@nodaro/shared 3.24.0 → 3.25.0
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/index.cjs +595 -128
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +375 -48
- package/dist/index.d.ts +375 -48
- package/dist/index.js +553 -129
- package/dist/index.js.map +1 -1
- package/package.json +1 -1
- package/src/__tests__/credit-identifiers.test.ts +19 -6
- package/src/__tests__/dialogue-capabilities.test.ts +3 -2
- package/src/__tests__/edit-plan-minute-ids.test.ts +82 -0
- package/src/__tests__/json-kinds.test.ts +179 -0
- package/src/__tests__/presentation-utils.test.ts +1 -0
- package/src/__tests__/preview-gate.test.ts +2 -2
- package/src/__tests__/pricing-default-duration.test.ts +56 -15
- package/src/__tests__/private-host.test.ts +24 -0
- package/src/__tests__/render-nodes.test.ts +22 -3
- package/src/__tests__/tts-capabilities.test.ts +61 -1
- package/src/__tests__/video-link-input.test.ts +148 -0
- package/src/__tests__/video-link-needs.test.ts +67 -0
- package/src/__tests__/video-request-normalize.test.ts +18 -3
- package/src/credit-identifiers.ts +48 -37
- package/src/edit-plan-contract.ts +67 -1
- package/src/fan-out-rows.ts +2 -2
- package/src/index.ts +45 -1
- package/src/json-kinds.ts +188 -0
- package/src/model-catalog.ts +121 -8
- package/src/model-constants.ts +68 -42
- package/src/node-default-mappings.ts +1 -0
- package/src/pipeline-types.ts +13 -2
- package/src/presentation-utils.ts +8 -1
- package/src/private-host.ts +91 -0
- package/src/producer-types.ts +3 -0
- package/src/render-nodes.ts +12 -1
- package/src/scene-node-types.ts +17 -5
- package/src/social-post-readers.ts +37 -0
- package/src/video-link.ts +238 -0
- package/src/voices.ts +16 -9
|
@@ -0,0 +1,188 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* What a JSON pip CARRIES — the handle typing behind "an EDL is not a
|
|
3
|
+
* transcript" (decided 2026-10-07) and its reverse, "a transcript is not an
|
|
4
|
+
* EDL" (decided 2026-10-08).
|
|
5
|
+
*
|
|
6
|
+
* Every JSON output colours the same "look" data pip, so a validator that sees
|
|
7
|
+
* only the source NODE type cannot tell Edit Plan's edit list from a
|
|
8
|
+
* Transcribe's transcript. Wired into Add Captions' Transcript input, the edit
|
|
9
|
+
* list is accepted at connect time and only fails (or captions nothing) at run
|
|
10
|
+
* time. This module is the one place that knows the difference, read by the
|
|
11
|
+
* canvas validators and the server's edge normalization alike (MCP workflow
|
|
12
|
+
* writes, the REST delta lane, the Copilot).
|
|
13
|
+
*
|
|
14
|
+
* Output kinds are DATA, never a list of node names:
|
|
15
|
+
* - a render node's `json` pip answers its registry `jsonKind`
|
|
16
|
+
* (`RENDER_NODE_TYPES`), so a render registered later — Speaker View, whose
|
|
17
|
+
* json is an EDL — is covered by its own registry entry;
|
|
18
|
+
* - the other JSON producers declare a kind per output pip in
|
|
19
|
+
* `JSON_OUTPUT_KINDS` below. EVERY node with a JSON output declares one
|
|
20
|
+
* (decided 2026-10-08): `transcript`, `edl`, or `other` for JSON that is
|
|
21
|
+
* neither (a scrape, an analysis, a collection). `other` is a positive
|
|
22
|
+
* statement, not a gap — it connects exactly as an undeclared pip always
|
|
23
|
+
* has. A guard test enumerates the node definitions' JSON pips and fails
|
|
24
|
+
* on one with no declaration, so a new JSON node cannot ship unclassified
|
|
25
|
+
* and a Transcript producer cannot slip past the EDL/Transcript block.
|
|
26
|
+
* Input kinds follow the platform's handle vocabulary: a `transcript` pip
|
|
27
|
+
* expects a Transcript, an `edl` pip an EDL.
|
|
28
|
+
*/
|
|
29
|
+
import { RENDER_NODE_TYPES, type RenderNodeDescriptor } from "./render-nodes.js"
|
|
30
|
+
|
|
31
|
+
/** The two JSON shapes the platform's handle vocabulary tells apart. */
|
|
32
|
+
export type JsonKind = "transcript" | "edl"
|
|
33
|
+
|
|
34
|
+
/**
|
|
35
|
+
* What a JSON output pip DECLARES: one of the two kinds, or `other` — JSON that
|
|
36
|
+
* is neither a Transcript nor an EDL. Only `JsonKind` ever takes part in a
|
|
37
|
+
* connection check; `other` records that the pip was classified.
|
|
38
|
+
*/
|
|
39
|
+
export type JsonOutputDeclaration = JsonKind | "other"
|
|
40
|
+
|
|
41
|
+
interface JsonOutputs {
|
|
42
|
+
/** The pip an edge with no `sourceHandle` reads — absent when that read is not JSON (Text to Dialogue's is its audio). */
|
|
43
|
+
readonly primary?: string
|
|
44
|
+
readonly pips: Readonly<Record<string, JsonOutputDeclaration>>
|
|
45
|
+
}
|
|
46
|
+
|
|
47
|
+
/** JSON outputs of the nodes that are not renders: pip id → what it carries. */
|
|
48
|
+
const JSON_OUTPUT_KINDS: Readonly<Record<string, JsonOutputs>> = Object.freeze({
|
|
49
|
+
// Edit Plan: the plan on `edl` — a Tighten EDL, a clip set (`Edl[]`) or a chapter list.
|
|
50
|
+
"edit-plan": { primary: "edl", pips: { edl: "edl" } },
|
|
51
|
+
// Transcribe: its json is the normalized Transcript (its `text` pip is plain text, no kind).
|
|
52
|
+
transcribe: { primary: "json", pips: { json: "transcript" } },
|
|
53
|
+
// Camera Switch: the switched edit and the renamed (still a Transcript) transcript.
