutionsAtAspectRatio(modelId: string, aspectRatio: unknown): readonly string[] | undefined;
|
|
581
|
+
/**
|
|
582
|
+
* The size `modelId` renders `resolution` at, at `aspectRatio`: the same size
|
|
583
|
+
* where that ratio renders it (or the model limits no ratio), else the largest
|
|
584
|
+
* size the ratio renders at or below it, else its smallest. A size the model
|
|
585
|
+
* does not list comes back as it is: snapping the lever itself is
|
|
586
|
+
* `normalizeModelInput`'s. A route that prices a run without that function
|
|
587
|
+
* prices and sends the size this returns, so the two agree.
|
|
588
|
+
*/
|
|
589
|
+
declare function resolutionAtAspectRatio(modelId: string, aspectRatio: unknown, resolution: string | undefined): string | undefined;
|
|
546
590
|
/**
|
|
547
591
|
* Coerce a model's parameters into a combination the model actually accepts.
|
|
548
592
|
*
|
|
@@ -557,10 +601,11 @@ declare function autoAspectNeedsSourceImage(modelId: string | undefined, aspectR
|
|
|
557
601
|
*
|
|
558
602
|
* Rules, mirroring the config panels' provider-change snap:
|
|
559
603
|
* - Model has no such lever → drop the value (sending it 400s upstream).
|
|
560
|
-
* - "auto" on a model whose list lacks it,
|
|
561
|
-
*
|
|
562
|
-
*
|
|
563
|
-
*
|
|
604
|
+
* - "auto" on a model whose list lacks it, or whose own auto does not keep
|
|
605
|
+
* the photo's shape (`autoAspectFromSourceImage`), with a known
|
|
606
|
+
* `context.sourceImage` → the listed ratio nearest the source's shape
|
|
607
|
+
* (`nearestCatalogAspectRatio`). "auto" asks to keep the source photo's
|
|
608
|
+
* shape; the first listed ratio would reframe every source of another shape.
|
|
564
609
|
* - Value outside the model's allow-list → snap to the model's default
|
|
565
610
|
* (`defaultResolutionFor`) or the first valid option.
|
|
566
611
|
* - Then apply cross-field constraints that only hold for certain models.
|
|
@@ -732,4 +777,4 @@ declare function modelIdsByKindMode(kind: ModelKind | null, modes: readonly Mode
|
|
|
732
777
|
includeHidden?: boolean;
|
|
733
778
|
}): [string, ...string[]];
|
|
734
779
|
|
|
735
|
-
export { getDurationsForModel as A, getModel as B, getQualityOptions as C, getResolutionOptions as D, groupByFamily as E, groupByKindAndFamily as F, hasFeature as G, isAutoAspectToken as H, isModelMakerId as I, listModels as J, modelIdsByKindMode as K, type LabeledOption as L, MODEL_CATALOG as M, type NormalizedModelInput as N, modelMakerOf as O, type PriceVariant as P, modelTakesSeed as Q, modelsWithFeature as R, SPEECH_UNIT_PRICE_NOTE as S, type TtsCapabilities as T, nearestCatalogAspectRatio as U, type ValidationField as V, normalizeModelInput as W, normalizeVideoRequestParams as X, qualityOptionsByKind as Y,
|
|
780
|
+
export { resolutionsAtAspectRatio as $, getDurationsForModel as A, getModel as B, getQualityOptions as C, getResolutionOptions as D, groupByFamily as E, groupByKindAndFamily as F, hasFeature as G, isAutoAspectToken as H, isModelMakerId as I, listModels as J, modelIdsByKindMode as K, type LabeledOption as L, MODEL_CATALOG as M, type NormalizedModelInput as N, modelMakerOf as O, type PriceVariant as P, modelTakesSeed as Q, modelsWithFeature as R, SPEECH_UNIT_PRICE_NOTE as S, type TtsCapabilities as T, nearestCatalogAspectRatio as U, type ValidationField as V, normalizeModelInput as W, normalizeVideoRequestParams as X, qualityOptionsByKind as Y, resolutionAtAspectRatio as Z, resolutionOptionsByKind as _, MODEL_KINDS as a, validateModelInput as a0, MODEL_MAKERS as b, MODEL_RECOMMENDATIONS as c, type ModelCatalogEntry as d, type ModelInputAdjustment as e, type ModelInputContext as f, type ModelKind as g, type ModelMaker as h, type ModelMakerId as i, type ModelMenuOption as j, type ModelMode as k, type ModelRecommendation as l, type ModelValidationIssue as m, type NormalizedVideoRequest as n, type SourceImageSize as o, type TtsSettingLever as p, aspectRatioOptionsByKind as q, autoAspectNeedsSourceImage as r, backgroundOptionsByKind as s, buildModelMenu as t, creditRangesAll as u, defaultResolutionFor as v, durationsByMode as w, fitAspectRatioToModel as x, getAspectRatioOptions as y, getCreditRange as z };
|
|
@@ -151,6 +151,18 @@ declare const PROMPT_HARD_CEILING = 30000;
|
|
|
151
151
|
* route's maxTokens.
|
|
152
152
|
*/
|
|
153
153
|
declare const LLM_TEXT_INPUT_MAX = 100000;
|
|
154
|
+
/**
|
|
155
|
+
* Ceiling for the `system` prompt of the structured-output routes
|
|
156
|
+
* (`POST /v1/llm/structured` and `POST /v1/llm/structured/jobs`). A client
|
|
157
|
+
* that has the model write in its own vocabulary sends that vocabulary as the
|
|
158
|
+
* system prompt, and it grows with the client's catalogs (Nodaro Studio's
|
|
159
|
+
* production format is close to 100,000 chars), so this is twice
|
|
160
|
+
* {@link LLM_TEXT_INPUT_MAX}. The routes' `input`, and every other LLM text
|
|
161
|
+
* field, stay at LLM_TEXT_INPUT_MAX. 200,000 chars is roughly 50,000 to
|
|
162
|
+
* 100,000 tokens, and a body at both ceilings still fits the API's 1 MB
|
|
163
|
+
* request limit.