|
|
54
|
+
"camera-switch": { primary: "edl", pips: { edl: "edl", transcript: "transcript" } },
|
|
55
|
+
// Text to Dialogue: the Transcript built from the model's own word timings (its default read is the audio).
|
|
56
|
+
"text-to-dialogue": { pips: { json: "transcript" } },
|
|
57
|
+
|
|
58
|
+
// ── JSON that is neither a Transcript nor an EDL (decided 2026-10-08) ──
|
|
59
|
+
// Forced Alignment: `{ alignment: [{ word, start, end }] }` in seconds — a word list, not a Transcript.
|
|
60
|
+
"forced-alignment": { primary: "data", pips: { data: "other" } },
|
|
61
|
+
// Video Analysis / Video Audit: the canonical analysis result (an audit IS an analysis).
|
|
62
|
+
"video-analysis": { primary: "json", pips: { json: "other" } },
|
|
63
|
+
"video-audit": { primary: "json", pips: { json: "other" } },
|
|
64
|
+
// Silence Detect: `{ version, ranges, durationMs }`. Audio Sync: `{ version, reference, offsets, notes }`.
|
|
65
|
+
"silence-detect": { primary: "json", pips: { json: "other" } },
|
|
66
|
+
"audio-sync": { primary: "json", pips: { json: "other" } },
|
|
67
|
+
// Scrapers and searches: the scraped page / ad array / post list.
|
|
68
|
+
"web-scrape": { primary: "json", pips: { json: "other" } },
|
|
69
|
+
"meta-ads-scrape": { primary: "json", pips: { json: "other" } },
|
|
70
|
+
"instagram-scrape": { primary: "json", pips: { json: "other" } },
|
|
71
|
+
"social-search": { primary: "json", pips: { json: "other" } },
|
|
72
|
+
"telegram-channel-feed": { primary: "json", pips: { json: "other" } },
|
|
73
|
+
// Content Recipe: the recipe's structured result.
|
|
74
|
+
"content-recipe": { primary: "json", pips: { json: "other" } },
|
|
75
|
+
// Collections and libraries: the stored rows, whatever the user keeps in them.
|
|
76
|
+
"collection-read": { primary: "json", pips: { json: "other" } },
|
|
77
|
+
"collection-write": { primary: "json", pips: { json: "other" } },
|
|
78
|
+
"inspiration-read": { primary: "json", pips: { json: "other" } },
|
|
79
|
+
"competitor-read": { primary: "json", pips: { json: "other" } },
|
|
80
|
+
// Describe to Picker: the picker values read off an image.
|
|
81
|
+
"describe-to-picker": { primary: "picker-json", pips: { "picker-json": "other" } },
|
|
82
|
+
})
|
|
83
|
+
|
|
84
|
+
/** The pip a render's order reads/writes its json on. */
|
|
85
|
+
const RENDER_JSON_PIP = "json"
|
|
86
|
+
|
|
87
|
+
/** The input pips whose kind is fixed by the handle vocabulary. */
|
|
88
|
+
const JSON_INPUT_KINDS: Readonly<Record<string, JsonKind>> = Object.freeze({
|
|
89
|
+
transcript: "transcript",
|
|
90
|
+
edl: "edl",
|
|
91
|
+
})
|
|
92
|
+
|
|
93
|
+
const own = <T>(table: Readonly<Record<string, T>>, key: string): T | undefined =>
|
|
94
|
+
Object.hasOwn(table, key) ? table[key] : undefined
|
|
95
|
+
|
|
96
|
+
/**
|
|
97
|
+
* What the JSON output `sourceHandle` of a node of `nodeType` DECLARES —
|
|
98
|
+
* `transcript`, `edl`, `other` — or `undefined` when the pip is not declared at
|
|
99
|
+
* all (a non-JSON pip, an unknown handle, an unregistered node). `renders` is
|
|
100
|
+
* the render-node registry (injectable for tests). The guard test reads this to
|
|
101
|
+
* prove every JSON pip is classified.
|
|
102
|
+
*/
|
|
103
|
+
export function declaredJsonOutput(
|
|
104
|
+
nodeType: string | null | undefined,
|
|
105
|
+
sourceHandle: string | null | undefined,
|
|
106
|
+
renders: Readonly<Record<string, RenderNodeDescriptor>> = RENDER_NODE_TYPES,
|
|
107
|
+
): JsonOutputDeclaration | undefined {
|
|
108
|
+
if (typeof nodeType !== "string") return undefined
|
|
109
|
+
const render = own(renders, nodeType)
|
|
110
|
+
if (render) return (sourceHandle ?? RENDER_JSON_PIP) === RENDER_JSON_PIP ? render.jsonKind : undefined
|
|
111
|
+
const outputs = own(JSON_OUTPUT_KINDS, nodeType)
|
|
112
|
+
if (!outputs) return undefined
|
|
113
|
+
const pip = sourceHandle ?? outputs.primary
|
|
114
|
+
return typeof pip === "string" ? own(outputs.pips, pip) : undefined
|
|
115
|
+
}
|
|
116
|
+
|
|
117
|
+
/** The declared JSON output pips of the non-render nodes, as `[nodeType, pipId, declaration]` rows. */
|
|
118
|
+
export function declaredJsonOutputRows(): ReadonlyArray<readonly [string, string, JsonOutputDeclaration]> {
|
|
119
|
+
return Object.entries(JSON_OUTPUT_KINDS).flatMap(([type, o]) =>
|
|
120
|
+
Object.entries(o.pips).map(([pip, kind]) => [type, pip, kind] as const),
|
|
121
|
+
)
|
|
122
|
+
}
|
|
123
|
+
|
|
124
|
+
/**
|
|
125
|
+
* What the JSON output `sourceHandle` of a node of `nodeType` carries, or
|
|
126
|
+
* `undefined` when it is not known to carry one of the two kinds (anything
|
|
127
|
+
* declared `other`, a non-JSON pip, anything unregistered).
|
|
128
|
+
*/
|
|
129
|
+
export function jsonOutputKind(
|
|
130
|
+
nodeType: string | null | undefined,
|
|
131
|
+
sourceHandle: string | null | undefined,
|
|
132
|
+
renders: Readonly<Record<string, RenderNodeDescriptor>> = RENDER_NODE_TYPES,
|
|
133
|
+
): JsonKind | undefined {
|
|
134
|
+
const declared = declaredJsonOutput(nodeType, sourceHandle, renders)
|
|
135
|
+
return declared === "other" ? undefined : declared
|
|
136
|
+
}
|
|
137
|
+
|
|
138
|
+
/** What the JSON input `targetHandle` of a node expects, or `undefined`. */
|
|
139
|
+
export function jsonInputKind(
|
|
140
|
+
_nodeType: string | null | undefined,
|
|
141
|
+
targetHandle: string | null | undefined,
|
|
142
|
+
): JsonKind | undefined {
|
|
143
|
+
return typeof targetHandle === "string" ? own(JSON_INPUT_KINDS, targetHandle) : undefined
|
|
144
|
+
}
|
|
145
|
+
|
|
146
|
+
export interface JsonKindMismatch {
|
|
147
|
+
readonly output: JsonKind
|
|
148
|
+
readonly input: JsonKind
|
|
149
|
+
}
|
|
150
|
+
|
|
151
|
+
/**
|
|
152
|
+
* The connections this blocks: a JSON output of one kind wired into an input of
|
|
153
|
+
* the other (decided 2026-10-07 / 2026-10-08) — an EDL into a Transcript input
|
|
154
|
+
* and a Transcript into an EDL input. Both directions read the same two
|
|
155
|
+
* tables. `null` for every other pair — including pairs it cannot classify,
|
|
156
|
+
* which keep today's behaviour.
|
|
157
|
+
*/
|
|
158
|
+
export function jsonKindMismatch(
|
|
159
|
+
sourceType: string | null | undefined,
|
|
160
|
+
sourceHandle: string | null | undefined,
|
|
161
|
+
targetType: string | null | undefined,
|
|
162
|
+
targetHandle: string | null | undefined,
|
|
163
|
+
renders: Readonly<Record<string, RenderNodeDescriptor>> = RENDER_NODE_TYPES,
|
|
164
|
+
): JsonKindMismatch | null {
|
|
165
|
+
const input = jsonInputKind(targetType, targetHandle)
|
|
166
|
+
if (!input) return null
|
|
167
|
+
const output = jsonOutputKind(sourceType, sourceHandle, renders)
|
|
168
|
+
return output && output !== input ? { output, input } : null
|
|
169
|
+
}
|
|
170
|
+
|
|
171
|
+
/** The sentence a rejected connection is explained with. */
|
|
172
|
+
export function jsonKindMismatchMessage(
|
|
173
|
+
mismatch: JsonKindMismatch,
|
|
174
|
+
labels: { readonly sourceLabel?: string; readonly targetLabel?: string } = {},
|
|
175
|
+
): string {
|
|
176
|
+
const source = labels.sourceLabel ?? "This output"
|
|
177
|
+
const target = labels.targetLabel ?? "this node"
|
|
178
|
+
if (mismatch.output === "transcript") {
|
|
179
|
+
return (
|
|
180
|
+
`${source} outputs a transcript, not an edit list (EDL), so it cannot feed the EDL input of ${target}. ` +
|
|
181
|
+
`Wire an Edit Plan's or Camera Switch's edl there instead.`
|
|
182
|
+
)
|
|
183
|
+
}
|
|
184
|
+
return (
|
|
185
|
+
`${source} outputs an edit list (EDL), not a transcript, so it cannot feed the Transcript input of ${target}. ` +
|
|
186
|
+
`Wire a Transcribe node's json (or Apply EDL's or Camera Switch's transcript) there instead.`
|
|
187
|
+
)
|
|
188
|
+
}
|
package/src/model-catalog.ts
CHANGED
|
@@ -220,6 +220,14 @@ export interface ModelCatalogEntry {
|
|
|
220
220
|
unlistedResolutionRendersAs?: string
|
|
221
221
|
qualities?: readonly string[]
|
|
222
222
|
durations?: readonly number[]
|
|
223
|
+
/**
|
|
224
|
+
* The length (seconds) this model renders when the request names no duration
|
|
225
|
+
* — a capability fact, and the length an unset duration is CHARGED at
|
|
226
|
+
* (`pricedOutputDurationSec`). A member of `durations`. REQUIRED on every
|
|
227
|
+
* video model priced by length (`pricing-default-duration-sync.test.ts`), and
|
|
228
|
+
* where the KIE config pins `extraParams.duration` it must equal that pin.
|
|
229
|
+
*/
|
|
230
|
+
defaultDuration?: number
|
|
223
231
|
/**
|
|
224
232
|
* The model accepts an AUTO duration (`duration: -1`, see
|
|
225
233
|
* `VIDEO_DURATION_AUTO`): it picks the clip length itself — the source clip's
|
|
@@ -288,7 +296,7 @@ export const MODEL_RECOMMENDATIONS: readonly ModelRecommendation[] = [
|
|
|
288
296
|
{ intent: "cheap batch video clips", modelIds: ["veo3.1", "wan-turbo", "bytedance-lite"], note: "VEO 3.1 Fast is the best price/quality balance with native audio." },
|
|
289
297
|
{ intent: "video with start + end frame", modelIds: ["veo3", "veo3.1", "kling-turbo", "minimax", "hailuo-standard", "seedance-2"], note: "All listed support an end frame; VEO uses imageUrls[start, end]." },
|
|
290
298
|
{ intent: "music / song generation", modelIds: ["suno-v6", "suno-v6_wild", "suno-v6_mini", "suno-v5_5"], note: "V6 is the default flagship; V6 Wild for bolder, less predictable results; V6 Mini when speed matters; v5.5 / v5 / v4 keep their own character. Same price." },
|
|
291
|
-
{ intent: "voice over / narration", modelIds: ["elevenlabs-v4", "elevenlabs-v3", "elevenlabs-turbo"], note: "v4 is the default: [audio tags] for emotion and up to 10,000 characters per request. v3 is the previous expressive model, still selectable. Turbo is
|
|
299
|
+
{ intent: "voice over / narration", modelIds: ["elevenlabs-v4", "elevenlabs-v3", "elevenlabs-v4-turbo", "elevenlabs-turbo"], note: "v4 is the default: [audio tags] for emotion and up to 10,000 characters per request. v3 is the previous expressive model, still selectable. v4 Turbo keeps v4's tags at Turbo's credit price; Turbo v2.5 is the cheapest for plain narration without tags." },
|
|
292
300
|
{ intent: "lip-sync a portrait to audio", modelIds: ["kling-avatar-pro", "kling-avatar", "infinitalk"], note: "Pro for best mouth shape; InfiniTalk for resolution control." },
|
|
293
301
|
{ intent: "transcription / captions", modelIds: ["elevenlabs-stt", "incredibly-fast-whisper", "whisper"], note: "Captions need WORD timestamps: ElevenLabs STT (always) or Incredibly Fast Whisper. Plain Whisper returns phrase segments only." },
|
|
294
302
|
{ intent: "motion transfer (drive a subject by another video)", modelIds: ["motion-transfer", "kling-3.0-motion"], note: "Kling 2.6 base is cheap; Kling 3.0 is premium." },
|
|
@@ -304,6 +312,10 @@ const NANO_BANANA_RATIOS = ["1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "
|
|
|
304
312
|
// nano-banana-2-lite). Kept separate so the wider set can't leak to the
|
|
305
313
|
// rest of the family.