|
|
164
|
+
*/
|
|
165
|
+
declare const LLM_STRUCTURED_SYSTEM_MAX = 200000;
|
|
154
166
|
/**
|
|
155
167
|
* Per-provider maximum VIDEO prompt length (chars), VERIFIED against each model's
|
|
156
168
|
* official docs.kie.ai schema (2026-06). Absent → {@link VIDEO_PROMPT_MAX} (8000)
|
|
@@ -374,13 +386,13 @@ type ImageGenProvider = typeof IMAGE_GEN_PROVIDERS[number];
|
|
|
374
386
|
type ImageI2IProvider = typeof IMAGE_I2I_PROVIDERS[number];
|
|
375
387
|
type ImageEditProvider = typeof IMAGE_EDIT_PROVIDERS[number];
|
|
376
388
|
/** Image-to-video providers */
|
|
377
|
-
declare const IMAGE_TO_VIDEO_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "kling-master", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "hailuo-2.3-pro", "hailuo-2.3", "hailuo-standard", "wan-i2v", "wan-turbo", "bytedance-lite", "bytedance-pro", "bytedance-pro-fast", "grok-i2v", "wan-2.7-i2v", "happyhorse-i2v", "happyhorse-ref2v", "runway-kie", "kling-3-omni", "gemini-omni-video", "gemini-omni-flash", "wan-3", "wan-3-prime", "vidu-q4", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5"];
|
|
389
|
+
declare const IMAGE_TO_VIDEO_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "kling-master", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "hailuo-2.3-pro", "hailuo-2.3", "hailuo-standard", "wan-i2v", "wan-turbo", "bytedance-lite", "bytedance-pro", "bytedance-pro-fast", "grok-i2v", "wan-2.7-i2v", "happyhorse-i2v", "happyhorse-ref2v", "runway-kie", "kling-3-omni", "gemini-omni-video", "gemini-omni-flash", "gemini-omni-flash-1-2", "wan-3", "wan-3-prime", "vidu-q4", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5"];
|
|
378
390
|
type ImageToVideoProvider = typeof IMAGE_TO_VIDEO_PROVIDERS[number];
|
|
379
391
|
/** Text-to-video providers */
|
|
380
|
-
declare const TEXT_TO_VIDEO_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "grok", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "wan", "hailuo-standard", "bytedance-lite", "bytedance-pro", "wan-turbo", "wan-2.7-t2v", "happyhorse", "runway-kie", "gemini-omni-video", "gemini-omni-flash", "wan-3", "wan-3-prime", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5", "kling-3-omni", "kling-master", "hailuo-2.3", "hailuo-2.3-pro", "bytedance-pro-fast", "happyhorse-ref2v", "vidu-q4"];
|
|
392
|
+
declare const TEXT_TO_VIDEO_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "grok", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "wan", "hailuo-standard", "bytedance-lite", "bytedance-pro", "wan-turbo", "wan-2.7-t2v", "happyhorse", "runway-kie", "gemini-omni-video", "gemini-omni-flash", "gemini-omni-flash-1-2", "wan-3", "wan-3-prime", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5", "kling-3-omni", "kling-master", "hailuo-2.3", "hailuo-2.3-pro", "bytedance-pro-fast", "happyhorse-ref2v", "vidu-q4"];
|
|
381
393
|
type TextToVideoProvider = typeof TEXT_TO_VIDEO_PROVIDERS[number];
|
|
382
394
|
/** Unified video-generation providers (image-to-video ∪ text-to-video) for the generate-video node. */
|
|
383
|
-
declare const VIDEO_GEN_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "kling-master", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "hailuo-2.3-pro", "hailuo-2.3", "hailuo-standard", "wan-i2v", "wan-turbo", "bytedance-lite", "bytedance-pro", "bytedance-pro-fast", "grok-i2v", "wan-2.7-i2v", "happyhorse-i2v", "happyhorse-ref2v", "runway-kie", "kling-3-omni", "gemini-omni-video", "gemini-omni-flash", "wan-3", "wan-3-prime", "vidu-q4", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5", ...("seedance-2-5" | "minimax" | "veo3" | "kling-3.0" | "seedance-2" | "veo3.1" | "wan-turbo" | "bytedance-lite" | "kling-turbo" | "hailuo-standard" | "grok" | "minimax-h3" | "hailuo-2.3-pro" | "hailuo-2.3" | "veo3_lite" | "gemini-omni-video" | "gemini-omni-flash" | "kling" | "kling-master" | "kling-3-omni" | "kling-4" | "kling-4-keyframes" | "grok-imagine-video-1.5" | "seedance-2-fast" | "seedance-2-mini" | "wan-3" | "wan-3-prime" | "wan" | "wan-2.7-t2v" | "ltx-2.3-pro" | "ltx-2.3-fast" | "happyhorse" | "happyhorse-ref2v" | "vidu-q4" | "bytedance-pro" | "bytedance-pro-fast" | "runway-kie")[]];
|
|
395
|
+
declare const VIDEO_GEN_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "kling-master", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "hailuo-2.3-pro", "hailuo-2.3", "hailuo-standard", "wan-i2v", "wan-turbo", "bytedance-lite", "bytedance-pro", "bytedance-pro-fast", "grok-i2v", "wan-2.7-i2v", "happyhorse-i2v", "happyhorse-ref2v", "runway-kie", "kling-3-omni", "gemini-omni-video", "gemini-omni-flash", "gemini-omni-flash-1-2", "wan-3", "wan-3-prime", "vidu-q4", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5", ...("seedance-2-5" | "minimax" | "veo3" | "kling-3.0" | "seedance-2" | "veo3.1" | "wan-turbo" | "bytedance-lite" | "kling-turbo" | "hailuo-standard" | "grok" | "minimax-h3" | "hailuo-2.3-pro" | "hailuo-2.3" | "veo3_lite" | "gemini-omni-video" | "gemini-omni-flash" | "gemini-omni-flash-1-2" | "kling" | "kling-master" | "kling-3-omni" | "kling-4" | "kling-4-keyframes" | "grok-imagine-video-1.5" | "seedance-2-fast" | "seedance-2-mini" | "wan-3" | "wan-3-prime" | "wan" | "wan-2.7-t2v" | "ltx-2.3-pro" | "ltx-2.3-fast" | "happyhorse" | "happyhorse-ref2v" | "vidu-q4" | "bytedance-pro" | "bytedance-pro-fast" | "runway-kie")[]];
|
|
384
396
|
type VideoGenProvider = typeof VIDEO_GEN_PROVIDERS[number];
|
|
385
397
|
/**
|
|
386
398
|
* Video models that expose DISTINCT provider ids per mode (text-to-video vs
|
|
@@ -462,7 +474,7 @@ declare function resolveVideoModeForInputs(provider: string | undefined, inputs:
|
|
|
462
474
|
readonly hasImageRefs: boolean;
|
|
463
475
|
}): "image-to-video" | "text-to-video";
|
|
464
476
|
/** The prompt syntax a video provider addresses its images in, by array order. */
|
|
465
|
-
type VideoReferenceSyntax = "@image_N" | "@image(n)" | "[Image N]";
|
|
477
|
+
type VideoReferenceSyntax = "@image_N" | "@image(n)" | "[Image N]" | "image N";
|
|
466
478
|
/**
|
|
467
479
|
* The syntax this video model's frame / identity binding sentences are written
|
|
468
480
|
* in — the catalog's `referenceSyntax`, else the platform's `@image_N`. One
|
|
@@ -953,6 +965,20 @@ declare function kling4KeyframeFrameCap(durationSec: number | string | undefined
|
|
|
953
965
|
*/
|
|
954
966
|
declare const GEMINI_OMNI_PROVIDERS: Set<string>;
|
|
955
967
|
declare function isGeminiOmniProvider(provider: string | undefined): boolean;
|
|
968
|
+
/**
|
|
969
|
+
* The Gemini Omni SKUs billed per second of video (Gemini Omni Flash 1.2). What
|
|
970
|
+
* the family shares (V2V through the i2v handle, the panels) comes from
|
|
971
|
+
* {@link GEMINI_OMNI_PROVIDERS}. These differ in two ways: the price is a
|
|
972
|
+
* `:<N>s:<res>` ladder, and a run with a source video bills that clip's own
|
|
973
|
+
* seconds and nothing else (the output follows the clip), so they also answer
|
|
974
|
+
* {@link billsReferenceVideoSeconds}.
|
|
975
|
+
*/
|
|
976
|
+
declare const GEMINI_OMNI_PER_SECOND_PROVIDERS: Set<string>;
|
|
977
|
+
declare function isGeminiOmniPerSecondProvider(provider: string | undefined): boolean;
|
|
978
|
+
/** The longest source clip Gemini Omni Flash 1.2 takes (docs.kie.ai/market/google/gemini-omni-flash-1-2). */
|
|
979
|
+
declare const GEMINI_OMNI_12_MAX_SOURCE_SEC = 15;
|
|
980
|
+
/** The widest trim window a request may set on a Gemini Omni source clip, in seconds. */
|
|
981
|
+
declare function videoSourceTrimMaxSec(provider: string | undefined): number;
|
|
956
982
|
/**
|
|
957
983
|
* Video models that are CATALOGUED and priced but have NO working dispatch
|
|
958
984
|
* path today — selecting one fails at the router, not at validation.
|
|
@@ -1288,8 +1314,9 @@ declare function checkRefVideoDurations(provider: string, durationsSec: readonly
|
|
|
1288
1314
|
declare const VIDEO_REF_RATE_PROVIDERS: Set<string>;
|
|
1289
1315
|
/**
|
|
1290
1316
|
* Video models whose reference-VIDEO runs also bill the clips' own seconds:
|
|
1291
|
-
* the Seedance 2 family (`unit × (input + output)`)
|
|
1292
|
-
* VIDEO_REF_RATE_PROVIDERS (Kling 4.0: `unit × (output + max(output, input))`)
|
|
1317
|
+
* the Seedance 2 family (`unit × (input + output)`), the
|
|
1318
|
+
* VIDEO_REF_RATE_PROVIDERS (Kling 4.0: `unit × (output + max(output, input))`)
|
|
1319
|
+
* and the per-second Gemini Omni SKUs (`unit × source clip`, the output follows it).