|
|
306
314
|
const NANO_BANANA_2_LITE_RATIOS = ["auto", "1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "4:5", "5:4", "21:9", "4:1", "1:4", "8:1", "1:8"] as const
|
|
315
|
+
// Nano Banana 2.1 — its own documented enum, the same fifteen ratios as Lite
|
|
316
|
+
// (docs.kie.ai/market/google/nanobanana-2-1). Declared separately so a change to
|
|
317
|
+
// either model's set cannot silently move the other.
|
|
318
|
+
const NANO_BANANA_2_1_RATIOS = ["auto", "1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3", "4:5", "5:4", "21:9", "4:1", "1:4", "8:1", "1:8"] as const
|
|
307
319
|
// Note: these arrays are synchronized with the frontend's
|
|
308
320
|
// `model-options.ts` constants of the same names. The frontend now imports
|
|
309
321
|
// them via the `getAspectRatioOptions(modelId)` helper; if you update one
|
|
@@ -389,6 +401,26 @@ const IMAGE_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
389
401
|
{ identifier: "nano-banana-2:4K", credits: 50, note: "4K" },
|
|
390
402
|
],
|
|
391
403
|
},
|
|
404
|
+
"nano-banana-2-1": {
|
|
405
|
+
id: "nano-banana-2-1",
|
|
406
|
+
kind: "image",
|
|
407
|
+
// Single id for both modes, like nano-banana-2: optional input images via
|
|
408
|
+
// image_input (up to 10) on the one endpoint.
|
|
409
|
+
modes: ["t2i", "i2i"] as const,
|
|
410
|
+
family: "Google",
|
|
411
|
+
label: "Nano Banana 2.1",
|
|
412
|
+
series: "Nano Banana",
|
|
413
|
+
description: "Google's high-efficiency image model for generation and editing — sharper quality, prompt adherence, character consistency and text rendering, up to 10 references and 4K output.",
|
|
414
|
+
useCases: ["realistic", "character", "typography", "high-res"],
|
|
415
|
+
features: ["reference-image"],
|
|
416
|
+
aspectRatios: NANO_BANANA_2_1_RATIOS,
|
|
417
|
+
resolutions: ["1K", "2K", "4K"],
|
|
418
|
+
pricing: [
|
|
419
|
+
{ identifier: "nano-banana-2-1", credits: 10, note: "1K default" },
|
|
420
|
+
{ identifier: "nano-banana-2-1:2K", credits: 20, note: "2K" },
|
|
421
|
+
{ identifier: "nano-banana-2-1:4K", credits: 30, note: "4K" },
|
|
422
|
+
],
|
|
423
|
+
},
|
|
392
424
|
"nano-banana-2-lite": {
|
|
393
425
|
id: "nano-banana-2-lite",
|
|
394
426
|
kind: "image",
|
|
@@ -1201,6 +1233,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1201
1233
|
useCases: ["realistic", "motion", "narrative"],
|
|
1202
1234
|
features: ["end-frame"],
|
|
1203
1235
|
durations: [5],
|
|
1236
|
+
defaultDuration: 5,
|
|
1204
1237
|
pricing: [{ identifier: "minimax", credits: 143, note: "5s, 1080p" }],
|
|
1205
1238
|
},
|
|
1206
1239
|
// MiniMax Hailuo 3 (KIE "minimax-h3") — t2v / i2v (first and/or last frame) /
|
|
@@ -1226,6 +1259,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1226
1259
|
features: ["end-frame", "audio", "reference-image", "video-reference"],
|
|
1227
1260
|
aspectRatios: VIDEO_RATIOS_SEEDANCE_2,
|
|
1228
1261
|
durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1262
|
+
defaultDuration: 6,
|
|
1229
1263
|
// First entry = default (the UI dropdown + fail-safe snap use opts[0], and
|
|
1230
1264
|
// 2K is the KIE-side default). Uppercase values are the exact KIE wire enum.
|
|
1231
1265
|
resolutions: ["2K", "768P"],
|
|
@@ -1251,6 +1285,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1251
1285
|
description: "Hailuo 2.3 Pro — newer Hailuo with 768P / 1080P resolutions.",
|
|
1252
1286
|
useCases: ["realistic", "motion"],
|
|
1253
1287
|
durations: [6, 10],
|
|
1288
|
+
defaultDuration: 6,
|
|
1254
1289
|
resolutions: ["768P", "1080P"],
|
|
1255
1290
|
pricing: [
|
|
1256
1291
|
{ identifier: "hailuo-2.3-pro", credits: 200, note: "10s default" },
|
|
@@ -1270,6 +1305,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1270
1305
|
description: "Cheaper Hailuo 2.3 tier — good baseline quality.",
|
|
1271
1306
|
useCases: ["realistic", "cheap"],
|
|
1272
1307
|
durations: [6, 10],
|
|
1308
|
+
defaultDuration: 6,
|
|
1273
1309
|
pricing: [
|
|
1274
1310
|
{ identifier: "hailuo-2.3", credits: 75, note: "6s default" },
|
|
1275
1311
|
{ identifier: "hailuo-2.3:6s", credits: 80, note: "6s" },
|
|
@@ -1287,6 +1323,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1287
1323
|
useCases: ["cheap", "motion"],
|
|
1288
1324
|
features: ["end-frame"],
|
|
1289
1325
|
durations: [6, 10],
|
|
1326
|
+
defaultDuration: 6,
|
|
1290
1327
|
resolutions: ["512P", "768P"],
|
|
1291
1328
|
pricing: [
|
|
1292
1329
|
{ identifier: "hailuo-standard", credits: 75, note: "6s default" },
|
|
@@ -1315,6 +1352,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1315
1352
|
// `veo3_lite` DO serve it and keep the flag.
|
|
1316
1353
|
features: ["end-frame", "audio"],
|
|
1317
1354
|
durations: [4, 6, 8],
|
|
1355
|
+
defaultDuration: 8,
|
|
1318
1356
|
aspectRatios: VIDEO_RATIOS_HV,
|
|
1319
1357
|
// 720p (default) + 1080p inline. 4K generates the base at 1080p then chains
|
|
1320
1358
|
// KIE's /api/v1/veo/get-4k-video endpoint (worker handleImageToVideo / handleTextToVideo).
|
|
@@ -1336,6 +1374,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1336
1374
|
useCases: ["cinematic", "fast", "general"],
|
|
1337
1375
|
features: ["end-frame", "audio", "reference-image"],
|
|
1338
1376
|
durations: [4, 6, 8],
|
|
1377
|
+
defaultDuration: 8,
|
|
1339
1378
|
aspectRatios: VIDEO_RATIOS_HV,
|
|
1340
1379
|
resolutions: ["720p", "1080p", "4k"],
|
|
1341
1380
|
pricing: [
|
|
@@ -1356,6 +1395,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1356
1395
|
useCases: ["cinematic", "cheap", "high-volume"],
|
|
1357
1396
|
features: ["end-frame", "audio", "reference-image"],
|
|
1358
1397
|
durations: [4, 6, 8],
|
|
1398
|
+
defaultDuration: 8,
|
|
1359
1399
|
aspectRatios: VIDEO_RATIOS_HV,
|
|
1360
1400
|
resolutions: ["720p", "1080p", "4k"],
|
|
1361
1401
|
pricing: [
|
|
@@ -1378,6 +1418,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1378
1418
|
aspectRatios: VIDEO_RATIOS_HV,
|
|
1379
1419
|
resolutions: ["720p", "1080p", "4k"],
|
|
1380
1420
|
durations: [4, 6, 8, 10],
|
|
1421
|
+
defaultDuration: 8,
|
|
1381
1422
|
pricing: [
|
|
1382
1423
|
{ identifier: "gemini-omni-video", credits: 315, note: "default — 720p/1080p 4s" },
|
|
1383
1424
|
{ identifier: "gemini-omni-video:4", credits: 230, note: "720p/1080p 4s" },
|
|
@@ -1413,6 +1454,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1413
1454
|
aspectRatios: VIDEO_RATIOS_HV,
|
|
1414
1455
|
resolutions: ["720p", "1080p", "4k"],
|
|
1415
1456
|
durations: [4, 6, 8, 10],
|
|
1457
|
+
defaultDuration: 8,
|
|
1416
1458
|
pricing: [
|
|
1417
1459
|
{ identifier: "gemini-omni-flash", credits: 270, note: "default — 8s 720p/1080p (the credit builder's duration fallback)" },
|
|
1418
1460
|
{ identifier: "gemini-omni-flash:4", credits: 160, note: "720p/1080p 4s" },
|
|
@@ -1441,6 +1483,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1441
1483
|
features: ["audio"],
|
|
1442
1484
|
aspectRatios: VIDEO_RATIOS_HVS,
|
|
1443
1485
|
durations: [5, 10],
|
|
1486
|
+
defaultDuration: 5,
|
|
1444
1487
|
pricing: [
|
|
1445
1488
|
{ identifier: "kling", credits: 280, note: "10s no audio default" },
|
|
1446
1489
|
{ identifier: "kling:5s", credits: 138, note: "5s no audio" },
|
|
@@ -1460,6 +1503,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1460
1503
|
useCases: ["fast", "motion", "cheap"],
|
|
1461
1504
|
features: ["end-frame"],
|
|
1462
1505
|
durations: [5, 10],
|
|
1506
|
+
defaultDuration: 5,
|
|
1463
1507
|
pricing: [
|
|
1464
1508
|
{ identifier: "kling-turbo", credits: 125, note: "5s default" },
|
|
1465
1509
|
{ identifier: "kling-turbo:5s", credits: 110 },
|
|
@@ -1477,6 +1521,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1477
1521
|
useCases: ["premium", "narrative", "cinematic"],
|
|
1478
1522
|
features: ["end-frame", "audio"],
|
|
1479
1523
|
durations: [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1524
|
+
defaultDuration: 5,
|
|
1480
1525
|
resolutions: ["720P", "1080P"],
|
|
1481
1526
|
pricing: [
|
|
1482
1527
|
{ identifier: "kling-3.0", credits: 500, note: "5s 1080p with audio" },
|
|
@@ -1499,6 +1544,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1499
1544
|
description: "Master tier I2V — strong cinematic quality.",
|
|
1500
1545
|
useCases: ["cinematic", "premium"],
|
|
1501
1546
|
durations: [5, 10],
|
|
1547
|
+
defaultDuration: 5,
|
|
1502
1548
|
pricing: [
|
|
1503
1549
|
{ identifier: "kling-master", credits: 400, note: "5s default" },
|
|
1504
1550
|
{ identifier: "kling-master:5s", credits: 400 },
|
|
@@ -1516,6 +1562,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1516
1562
|
useCases: ["motion", "cinematic", "reference"],
|
|
1517
1563
|
features: ["end-frame", "reference-image"],
|
|
1518
1564
|
durations: [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1565
|
+
defaultDuration: 5,
|
|
1519
1566
|
resolutions: ["720p", "1080p"],
|
|
1520
1567
|
aspectRatios: VIDEO_RATIOS_HVS,
|
|
1521
1568
|
pricing: [
|
|
@@ -1538,6 +1585,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1538
1585
|
description: "Grok image-to-video — stylized motion. Up to 15s.",
|
|
1539
1586
|
useCases: ["stylized", "motion"],
|
|
1540
1587
|
durations: [6, 10],
|
|
1588
|
+
defaultDuration: 6,
|
|
1541
1589
|
resolutions: ["480p", "720p"],