|
|
1293
1320
|
* The reservation holds the worst case from the probed clips and the
|
|
1294
1321
|
* settlement meters down to what was delivered (backend
|
|
1295
1322
|
* ee/billing/seedance2-ref-video-credits.ts) — this is the one gate every lane
|
|
@@ -1356,8 +1383,8 @@ declare function videoProviderRequiresImage(provider: string | undefined): boole
|
|
|
1356
1383
|
* reference mode takes a lone last frame as the sole reference image plus a
|
|
1357
1384
|
* closing-frame prompt hint. NOTHING else does: VEO ships `[imageUrl,
|
|
1358
1385
|
* endFrameUrl]` verbatim (a lone end frame would reach KIE as `[null, url]`),
|
|
1359
|
-
* the Gemini Omni branch drops the frame when there is no start frame
|
|
1360
|
-
* generic KIE path sends `end_frame` with no image param at all. So the
|
|
1386
|
+
* the per-clip Gemini Omni branch drops the frame when there is no start frame (Omni
|
|
1387
|
+
* Flash 1.2 folds it, with its own builder), and the generic KIE path sends `end_frame` with no image param at all. So the
|
|
1361
1388
|
* exemption is keyed to the FOLD, not to "carries image references" — the
|
|
1362
1389
|
* catalog has no capability for it, and these three family predicates are the
|
|
1363
1390
|
* same membership the provider layer itself branches on (the backend's
|
|
@@ -1702,4 +1729,4 @@ declare function applyDefaultVideoSelection<D>(input: {
|
|
|
1702
1729
|
duration: D | number | undefined;
|
|
1703
1730
|
};
|
|
1704
1731
|
|
|
1705
|
-
export {
|
|
1732
|
+
export { IMAGE_MASK_MODE as $, AI_WRITER_PROVIDERS as A, type FaceSwapProvider as B, CHARACTER_MOTION_PROVIDERS as C, DEFAULT_DIALOGUE_PROVIDER as D, EXTEND_VIDEO_PROVIDERS as E, FACE_SWAP_PROVIDERS as F, GEMINI_OMNI_12_MAX_SOURCE_SEC as G, GEMINI_OMNI_PER_SECOND_PROVIDERS as H, GEMINI_OMNI_PROVIDERS as I, GUIDANCE_SCALE_SUPPORT as J, GVP_ANCHOR_CHOICES as K, GVP_DEFAULT_PROVIDER as L, GVP_END_FRAME_PROVIDERS as M, GVP_EXTEND_PROVIDERS as N, GVP_SUPPORTED_PROVIDERS as O, type GvpAnchorChoice as P, type GvpAnchorWireMode as Q, HIGH_QUALITY_PROVIDERS as R, I2I_MASK_SUPPORT as S, I2I_STRENGTH_SUPPORT as T, IDEOGRAM_PROVIDERS as U, IMAGE_ASPECT_RATIO_VALUES as V, IMAGE_BACKGROUND_VALUES as W, IMAGE_EDIT_PROVIDERS as X, IMAGE_EDIT_TARGET_RESOLUTION_VALUES as Y, IMAGE_GEN_PROVIDERS as Z, IMAGE_I2I_PROVIDERS as _, ASPECT_RATIO_DIMENSIONS as a, SEEDANCE_2_CONTINUATION_REF_SEC as a$, IMAGE_PROMPT_MAX as a0, IMAGE_QUALITY_VALUES as a1, IMAGE_RENDERING_SPEED_VALUES as a2, IMAGE_RESOLUTION_VALUES as a3, IMAGE_TO_VIDEO_PROVIDERS as a4, type ImageAspectRatio as a5, type ImageBackground as a6, type ImageEditProvider as a7, type ImageEditTargetResolution as a8, type ImageGenProvider as a9, MUSIC_PROVIDERS as aA, MUSIC_PROVIDER_LABELS as aB, type ModelPromptingStyle as aC, type ModifyImageProvider as aD, type MotionTransferProviderType as aE, type MusicProvider as aF, NATIVE_ADAPTIVE_ASPECT as aG, NATIVE_NEGATIVE_PROMPT_MODELS as aH, NATIVE_NEGATIVE_VIDEO_PROVIDERS as aI, NEGATIVE_PROMPT_MAX as aJ, OBJECT_MOTION_PROVIDERS as aK, type ObjectMotionProvider as aL, PRICING_DEFAULT_DURATION_SEC as aM, PRICING_DEFAULT_RESOLUTION as aN, PROMPT_HARD_CEILING as aO, QA_CHECK_PROVIDERS as aP, type QaCheckProvider as aQ, REF_IMAGE_MAX_LIMITS as aR, RENDERING_SPEED_SUPPORT as aS, REPLICATE_LIP_SYNC_PROVIDERS as aT, RETIRED_VIDEO_PROVIDERS as aU, type RefAudioCheck as aV, type RefAudioProbe as aW, type RefAudioRule as aX, type RefVideoDurationLimit as aY, SCRIPT_PROVIDERS as aZ, SEEDANCE_2_5_REF_LIMITS as a_, type ImageI2IProvider as aa, type ImageMaskMode as ab, type ImageQuality as ac, type ImageRenderingSpeed as ad, type ImageResolution as ae, type ImageToVideoProvider as af, KLING_4_DEFAULT_RESOLUTION as ag, KLING_4_MAX_MIDDLE_FRAMES as ah, KLING_4_PROVIDERS as ai, LIP_SYNC_PROVIDERS as aj, LLM_STRUCTURED_SYSTEM_MAX as ak, LLM_TEXT_INPUT_MAX as al, LOCATION_ATMOSPHERE_PROVIDERS as am, type LipSyncProvider as an, type LocationAtmosphereProvider as ao, MAX_IMAGE_PROMPT_CHARS_BY_PROVIDER as ap, MAX_NEGATIVE_PROMPT_CHARS_BY_PROVIDER as aq, MAX_REQUEST_IMAGE_LIMIT as ar, MAX_TTS_CHARS_BY_PROVIDER as as, MAX_VIDEO_PROMPT_CHARS_BY_PROVIDER as at, MINIMAX_H3_DEFAULT_RESOLUTION as au, MINIMAX_H3_PROVIDERS as av, MODELS_WITH_REFERENCE_IMAGE_SUPPORT as aw, MODIFY_IMAGE_PROVIDERS as ax, MOTION_TRANSFER_PROVIDERS as ay, MUSIC_CREDIT_ID as az, AUDIO_ADDON_PROVIDERS as b, VIDEO_TO_VIDEO_PROVIDERS as b$, SEEDANCE_2_EXTEND_STITCH as b0, SEEDANCE_2_PROVIDERS as b1, SEEDANCE_2_R2V_MAX_AUDIO_SEC_BY_PROVIDER as b2, SEEDANCE_2_R2V_MIN_REF_VIDEO_SEC as b3, SEEDANCE_2_REF_LIMITS as b4, SEEDANCE_LIP_SYNC_PROVIDERS as b5, SEEDANCE_VIDEO_EDIT_PROVIDERS as b6, SEEDANCE_VIDEO_EDIT_SHAPE as b7, SEED_SUPPORT as b8, SUNO_ACTIVE_MODELS as b9, type TaskChainedEditProvider as bA, type TextToAudioProvider as bB, type TextToVideoProvider as bC, type TranscribeLane as bD, type TranscribeProvider as bE, type TtsOutputFormat as bF, type TtsProvider as bG, UPSCALE_IMAGE_PROVIDERS as bH, type UpscaleImageProvider as bI, VARIABLE_PRICING_MODELS as bJ, VEO_PROVIDERS as bK, VIDEO_AUDIO_CAPABILITY as bL, VIDEO_DURATION_TIERS as bM, VIDEO_GEN_COLLAPSED_T2V_IDS as bN, VIDEO_GEN_PROVIDERS as bO, VIDEO_INPUT_LIP_SYNC_PROVIDERS as bP, VIDEO_MODEL_CAPS as bQ, VIDEO_MODE_ALIASES as bR, VIDEO_PROMPT_MAX as bS, VIDEO_PROVIDERS_REQUIRING_IMAGE as bT, VIDEO_PROVIDERS_WITHOUT_DISPATCH as bU, VIDEO_PROVIDERS_WITH_SEPARATE_FRAME_CHANNEL as bV, VIDEO_REF_AUDIO_RULES as bW, VIDEO_REF_LIMITS_BY_PROVIDER as bX, VIDEO_REF_RATE_PROVIDERS as bY, VIDEO_REF_VIDEO_DURATION_LIMITS as bZ, VIDEO_TO_VIDEO_NODE_PROVIDERS as b_, SUNO_ADD_TRACK_MODELS as ba, SUNO_DURATION_MODELS as bb, SUNO_HARD_CEILING as bc, SUNO_LEGACY_MODELS as bd, SUNO_MODELS as be, SUNO_TEXT_MAX as bf, SUNO_TITLE_MAX as bg, type SceneInputMode as bh, SceneInputModeSchema as bi, type ScriptProvider as bj, type SeedanceVideoEditProvider as bk, type SunoAddTrackModel as bl, type SunoModel as bm, T2I_TO_I2I_VARIANT as bn, TASK_CHAINED_EDIT_PROVIDERS as bo, TASK_CHAINED_EDIT_PROVIDER_IDS as bp, TEXT_TO_AUDIO_PROVIDERS as bq, TEXT_TO_VIDEO_PROVIDERS as br, TRANSCRIBE_LANES as bs, TRANSCRIBE_PROVIDERS as bt, TRANSCRIBE_PROVIDER_CAPABILITIES as bu, TTS_OUTPUT_FORMATS as bv, TTS_OUTPUT_FORMAT_FILES as bw, TTS_PROVIDERS as bx, TTS_TEXT_MAX as by, TWO_K_RESOLUTION_PROVIDERS as bz, type AiWriterProvider as c, seedance2AudioLimitSec as c$, VIDEO_UPSCALE_PROVIDERS as c0, VIDEO_VARIABLE_PRICING as c1, VIDU_Q4_DEFAULT_RESOLUTION as c2, VIDU_Q4_PROVIDERS as c3, VOICE_DESIGN_MODELS as c4, type VideoAudioCapability as c5, type VideoAudioMode as c6, type VideoGenProvider as c7, type VideoModeAlias as c8, type VideoModelCapabilities as c9, isGeminiOmniProvider as cA, isGvpSupportedProvider as cB, isKling4Provider as cC, isLegacySunoModel as cD, isMinimaxH3Provider as cE, isSeedance2Provider as cF, isSeedanceVideoEditProvider as cG, isVeoProvider as cH, isViduQ4Provider as cI, isWan3Provider as cJ, kling4KeyframeFrameCap as cK, maxSegmentSecFor as cL, maxSegmentsFor as cM, maxVideoDurationSec as cN, minSegmentSecFor as cO, modelsForInputMode as cP, normalizeKling4Resolution as cQ, normalizeMinimaxH3Resolution as cR, normalizeViduQ4Resolution as cS, normalizeWan3Resolution as cT, preferredInputModeForModel as cU, pricedOutputDurationSec as cV, requestImageLimit as cW, resolveGvpAnchorWire as cX, resolveMusicProvider as cY, resolveVideoModeForInputs as cZ, resolveVideoProviderForMode as c_, type VideoReferenceSyntax as ca, type VideoToVideoNodeProvider as cb, type VideoToVideoProvider as cc, type VideoUpscaleProvider as cd, type VoiceDesignModel as ce, WAN_3_DEFAULT_RESOLUTION as cf, WAN_3_PROVIDERS as cg, applyDefaultVideoSelection as ch, applyVideoAudioToggle as ci, applyVideoNegativePrompt as cj, billsReferenceVideoSeconds as ck, canonicalVideoProvider as cl, checkRefAudioRules as cm, checkRefVideoDurations as cn, defaultVideoAspectRatio as co, findSeedance2AudioOverLimit as cp, getMaxImagePromptChars as cq, getMaxNegativePromptChars as cr, getMaxSunoPromptChars as cs, getMaxSunoStyleChars as ct, getMaxTtsChars as cu, getMaxVideoPromptChars as cv, getVideoAudioCapability as cw, hasContiguousSegmentDurations as cx, imageReferenceLimit as cy, isGeminiOmniPerSecondProvider as cz, COMPOSER_PLAN_FIELDS as d, segmentDurationsFor as d0, snapToNearestDuration as d1, sunoModelHonoursDuration as d2, supportsAutoVideoDuration as d3, supportsEndAnchor as d4, supportsExtendRender as d5, transcribeLaneSupportsWordTimestamps as d6, transcribeProvidersWithWordTimestamps as d7, videoDefaultDurationSec as d8, videoModelCanSpeakDialogue as d9, videoModelSupportsAudio as da, videoNegativeSuffix as db, videoProviderAppendsFrames as dc, videoProviderFoldsLoneEndFrame as dd, videoProviderRefModeFrameSeats as de, videoProviderReferenceSyntax as df, videoProviderRequiresImage as dg, videoProviderRunsFromReferences as dh, videoSourceTrimMaxSec as di, COMPOSER_PLAN_MAP as e, CREDIT_BASE_USD as f, type CharacterMotionProvider as g, DEFAULT_MUSIC_PROVIDER as h, DEFAULT_REF_IMAGE_MAX as i, DEFAULT_REQUEST_IMAGE_LIMIT as j, DEFAULT_SUNO_MODEL as k, DEFAULT_TEXT_TO_AUDIO_PROVIDER as l, DEFAULT_TRANSCRIBE_NODE_PROVIDER as m, DEFAULT_TRANSCRIBE_PROVIDER as n, DEFAULT_TTS_OUTPUT_FORMAT as o, DEFAULT_TTS_PROVIDER as p, DEFAULT_VIDEO_DURATION_SEC as q, DEFAULT_VIDEO_PROVIDER as r, DEFAULT_VOICE_DESIGN_MODEL as s, DIALOGUE_PROVIDERS as t, DURATION_PRICED_PROVIDERS as u, type DialogueProvider as v, type ExtendVideoProvider as w, FAL_LIP_SYNC_PROVIDERS as x, FLEXIBLE_INPUT_LIP_SYNC_PROVIDERS as y, FRAME_MODE_ADAPTIVE_ONLY_ASPECT as z };
|
|
@@ -151,6 +151,18 @@ declare const PROMPT_HARD_CEILING = 30000;
|
|
|
151
151
|
* route's maxTokens.
|
|
152
152
|
*/
|
|
153
153
|
declare const LLM_TEXT_INPUT_MAX = 100000;
|
|
154
|
+
/**
|
|
155
|
+
* Ceiling for the `system` prompt of the structured-output routes
|
|
156
|
+
* (`POST /v1/llm/structured` and `POST /v1/llm/structured/jobs`). A client
|
|
157
|
+
* that has the model write in its own vocabulary sends that vocabulary as the
|
|
158
|
+
* system prompt, and it grows with the client's catalogs (Nodaro Studio's
|
|
159
|
+
* production format is close to 100,000 chars), so this is twice
|
|
160
|
+
* {@link LLM_TEXT_INPUT_MAX}. The routes' `input`, and every other LLM text
|
|
161
|
+
* field, stay at LLM_TEXT_INPUT_MAX. 200,000 chars is roughly 50,000 to
|
|
162
|
+
* 100,000 tokens, and a body at both ceilings still fits the API's 1 MB
|
|
163
|
+
* request limit.