|
|
1542
1590
|
pricing: [
|
|
1543
1591
|
{ identifier: "grok-i2v", credits: 150, note: "6s default" },
|
|
@@ -1566,6 +1614,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1566
1614
|
useCases: ["stylized", "motion"],
|
|
1567
1615
|
aspectRatios: ["16:9", "9:16", "1:1", "4:3", "3:4"],
|
|
1568
1616
|
durations: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1617
|
+
defaultDuration: 8,
|
|
1569
1618
|
resolutions: ["480p", "720p"],
|
|
1570
1619
|
pricing: [
|
|
1571
1620
|
{ identifier: "grok-imagine-video-1.5", credits: 295, note: "default 8s 480p — see :Ns:res variants" },
|
|
@@ -1588,6 +1637,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1588
1637
|
useCases: ["motion", "narrative"],
|
|
1589
1638
|
features: ["end-frame"],
|
|
1590
1639
|
durations: [4, 8, 12],
|
|
1640
|
+
defaultDuration: 8,
|
|
1591
1641
|
resolutions: ["480p", "720p", "1080p"],
|
|
1592
1642
|
pricing: [
|
|
1593
1643
|
{ identifier: "seedance", credits: 250, note: "8s default" },
|
|
@@ -1609,6 +1659,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1609
1659
|
features: ["end-frame", "audio", "reference-image", "video-reference"],
|
|
1610
1660
|
aspectRatios: VIDEO_RATIOS_SEEDANCE_2,
|
|
1611
1661
|
durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1662
|
+
defaultDuration: 8,
|
|
1612
1663
|
autoDuration: true,
|
|
1613
1664
|
resolutions: ["480p", "720p", "1080p", "4k"],
|
|
1614
1665
|
pricing: [
|
|
@@ -1635,6 +1686,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1635
1686
|
features: ["end-frame", "audio", "reference-image", "video-reference"],
|
|
1636
1687
|
aspectRatios: VIDEO_RATIOS_SEEDANCE_2,
|
|
1637
1688
|
durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1689
|
+
defaultDuration: 8,
|
|
1638
1690
|
autoDuration: true,
|
|
1639
1691
|
resolutions: ["480p", "720p"],
|
|
1640
1692
|
pricing: [
|
|
@@ -1657,6 +1709,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1657
1709
|
features: ["end-frame", "audio", "reference-image", "video-reference"],
|
|
1658
1710
|
aspectRatios: VIDEO_RATIOS_SEEDANCE_2,
|
|
1659
1711
|
durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1712
|
+
defaultDuration: 8,
|
|
1660
1713
|
autoDuration: true,
|
|
1661
1714
|
resolutions: ["480p", "720p"],
|
|
1662
1715
|
pricing: [
|
|
@@ -1685,6 +1738,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1685
1738
|
features: ["end-frame", "audio", "reference-image", "video-reference"],
|
|
1686
1739
|
aspectRatios: VIDEO_RATIOS_SEEDANCE_2,
|
|
1687
1740
|
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],
|
|
1741
|
+
defaultDuration: 8,
|
|
1688
1742
|
autoDuration: true,
|
|
1689
1743
|
resolutions: ["480p", "720p", "1080p"],
|
|
1690
1744
|
pricing: [
|
|
@@ -1727,6 +1781,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1727
1781
|
features: ["end-frame", "audio", "reference-image"],
|
|
1728
1782
|
aspectRatios: VIDEO_RATIOS_WAN_3,
|
|
1729
1783
|
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],
|
|
1784
|
+
defaultDuration: 5,
|
|
1730
1785
|
resolutions: ["480p", "720p", "1080p"],
|
|
1731
1786
|
pricing: [
|
|
1732
1787
|
{ identifier: "wan-3", credits: 200, note: "default 5s 720p — see :Ns:res variants for exact" },
|
|
@@ -1752,6 +1807,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1752
1807
|
features: ["end-frame", "audio", "reference-image"],
|
|
1753
1808
|
aspectRatios: VIDEO_RATIOS_WAN_3,
|
|
1754
1809
|
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],
|
|
1810
|
+
defaultDuration: 5,
|
|
1755
1811
|
resolutions: ["480p", "720p", "1080p"],
|
|
1756
1812
|
pricing: [
|
|
1757
1813
|
{ identifier: "wan-3-prime", credits: 320, note: "default 5s 720p — see :Ns:res variants for exact" },
|
|
@@ -1774,6 +1830,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1774
1830
|
description: "Wan 2.6 image-to-video — 5/10/15s at 720p/1080p.",
|
|
1775
1831
|
useCases: ["motion", "narrative"],
|
|
1776
1832
|
durations: [5, 10, 15],
|
|
1833
|
+
defaultDuration: 5,
|
|
1777
1834
|
resolutions: ["720p", "1080p"],
|
|
1778
1835
|
pricing: [
|
|
1779
1836
|
{ identifier: "wan-i2v", credits: 175, note: "5s 720p default" },
|
|
@@ -1793,6 +1850,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1793
1850
|
useCases: ["cheap", "fast"],
|
|
1794
1851
|
aspectRatios: VIDEO_RATIOS_HVS,
|
|
1795
1852
|
durations: [5],
|
|
1853
|
+
defaultDuration: 5,
|
|
1796
1854
|
resolutions: ["480p", "720p"],
|
|
1797
1855
|
// Different KIE endpoints (i2v vs t2v) → different costs under
|
|
1798
1856
|
// composite ids. Route picks endpoint based on whether image was supplied.