|
|
164
|
+
*/
|
|
165
|
+
declare const LLM_STRUCTURED_SYSTEM_MAX = 200000;
|
|
154
166
|
/**
|
|
155
167
|
* Per-provider maximum VIDEO prompt length (chars), VERIFIED against each model's
|
|
156
168
|
* official docs.kie.ai schema (2026-06). Absent → {@link VIDEO_PROMPT_MAX} (8000)
|
|
@@ -374,13 +386,13 @@ type ImageGenProvider = typeof IMAGE_GEN_PROVIDERS[number];
|
|
|
374
386
|
type ImageI2IProvider = typeof IMAGE_I2I_PROVIDERS[number];
|
|
375
387
|
type ImageEditProvider = typeof IMAGE_EDIT_PROVIDERS[number];
|
|
376
388
|
/** Image-to-video providers */
|
|
377
|
-
declare const IMAGE_TO_VIDEO_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "kling-master", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "hailuo-2.3-pro", "hailuo-2.3", "hailuo-standard", "wan-i2v", "wan-turbo", "bytedance-lite", "bytedance-pro", "bytedance-pro-fast", "grok-i2v", "wan-2.7-i2v", "happyhorse-i2v", "happyhorse-ref2v", "runway-kie", "kling-3-omni", "gemini-omni-video", "gemini-omni-flash", "wan-3", "wan-3-prime", "vidu-q4", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5"];
|
|
389
|
+
declare const IMAGE_TO_VIDEO_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "kling-master", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "hailuo-2.3-pro", "hailuo-2.3", "hailuo-standard", "wan-i2v", "wan-turbo", "bytedance-lite", "bytedance-pro", "bytedance-pro-fast", "grok-i2v", "wan-2.7-i2v", "happyhorse-i2v", "happyhorse-ref2v", "runway-kie", "kling-3-omni", "gemini-omni-video", "gemini-omni-flash", "gemini-omni-flash-1-2", "wan-3", "wan-3-prime", "vidu-q4", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5"];
|
|
378
390
|
type ImageToVideoProvider = typeof IMAGE_TO_VIDEO_PROVIDERS[number];
|
|
379
391
|
/** Text-to-video providers */
|
|
380
|
-
declare const TEXT_TO_VIDEO_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "grok", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "wan", "hailuo-standard", "bytedance-lite", "bytedance-pro", "wan-turbo", "wan-2.7-t2v", "happyhorse", "runway-kie", "gemini-omni-video", "gemini-omni-flash", "wan-3", "wan-3-prime", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5", "kling-3-omni", "kling-master", "hailuo-2.3", "hailuo-2.3-pro", "bytedance-pro-fast", "happyhorse-ref2v", "vidu-q4"];
|
|
392
|
+
declare const TEXT_TO_VIDEO_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "grok", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "wan", "hailuo-standard", "bytedance-lite", "bytedance-pro", "wan-turbo", "wan-2.7-t2v", "happyhorse", "runway-kie", "gemini-omni-video", "gemini-omni-flash", "gemini-omni-flash-1-2", "wan-3", "wan-3-prime", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5", "kling-3-omni", "kling-master", "hailuo-2.3", "hailuo-2.3-pro", "bytedance-pro-fast", "happyhorse-ref2v", "vidu-q4"];
|
|
381
393
|
type TextToVideoProvider = typeof TEXT_TO_VIDEO_PROVIDERS[number];
|
|
382
394
|
/** Unified video-generation providers (image-to-video ∪ text-to-video) for the generate-video node. */
|
|
383
|
-
declare const VIDEO_GEN_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "kling-master", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "hailuo-2.3-pro", "hailuo-2.3", "hailuo-standard", "wan-i2v", "wan-turbo", "bytedance-lite", "bytedance-pro", "bytedance-pro-fast", "grok-i2v", "wan-2.7-i2v", "happyhorse-i2v", "happyhorse-ref2v", "runway-kie", "kling-3-omni", "gemini-omni-video", "gemini-omni-flash", "wan-3", "wan-3-prime", "vidu-q4", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5", ...("seedance-2-5" | "minimax" | "veo3" | "kling-3.0" | "seedance-2" | "veo3.1" | "wan-turbo" | "bytedance-lite" | "kling-turbo" | "hailuo-standard" | "grok" | "minimax-h3" | "hailuo-2.3-pro" | "hailuo-2.3" | "veo3_lite" | "gemini-omni-video" | "gemini-omni-flash" | "kling" | "kling-master" | "kling-3-omni" | "kling-4" | "kling-4-keyframes" | "grok-imagine-video-1.5" | "seedance-2-fast" | "seedance-2-mini" | "wan-3" | "wan-3-prime" | "wan" | "wan-2.7-t2v" | "ltx-2.3-pro" | "ltx-2.3-fast" | "happyhorse" | "happyhorse-ref2v" | "vidu-q4" | "bytedance-pro" | "bytedance-pro-fast" | "runway-kie")[]];
|
|
395
|
+
declare const VIDEO_GEN_PROVIDERS: readonly ["minimax", "veo3", "veo3.1", "veo3_lite", "kling", "kling-turbo", "kling-3.0", "kling-master", "seedance-2", "seedance-2-fast", "seedance-2-mini", "seedance-2-5", "minimax-h3", "hailuo-2.3-pro", "hailuo-2.3", "hailuo-standard", "wan-i2v", "wan-turbo", "bytedance-lite", "bytedance-pro", "bytedance-pro-fast", "grok-i2v", "wan-2.7-i2v", "happyhorse-i2v", "happyhorse-ref2v", "runway-kie", "kling-3-omni", "gemini-omni-video", "gemini-omni-flash", "gemini-omni-flash-1-2", "wan-3", "wan-3-prime", "vidu-q4", "kling-4", "kling-4-keyframes", "ltx-2.3-pro", "ltx-2.3-fast", "grok-imagine-video-1.5", ...("seedance-2-5" | "minimax" | "veo3" | "kling-3.0" | "seedance-2" | "veo3.1" | "wan-turbo" | "bytedance-lite" | "kling-turbo" | "hailuo-standard" | "grok" | "minimax-h3" | "hailuo-2.3-pro" | "hailuo-2.3" | "veo3_lite" | "gemini-omni-video" | "gemini-omni-flash" | "gemini-omni-flash-1-2" | "kling" | "kling-master" | "kling-3-omni" | "kling-4" | "kling-4-keyframes" | "grok-imagine-video-1.5" | "seedance-2-fast" | "seedance-2-mini" | "wan-3" | "wan-3-prime" | "wan" | "wan-2.7-t2v" | "ltx-2.3-pro" | "ltx-2.3-fast" | "happyhorse" | "happyhorse-ref2v" | "vidu-q4" | "bytedance-pro" | "bytedance-pro-fast" | "runway-kie")[]];
|
|
384
396
|
type VideoGenProvider = typeof VIDEO_GEN_PROVIDERS[number];
|
|
385
397
|
/**
|
|
386
398
|
* Video models that expose DISTINCT provider ids per mode (text-to-video vs
|
|
@@ -462,7 +474,7 @@ declare function resolveVideoModeForInputs(provider: string | undefined, inputs:
|
|
|
462
474
|
readonly hasImageRefs: boolean;
|
|
463
475
|
}): "image-to-video" | "text-to-video";
|
|
464
476
|
/** The prompt syntax a video provider addresses its images in, by array order. */
|
|
465
|
-
type VideoReferenceSyntax = "@image_N" | "@image(n)" | "[Image N]";
|
|
477
|
+
type VideoReferenceSyntax = "@image_N" | "@image(n)" | "[Image N]" | "image N";
|
|
466
478
|
/**
|
|
467
479
|
* The syntax this video model's frame / identity binding sentences are written
|
|
468
480
|
* in — the catalog's `referenceSyntax`, else the platform's `@image_N`. One
|
|
@@ -953,6 +965,20 @@ declare function kling4KeyframeFrameCap(durationSec: number | string | undefined
|
|
|
953
965
|
*/
|
|
954
966
|
declare const GEMINI_OMNI_PROVIDERS: Set<string>;
|
|
955
967
|
declare function isGeminiOmniProvider(provider: string | undefined): boolean;
|
|
968
|
+
/**
|
|
969
|
+
* The Gemini Omni SKUs billed per second of video (Gemini Omni Flash 1.2). What
|
|
970
|
+
* the family shares (V2V through the i2v handle, the panels) comes from
|
|
971
|
+
* {@link GEMINI_OMNI_PROVIDERS}. These differ in two ways: the price is a
|
|
972
|
+
* `:<N>s:<res>` ladder, and a run with a source video bills that clip's own
|
|
973
|
+
* seconds and nothing else (the output follows the clip), so they also answer
|
|
974
|
+
* {@link billsReferenceVideoSeconds}.
|
|
975
|
+
*/
|
|
976
|
+
declare const GEMINI_OMNI_PER_SECOND_PROVIDERS: Set<string>;
|
|
977
|
+
declare function isGeminiOmniPerSecondProvider(provider: string | undefined): boolean;
|
|
978
|
+
/** The longest source clip Gemini Omni Flash 1.2 takes (docs.kie.ai/market/google/gemini-omni-flash-1-2). */
|
|
979
|
+
declare const GEMINI_OMNI_12_MAX_SOURCE_SEC = 15;
|
|
980
|
+
/** The widest trim window a request may set on a Gemini Omni source clip, in seconds. */
|
|
981
|
+
declare function videoSourceTrimMaxSec(provider: string | undefined): number;
|
|
956
982
|
/**
|
|
957
983
|
* Video models that are CATALOGUED and priced but have NO working dispatch
|
|
958
984
|
* path today — selecting one fails at the router, not at validation.
|
|
@@ -1288,8 +1314,9 @@ declare function checkRefVideoDurations(provider: string, durationsSec: readonly
|
|
|
1288
1314
|
declare const VIDEO_REF_RATE_PROVIDERS: Set<string>;
|
|
1289
1315
|
/**
|
|
1290
1316
|
* Video models whose reference-VIDEO runs also bill the clips' own seconds:
|
|
1291
|
-
* the Seedance 2 family (`unit × (input + output)`)
|
|
1292
|
-
* VIDEO_REF_RATE_PROVIDERS (Kling 4.0: `unit × (output + max(output, input))`)
|
|
1317
|
+
* the Seedance 2 family (`unit × (input + output)`), the
|
|
1318
|
+
* VIDEO_REF_RATE_PROVIDERS (Kling 4.0: `unit × (output + max(output, input))`)
|
|
1319
|
+
* and the per-second Gemini Omni SKUs (`unit × source clip`, the output follows it).
|
|
1293
1320
|
* The reservation holds the worst case from the probed clips and the
|
|
1294
1321
|
* settlement meters down to what was delivered (backend
|
|
1295
1322
|
* ee/billing/seedance2-ref-video-credits.ts) — this is the one gate every lane
|
|
@@ -1356,8 +1383,8 @@ declare function videoProviderRequiresImage(provider: string | undefined): boole
|
|
|
1356
1383
|
* reference mode takes a lone last frame as the sole reference image plus a
|
|
1357
1384
|
* closing-frame prompt hint. NOTHING else does: VEO ships `[imageUrl,
|
|
1358
1385
|
* endFrameUrl]` verbatim (a lone end frame would reach KIE as `[null, url]`),
|
|
1359
|
-
* the Gemini Omni branch drops the frame when there is no start frame
|
|
1360
|
-
* generic KIE path sends `end_frame` with no image param at all. So the
|
|
1386
|
+
* the per-clip Gemini Omni branch drops the frame when there is no start frame (Omni
|
|
1387
|
+
* Flash 1.2 folds it, with its own builder), and the generic KIE path sends `end_frame` with no image param at all. So the
|
|
1361
1388
|
* exemption is keyed to the FOLD, not to "carries image references" — the
|
|
1362
1389
|
* catalog has no capability for it, and these three family predicates are the
|
|
1363
1390
|
* same membership the provider layer itself branches on (the backend's
|
|
@@ -1702,4 +1729,4 @@ declare function applyDefaultVideoSelection<D>(input: {
|
|
|
1702
1729
|
duration: D | number | undefined;
|
|
1703
1730
|
};
|
|
1704
1731
|
|
|
1705
|
-
export {
|
|
1732
|
+
export { IMAGE_MASK_MODE as $, AI_WRITER_PROVIDERS as A, type FaceSwapProvider as B, CHARACTER_MOTION_PROVIDERS as C, DEFAULT_DIALOGUE_PROVIDER as D, EXTEND_VIDEO_PROVIDERS as E, FACE_SWAP_PROVIDERS as F, GEMINI_OMNI_12_MAX_SOURCE_SEC as G, GEMINI_OMNI_PER_SECOND_PROVIDERS as H, GEMINI_OMNI_PROVIDERS as I, GUIDANCE_SCALE_SUPPORT as J, GVP_ANCHOR_CHOICES as K, GVP_DEFAULT_PROVIDER as L, GVP_END_FRAME_PROVIDERS as M, GVP_EXTEND_PROVIDERS as N, GVP_SUPPORTED_PROVIDERS as O, type GvpAnchorChoice as P, type GvpAnchorWireMode as Q, HIGH_QUALITY_PROVIDERS as R, I2I_MASK_SUPPORT as S, I2I_STRENGTH_SUPPORT as T, IDEOGRAM_PROVIDERS as U, IMAGE_ASPECT_RATIO_VALUES as V, IMAGE_BACKGROUND_VALUES as W, IMAGE_EDIT_PROVIDERS as X, IMAGE_EDIT_TARGET_RESOLUTION_VALUES as Y, IMAGE_GEN_PROVIDERS as Z, IMAGE_I2I_PROVIDERS as _, ASPECT_RATIO_DIMENSIONS as a, SEEDANCE_2_CONTINUATION_REF_SEC as a$, IMAGE_PROMPT_MAX as a0, IMAGE_QUALITY_VALUES as a1, IMAGE_RENDERING_SPEED_VALUES as a2, IMAGE_RESOLUTION_VALUES as a3, IMAGE_TO_VIDEO_PROVIDERS as a4, type ImageAspectRatio as a5, type ImageBackground as a6, type ImageEditProvider as a7, type ImageEditTargetResolution as a8, type ImageGenProvider as a9, MUSIC_PROVIDERS as aA, MUSIC_PROVIDER_LABELS as aB, type ModelPromptingStyle as aC, type ModifyImageProvider as aD, type MotionTransferProviderType as aE, type MusicProvider as aF, NATIVE_ADAPTIVE_ASPECT as aG, NATIVE_NEGATIVE_PROMPT_MODELS as aH, NATIVE_NEGATIVE_VIDEO_PROVIDERS as aI, NEGATIVE_PROMPT_MAX as aJ, OBJECT_MOTION_PROVIDERS as aK, type ObjectMotionProvider as aL, PRICING_DEFAULT_DURATION_SEC as aM, PRICING_DEFAULT_RESOLUTION as aN, PROMPT_HARD_CEILING as aO, QA_CHECK_PROVIDERS as aP, type QaCheckProvider as aQ, REF_IMAGE_MAX_LIMITS as aR, RENDERING_SPEED_SUPPORT as aS, REPLICATE_LIP_SYNC_PROVIDERS as aT, RETIRED_VIDEO_PROVIDERS as aU, type RefAudioCheck as aV, type RefAudioProbe as aW, type RefAudioRule as aX, type RefVideoDurationLimit as aY, SCRIPT_PROVIDERS as aZ, SEEDANCE_2_5_REF_LIMITS as a_, type ImageI2IProvider as aa, type ImageMaskMode as ab, type ImageQuality as ac, type ImageRenderingSpeed as ad, type ImageResolution as ae, type ImageToVideoProvider as af, KLING_4_DEFAULT_RESOLUTION as ag, KLING_4_MAX_MIDDLE_FRAMES as ah, KLING_4_PROVIDERS as ai, LIP_SYNC_PROVIDERS as aj, LLM_STRUCTURED_SYSTEM_MAX as ak, LLM_TEXT_INPUT_MAX as al, LOCATION_ATMOSPHERE_PROVIDERS as am, type LipSyncProvider as an, type LocationAtmosphereProvider as ao, MAX_IMAGE_PROMPT_CHARS_BY_PROVIDER as ap, MAX_NEGATIVE_PROMPT_CHARS_BY_PROVIDER as aq, MAX_REQUEST_IMAGE_LIMIT as ar, MAX_TTS_CHARS_BY_PROVIDER as as, MAX_VIDEO_PROMPT_CHARS_BY_PROVIDER as at, MINIMAX_H3_DEFAULT_RESOLUTION as au, MINIMAX_H3_PROVIDERS as av, MODELS_WITH_REFERENCE_IMAGE_SUPPORT as aw, MODIFY_IMAGE_PROVIDERS as ax, MOTION_TRANSFER_PROVIDERS as ay, MUSIC_CREDIT_ID as az, AUDIO_ADDON_PROVIDERS as b, VIDEO_TO_VIDEO_PROVIDERS as b$, SEEDANCE_2_EXTEND_STITCH as b0, SEEDANCE_2_PROVIDERS as b1, SEEDANCE_2_R2V_MAX_AUDIO_SEC_BY_PROVIDER as b2, SEEDANCE_2_R2V_MIN_REF_VIDEO_SEC as b3, SEEDANCE_2_REF_LIMITS as b4, SEEDANCE_LIP_SYNC_PROVIDERS as b5, SEEDANCE_VIDEO_EDIT_PROVIDERS as b6, SEEDANCE_VIDEO_EDIT_SHAPE as