|
|
@@ -1813,6 +1871,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1813
1871
|
description: "Wan 2.6 — text-to-video and video-to-video under a single id.",
|
|
1814
1872
|
useCases: ["v2v", "t2v", "restyle"],
|
|
1815
1873
|
durations: [5],
|
|
1874
|
+
defaultDuration: 5,
|
|
1816
1875
|
resolutions: ["720p", "1080p"],
|
|
1817
1876
|
pricing: [
|
|
1818
1877
|
{ identifier: "wan", credits: 175, note: "v2v 5s 720p" },
|
|
@@ -1888,6 +1947,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1888
1947
|
description: "Wan 2.7 image-to-video — 2–15s at 720p/1080p, supports start+end frame.",
|
|
1889
1948
|
useCases: ["motion", "narrative"],
|
|
1890
1949
|
durations: [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1950
|
+
defaultDuration: 5,
|
|
1891
1951
|
resolutions: ["720p", "1080p"],
|
|
1892
1952
|
features: ["end-frame"],
|
|
1893
1953
|
pricing: [
|
|
@@ -1905,6 +1965,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1905
1965
|
useCases: ["motion", "narrative"],
|
|
1906
1966
|
aspectRatios: VIDEO_RATIOS_HVS345,
|
|
1907
1967
|
durations: [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
1968
|
+
defaultDuration: 5,
|
|
1908
1969
|
resolutions: ["720p", "1080p"],
|
|
1909
1970
|
pricing: [
|
|
1910
1971
|
{ identifier: "wan-2.7-t2v", credits: 188, note: "5s 720p default" },
|
|
@@ -1931,6 +1992,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1931
1992
|
aspectRatios: ["16:9", "9:16"] as const,
|
|
1932
1993
|
resolutions: ["1080p", "2k", "4k"] as const,
|
|
1933
1994
|
durations: [6, 8, 10],
|
|
1995
|
+
defaultDuration: 6,
|
|
1934
1996
|
pricing: [
|
|
1935
1997
|
{ identifier: "ltx-2.3-pro", credits: 240, note: "default 1080p 6s" },
|
|
1936
1998
|
{ identifier: "ltx-2.3-pro:1080p:6s", credits: 240 },
|
|
@@ -1962,6 +2024,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1962
2024
|
// in credit-identifiers.ts is the per-band authority and snaps a 2k/4k
|
|
1963
2025
|
// request back onto 6/8/10s, so the reservation is always a real tier).
|
|
1964
2026
|
durations: [6, 8, 10, 12, 14, 16, 18, 20],
|
|
2027
|
+
defaultDuration: 6,
|
|
1965
2028
|
pricing: [
|
|
1966
2029
|
{ identifier: "ltx-2.3-fast", credits: 180, note: "default 1080p 6s" },
|
|
1967
2030
|
{ identifier: "ltx-2.3-fast:1080p:6s", credits: 180 },
|
|
@@ -1993,6 +2056,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
1993
2056
|
useCases: ["motion", "creative"],
|
|
1994
2057
|
aspectRatios: VIDEO_RATIOS_HAPPYHORSE_11,
|
|
1995
2058
|
durations: [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
2059
|
+
defaultDuration: 5,
|
|
1996
2060
|
resolutions: ["720p", "1080p"],
|
|
1997
2061
|
pricing: [
|
|
1998
2062
|
{ identifier: "happyhorse", credits: 282, note: "5s 720p default — per-second: ~5.7 cr/s @720p, ~7.3 cr/s @1080p" },
|
|
@@ -2008,6 +2072,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
2008
2072
|
description: "HappyHorse 1.1 image-to-video — 3–15s at 720p/1080p, aspect ratio inferred from input image, per-second pricing.",
|
|
2009
2073
|
useCases: ["motion", "creative"],
|
|
2010
2074
|
durations: [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
2075
|
+
defaultDuration: 5,
|
|
2011
2076
|
resolutions: ["720p", "1080p"],
|
|
2012
2077
|
pricing: [
|
|
2013
2078
|
{ identifier: "happyhorse-i2v", credits: 282, note: "5s 720p default — per-second: ~5.7 cr/s @720p, ~7.3 cr/s @1080p" },
|
|
@@ -2024,6 +2089,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
2024
2089
|
useCases: ["motion", "reference"],
|
|
2025
2090
|
aspectRatios: VIDEO_RATIOS_HAPPYHORSE_11,
|
|
2026
2091
|
durations: [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
2092
|
+
defaultDuration: 5,
|
|
2027
2093
|
resolutions: ["720p", "1080p"],
|
|
2028
2094
|
features: ["reference-image"],
|
|
2029
2095
|
pricing: [
|
|
@@ -2057,6 +2123,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
2057
2123
|
features: ["end-frame"],
|
|
2058
2124
|
aspectRatios: VIDEO_RATIOS_HVS,
|
|
2059
2125
|
durations: [5, 10],
|
|
2126
|
+
defaultDuration: 5,
|
|
2060
2127
|
resolutions: ["480p", "720p", "1080p"],
|
|
2061
2128
|
pricing: [{ identifier: "bytedance-lite", credits: 57 }],
|
|
2062
2129
|
},
|
|
@@ -2071,6 +2138,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
2071
2138
|
useCases: ["motion", "narrative"],
|
|
2072
2139
|
aspectRatios: VIDEO_RATIOS_HVS,
|
|
2073
2140
|
durations: [5, 10],
|
|
2141
|
+
defaultDuration: 5,
|
|
2074
2142
|
resolutions: ["480p", "720p", "1080p"],
|
|
2075
2143
|
pricing: [{ identifier: "bytedance-pro", credits: 175 }],
|
|
2076
2144
|
},
|
|
@@ -2084,6 +2152,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
2084
2152
|
description: "Faster Bytedance Pro variant.",
|
|
2085
2153
|
useCases: ["fast", "motion"],
|
|
2086
2154
|
durations: [5, 10],
|
|
2155
|
+
defaultDuration: 5,
|
|
2087
2156
|
resolutions: ["720p", "1080p"],
|
|
2088
2157
|
pricing: [{ identifier: "bytedance-pro-fast", credits: 90 }],
|
|
2089
2158
|
},
|
|
@@ -2105,6 +2174,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
2105
2174
|
description: "Runway Gen-3 routed through KIE. 5/10s at 720p/1080p.",
|
|
2106
2175
|
useCases: ["motion", "narrative"],
|
|
2107
2176
|
durations: [5, 10],
|
|
2177
|
+
defaultDuration: 5,
|
|
2108
2178
|
resolutions: ["720p", "1080p"],
|
|
2109
2179
|
pricing: [{ identifier: "runway-kie", credits: 30, note: "5s 720p" }],
|
|
2110
2180
|
},
|
|
@@ -2156,6 +2226,7 @@ const VIDEO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
2156
2226
|
description: "Extend ANY video: generates the continuation (audio included) and trim-stitches it into one seamless clip.",
|
|
2157
2227
|
useCases: ["extend"],
|
|
2158
2228
|
durations: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
|
|
2229
|
+
defaultDuration: 8,
|
|
2159
2230
|
resolutions: ["480p", "720p", "1080p"],
|
|
2160
2231
|
pricing: [
|
|
2161
2232
|
{ identifier: "seedance-2-extend", credits: 530, note: "default 8s 720p (ref-mode 50 + 3 stitch) — see :NsR variants" },
|
|
@@ -2506,14 +2577,26 @@ const TTS_LANGS_V3 = [
|
|
|
2506
2577
|
const TTS_LEVERS_V2 = ["stability", "similarity", "style", "speed", "speakerBoost"] as const
|
|
2507
2578
|
/**
|
|
2508
2579
|
* ElevenLabs Dialogue v4's total-character cap across lines — ONE constant, so
|
|
2509
|
-
* the live probe's answer lands in one place.