b7, SEED_SUPPORT as b8, SUNO_ACTIVE_MODELS as b9, type TaskChainedEditProvider as bA, type TextToAudioProvider as bB, type TextToVideoProvider as bC, type TranscribeLane as bD, type TranscribeProvider as bE, type TtsOutputFormat as bF, type TtsProvider as bG, UPSCALE_IMAGE_PROVIDERS as bH, type UpscaleImageProvider as bI, VARIABLE_PRICING_MODELS as bJ, VEO_PROVIDERS as bK, VIDEO_AUDIO_CAPABILITY as bL, VIDEO_DURATION_TIERS as bM, VIDEO_GEN_COLLAPSED_T2V_IDS as bN, VIDEO_GEN_PROVIDERS as bO, VIDEO_INPUT_LIP_SYNC_PROVIDERS as bP, VIDEO_MODEL_CAPS as bQ, VIDEO_MODE_ALIASES as bR, VIDEO_PROMPT_MAX as bS, VIDEO_PROVIDERS_REQUIRING_IMAGE as bT, VIDEO_PROVIDERS_WITHOUT_DISPATCH as bU, VIDEO_PROVIDERS_WITH_SEPARATE_FRAME_CHANNEL as bV, VIDEO_REF_AUDIO_RULES as bW, VIDEO_REF_LIMITS_BY_PROVIDER as bX, VIDEO_REF_RATE_PROVIDERS as bY, VIDEO_REF_VIDEO_DURATION_LIMITS as bZ, VIDEO_TO_VIDEO_NODE_PROVIDERS as b_, SUNO_ADD_TRACK_MODELS as ba, SUNO_DURATION_MODELS as bb, SUNO_HARD_CEILING as bc, SUNO_LEGACY_MODELS as bd, SUNO_MODELS as be, SUNO_TEXT_MAX as bf, SUNO_TITLE_MAX as bg, type SceneInputMode as bh, SceneInputModeSchema as bi, type ScriptProvider as bj, type SeedanceVideoEditProvider as bk, type SunoAddTrackModel as bl, type SunoModel as bm, T2I_TO_I2I_VARIANT as bn, TASK_CHAINED_EDIT_PROVIDERS as bo, TASK_CHAINED_EDIT_PROVIDER_IDS as bp, TEXT_TO_AUDIO_PROVIDERS as bq, TEXT_TO_VIDEO_PROVIDERS as br, TRANSCRIBE_LANES as bs, TRANSCRIBE_PROVIDERS as bt, TRANSCRIBE_PROVIDER_CAPABILITIES as bu, TTS_OUTPUT_FORMATS as bv, TTS_OUTPUT_FORMAT_FILES as bw, TTS_PROVIDERS as bx, TTS_TEXT_MAX as by, TWO_K_RESOLUTION_PROVIDERS as bz, type AiWriterProvider as c, seedance2AudioLimitSec as c$, VIDEO_UPSCALE_PROVIDERS as c0, VIDEO_VARIABLE_PRICING as c1, VIDU_Q4_DEFAULT_RESOLUTION as c2, VIDU_Q4_PROVIDERS as c3, VOICE_DESIGN_MODELS as c4, type VideoAudioCapability as c5, type VideoAudioMode as c6, type VideoGenProvider as c7, type VideoModeAlias as c8, type VideoModelCapabilities as c9, isGeminiOmniProvider as cA, isGvpSupportedProvider as cB, isKling4Provider as cC, isLegacySunoModel as cD, isMinimaxH3Provider as cE, isSeedance2Provider as cF, isSeedanceVideoEditProvider as cG, isVeoProvider as cH, isViduQ4Provider as cI, isWan3Provider as cJ, kling4KeyframeFrameCap as cK, maxSegmentSecFor as cL, maxSegmentsFor as cM, maxVideoDurationSec as cN, minSegmentSecFor as cO, modelsForInputMode as cP, normalizeKling4Resolution as cQ, normalizeMinimaxH3Resolution as cR, normalizeViduQ4Resolution as cS, normalizeWan3Resolution as cT, preferredInputModeForModel as cU, pricedOutputDurationSec as cV, requestImageLimit as cW, resolveGvpAnchorWire as cX, resolveMusicProvider as cY, resolveVideoModeForInputs as cZ, resolveVideoProviderForMode as c_, type VideoReferenceSyntax as ca, type VideoToVideoNodeProvider as cb, type VideoToVideoProvider as cc, type VideoUpscaleProvider as cd, type VoiceDesignModel as ce, WAN_3_DEFAULT_RESOLUTION as cf, WAN_3_PROVIDERS as cg, applyDefaultVideoSelection as ch, applyVideoAudioToggle as ci, applyVideoNegativePrompt as cj, billsReferenceVideoSeconds as ck, canonicalVideoProvider as cl, checkRefAudioRules as cm, checkRefVideoDurations as cn, defaultVideoAspectRatio as co, findSeedance2AudioOverLimit as cp, getMaxImagePromptChars as cq, getMaxNegativePromptChars as cr, getMaxSunoPromptChars as cs, getMaxSunoStyleChars as ct, getMaxTtsChars as cu, getMaxVideoPromptChars as cv, getVideoAudioCapability as cw, hasContiguousSegmentDurations as cx, imageReferenceLimit as cy, isGeminiOmniPerSecondProvider as cz, COMPOSER_PLAN_FIELDS as d, segmentDurationsFor as d0, snapToNearestDuration as d1, sunoModelHonoursDuration as d2, supportsAutoVideoDuration as d3, supportsEndAnchor as d4, supportsExtendRender as d5, transcribeLaneSupportsWordTimestamps as d6, transcribeProvidersWithWordTimestamps as d7, videoDefaultDurationSec as d8, videoModelCanSpeakDialogue as d9, videoModelSupportsAudio as da, videoNegativeSuffix as db, videoProviderAppendsFrames as dc, videoProviderFoldsLoneEndFrame as dd, videoProviderRefModeFrameSeats as de, videoProviderReferenceSyntax as df, videoProviderRequiresImage as dg, videoProviderRunsFromReferences as dh, videoSourceTrimMaxSec as di, COMPOSER_PLAN_MAP as e, CREDIT_BASE_USD as f, type CharacterMotionProvider as g, DEFAULT_MUSIC_PROVIDER as h, DEFAULT_REF_IMAGE_MAX as i, DEFAULT_REQUEST_IMAGE_LIMIT as j, DEFAULT_SUNO_MODEL as k, DEFAULT_TEXT_TO_AUDIO_PROVIDER as l, DEFAULT_TRANSCRIBE_NODE_PROVIDER as m, DEFAULT_TRANSCRIBE_PROVIDER as n, DEFAULT_TTS_OUTPUT_FORMAT as o, DEFAULT_TTS_PROVIDER as p, DEFAULT_VIDEO_DURATION_SEC as q, DEFAULT_VIDEO_PROVIDER as r, DEFAULT_VOICE_DESIGN_MODEL as s, DIALOGUE_PROVIDERS as t, DURATION_PRICED_PROVIDERS as u, type DialogueProvider as v, type ExtendVideoProvider as w, FAL_LIP_SYNC_PROVIDERS as x, FLEXIBLE_INPUT_LIP_SYNC_PROVIDERS as y, FRAME_MODE_ADAPTIVE_ONLY_ASPECT as z };
|
package/package.json
CHANGED
|
@@ -111,6 +111,99 @@ describe("autoAspectNeedsSourceImage", () => {
|
|
|
111
111
|
})
|
|
112
112
|
})
|
|
113
113
|
|
|
114
|
+
/**
|
|
115
|
+
* GPT Image 2.6's own "auto" renders a portrait photo square (a 768x1024
|
|
116
|
+
* source came back 1024x1024; GPT Image 2.5's auto kept it at 3:4). So every
|
|
117
|
+
* 2.6 entry keeps "auto" on offer but has it resolved from the photo, exactly
|
|
118
|
+
* as on a model without a native auto, while 2.5 keeps its native auto.
|
|
119
|
+
*/
|
|
120
|
+
describe("GPT Image 2.6 — 'auto' is resolved from the photo, never sent as the model's own", () => {
|
|
121
|
+
const GPT_IMAGE_2_6 = [
|
|
122
|
+
"gpt-image-2-6-flare", "gpt-image-2-6-flare-i2i",
|
|
123
|
+
"gpt-image-2-6-sunburst", "gpt-image-2-6-sunburst-i2i",
|
|
124
|
+
"gpt-image-2-6-economy", "gpt-image-2-6-economy-i2i",
|
|
125
|
+
]
|
|
126
|
+
const portrait = { width: 768, height: 1024 }
|
|
127
|
+
|
|
128
|
+
it("keeps 'auto' on offer in every 2.6 ratio list, so the menus are unchanged", () => {
|
|
129
|
+
for (const id of GPT_IMAGE_2_6) {
|
|
130
|
+
expect(MODEL_CATALOG[id]!.aspectRatios, id).toContain("auto")
|
|
131
|
+
expect(MODEL_CATALOG[id]!.autoAspectFromSourceImage, id).toBe(true)
|
|
132
|
+
}
|
|
133
|
+
})
|
|
134
|
+
|
|
135
|
+
it("asks for the photo's size on every 2.6 id, and not on 2.5 or GPT Image 2", () => {
|
|
136
|
+
for (const id of GPT_IMAGE_2_6) expect(autoAspectNeedsSourceImage(id, "auto"), id).toBe(true)
|
|
137
|
+
for (const id of GPT_IMAGE_2_6) expect(autoAspectNeedsSourceImage(id, "3:4"), id).toBe(false)
|
|
138
|
+
for (const id of ["gpt-image-2-5-flare-i2i", "gpt-image-2-5-sunburst-i2i", "gpt-image-2-5-flare", "gpt-image-2-i2i"]) {
|
|
139
|
+
expect(autoAspectNeedsSourceImage(id, "auto"), id).toBe(false)
|
|
140
|
+
}
|
|
141
|
+
})
|
|
142
|
+
|
|
143
|
+
it.each(["gpt-image-2-6-flare-i2i", "gpt-image-2-6-sunburst-i2i"])("%s: a 3:4 photo is sent 3:4, and says why", (id) => {
|
|
144
|
+
const out = normalizeModelInput(id, { aspectRatio: "auto" }, { sourceImage: portrait })
|
|
145
|
+
expect(out.aspectRatio).toBe("3:4")
|
|
146
|
+
expect(out.adjustments).toHaveLength(1)
|
|
147
|
+
expect(out.adjustments[0]).toMatchObject({ field: "aspectRatio", from: "auto", to: "3:4" })
|
|
148
|
+
expect(out.adjustments[0].reason).toContain(`own "auto" does not keep the photo's shape`)
|
|
149
|
+
})
|
|
150
|
+
|
|
151
|
+
it("Economy picks from its own shorter list", () => {
|
|
152
|
+
// 3:4 is not on Economy's list; 2:3 is the nearest that is.
|
|
153
|
+
expect(normalizeModelInput("gpt-image-2-6-economy-i2i", { aspectRatio: "auto" }, { sourceImage: portrait }).aspectRatio).toBe("2:3")
|
|
154
|
+
})
|
|
155
|
+
|
|
156
|
+
it("never picks a ratio 2.6 refuses (2.5's 27:16, 16:27, 9:8, 8:9)", () => {
|
|
157
|
+
const refused = new Set(["27:16", "16:27", "9:8", "8:9"])
|
|
158
|
+
for (const id of GPT_IMAGE_2_6) {
|
|
159
|
+
for (const sourceImage of [{ width: 2700, height: 1600 }, { width: 1600, height: 2700 }, { width: 1125, height: 1000 }, { width: 1000, height: 1125 }]) {
|
|
160
|
+
const out = normalizeModelInput(id, { aspectRatio: "auto" }, { sourceImage })
|
|
161
|
+
expect(refused.has(out.aspectRatio!), `${id} ${sourceImage.width}x${sourceImage.height} → ${out.aspectRatio}`).toBe(false)
|
|
162
|
+
expect(MODEL_CATALOG[id]!.aspectRatios).toContain(out.aspectRatio)
|
|
163
|
+
}
|
|
164
|
+
}
|
|
165
|
+
})
|
|
166
|
+
|
|
167
|
+
it("with no photo, or one that cannot be read, still sends the model's own 'auto', as before", () => {
|
|
168
|
+
for (const id of GPT_IMAGE_2_6) {
|
|
169
|
+
expect(normalizeModelInput(id, { aspectRatio: "auto" }).aspectRatio, id).toBe("auto")
|
|
170
|
+
expect(normalizeModelInput(id, { aspectRatio: "auto" }, { sourceImage: { width: 0, height: 0 } }).aspectRatio, id).toBe("auto")
|
|
171
|
+
}
|
|
172
|
+
})
|
|
173
|
+
|
|
174
|
+
it("Generate Image with a reference on a 2.6 model resolves 'auto' from the photo; with none it sends 'auto'", () => {
|
|
175
|
+
const withRef = resolveNormalizedImageGen({ provider: "gpt-image-2-6-flare", aspectRatio: "auto", refCount: 1, swapToI2i: true, sourceImage: portrait })
|
|
176
|
+
expect(withRef.modelId).toBe("gpt-image-2-6-flare-i2i")
|
|
177
|
+
expect(withRef.aspectRatio).toBe("3:4")
|
|
178
|
+
expect(imageGenAutoAspectNeedsSourceImage({ provider: "gpt-image-2-6-flare", aspectRatio: "auto", refCount: 1 })).toBe(true)
|
|
179
|
+
const noRef = resolveNormalizedImageGen({ provider: "gpt-image-2-6-flare", aspectRatio: "auto", refCount: 0, swapToI2i: true })
|
|
180
|
+
expect(noRef.aspectRatio).toBe("auto")
|
|
181
|
+
})
|
|
182
|
+
|
|
183
|
+
it("2.5 keeps its native 'auto' even when the photo's size is known", () => {
|
|
184
|
+
const out = normalizeModelInput("gpt-image-2-5-flare-i2i", { aspectRatio: "auto" }, { sourceImage: portrait })
|
|
185
|
+
expect(out.aspectRatio).toBe("auto")
|
|
186
|
+
expect(out.adjustments).toEqual([])
|
|
187
|
+
})
|
|
188
|
+
|
|
189
|
+
it("the measured ratio never moves the price id", () => {
|
|
190
|
+
for (const provider of ["gpt-image-2-6-flare-i2i", "gpt-image-2-6-sunburst-i2i", "gpt-image-2-6-economy-i2i"]) {
|
|
191
|
+
for (const quality of [undefined, "low", "medium", "high", "xhigh", "max", "auto"]) {
|
|
192
|
+
for (const resolution of [undefined, "1K", "2K", "4K"]) {
|
|
193
|
+
for (const refCount of [1, 5, 16]) {
|
|
194
|
+
const base = { provider, quality, resolution, refCount, swapToI2i: false }
|
|
195
|
+
const unset = resolveNormalizedImageGen({ ...base }).identifier
|
|
196
|
+
const auto = resolveNormalizedImageGen({ ...base, aspectRatio: "auto" }).identifier
|
|
197
|
+
const measured = resolveNormalizedImageGen({ ...base, aspectRatio: "auto", sourceImage: portrait })
|
|
198
|
+
expect(measured.identifier).toBe(auto)
|
|
199
|
+
expect(auto).toBe(unset)
|
|
200
|
+
}
|
|
201
|
+
}
|
|
202
|
+
}
|
|
203
|
+
}
|
|
204
|
+
})
|
|
205
|
+
})
|
|
206
|
+
|
|
114
207
|
describe("normalizeModelInput — 'auto' with a source image", () => {
|
|
115
208
|
const portrait = { width: 1104, height: 1472 }
|
|
116
209
|
|
|
@@ -60,6 +60,7 @@ describe("GVP_SUPPORTED_PROVIDERS", () => {
|
|
|
60
60
|
"veo3_lite",
|
|
61
61
|
"gemini-omni-video",
|
|
62
62
|
"gemini-omni-flash",
|
|
63
|
+
"gemini-omni-flash-1-2",
|
|
63
64
|
"kling-4",
|
|
64
65
|
"grok-i2v",
|
|
65
66
|
"wan-3",
|
|
@@ -214,6 +215,7 @@ describe("GVP_END_FRAME_PROVIDERS", () => {
|
|
|
214
215
|
// pro-node SKU any more.
|
|
215
216
|
"veo3.1",
|
|
216
217
|
"veo3_lite",
|
|
218
|
+
"gemini-omni-flash-1-2",
|
|
217
219
|
"wan-3",
|
|
218
220
|
"wan-3-prime",
|
|
219
221
|
])
|