|
|
2510
|
-
*
|
|
2511
|
-
* well inside the funnel's
|
|
2512
|
-
*
|
|
2513
|
-
*
|
|
2514
|
-
*
|
|
2580
|
+
* the live probe's answer lands in one place. Measured 2026-10-06: a
|
|
2581
|
+
* 10,000-character v4 dialogue returned 200, voiced its last line and finished
|
|
2582
|
+
* whole in 130 s, well inside the funnel's 300 s timeout
|
|
2583
|
+
* (`DIALOGUE_GENERATION_TIMEOUT_MS`); decided 2026-10-07 to raise the cap to
|
|
2584
|
+
* that number, twice v3 dialogue's 5,000. Length pricing charges
|
|
2585
|
+
* `min(total, cap)`, so the same unit rows reach 10,000 characters without a
|
|
2586
|
+
* new row. ElevenLabs recommends ≤ 2,000 for quality on either model.
|
|
2587
|
+
*/
|
|
2588
|
+
const DIALOGUE_V4_MAX_CHARS = 10000
|
|
2589
|
+
/**
|
|
2590
|
+
* ElevenLabs v4 Turbo's per-request character cap — ONE constant, so a change
|
|
2591
|
+
* of the owner's rule lands in one place. The vendor declares 10,000, the same
|
|
2592
|
+
* as v4 (`maximum_text_length_per_request`, measured 2026-10-07); live, both a
|
|
2593
|
+
* 10,000- and a 10,001-character request returned 200 (the vendor accepts over
|
|
2594
|
+
* its declared figure), so the platform clamps at the declared figure — the
|
|
2595
|
+
* smaller of the two. Decided 2026-10-07: the model takes that whole cap because
|
|
2596
|
+
* it ships priced per started 100 characters. An instance that still prices
|
|
2597
|
+
* speech flat charges the flat row for any length (the Text to Speech docs say so).
|
|
2515
2598
|
*/
|
|
2516
|
-
const
|
|
2599
|
+
const V4_TURBO_MAX_CHARS = 10000
|
|
2517
2600
|
|
|
2518
2601
|
const AUDIO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
2519
2602
|
// ── ElevenLabs TTS ──
|
|
@@ -2574,6 +2657,36 @@ const AUDIO_MODELS: Record<string, ModelCatalogEntry> = {
|
|
|
2574
2657
|
languages: TTS_LANGS_V3,
|
|
2575
2658
|
},
|
|
2576
2659
|
},
|
|
2660
|
+
"elevenlabs-v4-turbo": {
|
|
2661
|
+
id: "elevenlabs-v4-turbo",
|
|
2662
|
+
kind: "audio",
|
|
2663
|
+
modes: ["tts"] as const,
|
|
2664
|
+
family: "ElevenLabs",
|
|
2665
|
+
label: "ElevenLabs v4 Turbo",
|
|
2666
|
+
series: "ElevenLabs",
|
|
2667
|
+
description: "Faster, cheaper ElevenLabs v4 at lower fidelity — [audio tags], stability and similarity control, at Turbo's credit price. Prefer v4 when quality matters. Direct API.",
|
|
2668
|
+
useCases: ["tts", "voice-over", "narration", "expressive", "fast"],
|
|
2669
|
+
features: ["audio-tags", "voice-cloning"],
|
|
2670
|
+
// Flat per request at parity with Turbo v2.5 (what a run costs where length
|
|
2671
|
+
// pricing is off); per started 100 characters at Turbo's rate where it is on
|
|
2672
|
+
// (decided 2026-10-06).
|
|
2673
|
+
pricing: [
|
|
2674
|
+
{ identifier: "elevenlabs-v4-turbo", credits: 15 },
|
|
2675
|
+
{ identifier: "elevenlabs-v4-turbo:per-100-chars", credits: 2, note: SPEECH_UNIT_PRICE_NOTE },
|
|
2676
|
+
],
|
|
2677
|
+
tts: {
|
|
2678
|
+
// Every value below was measured 2026-10-07 from live calls to eleven_v4_turbo.
|
|
2679
|
+
audioTags: true, // measured 2026-10-07: a tagged line → 200; Scribe hears no tag word spoken and lists [chuckles] as an audio event
|
|
2680
|
+
ssmlBreaks: false,
|
|
2681
|
+
levers: ["stability", "similarity"], // measured 2026-10-07: /v1/models can_use_style / can_use_speaker_boost both false, as v4; stability + similarity_boost → 200
|
|
2682
|
+
languageCode: true, // measured 2026-10-07: language_code "he" with Hebrew text → 200
|
|
2683
|
+
timestamps: true, // measured 2026-10-07: /with-timestamps answers 200 at the same character cost as the plain call
|
|
2684
|
+
stitching: true, // measured 2026-10-07: previous_text / next_text → 200, at the same character cost (context not billed)
|
|
2685
|
+
maxChars: V4_TURBO_MAX_CHARS,
|
|
2686
|
+
// The curated picker, the same 46 as v4 and v3 (the model lists 85; widening is its own change).
|
|
2687
|
+
languages: TTS_LANGS_V3, // measured 2026-10-07: all 46 present in the model's 85-language list
|
|
2688
|
+
},
|
|
2689
|
+
},
|
|
2577
2690
|
"elevenlabs-turbo": {
|
|
2578
2691
|
id: "elevenlabs-turbo",
|
|
2579
2692
|
kind: "audio",
|