@nodaro/shared 3.64.0 → 3.65.0
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-PJJAZQVZ.js → chunk-222QM7CO.js} +3 -3
- package/dist/{chunk-PJJAZQVZ.js.map → chunk-222QM7CO.js.map} +1 -1
- package/dist/chunk-27HA6YOL.cjs +377 -0
- package/dist/chunk-27HA6YOL.cjs.map +1 -0
- package/dist/{chunk-TTGZ2L3O.js → chunk-2NKF476Z.js} +3 -3
- package/dist/{chunk-TTGZ2L3O.js.map → chunk-2NKF476Z.js.map} +1 -1
- package/dist/chunk-3AU52TSW.js +143 -0
- package/dist/chunk-3AU52TSW.js.map +1 -0
- package/dist/{chunk-7AJGUI7L.cjs → chunk-5NVGKXJV.cjs} +34 -17
- package/dist/chunk-5NVGKXJV.cjs.map +1 -0
- package/dist/{chunk-3AYM47GW.js → chunk-5SSMPYE5.js} +99 -9
- package/dist/chunk-5SSMPYE5.js.map +1 -0
- package/dist/{chunk-KHH54GRP.cjs → chunk-6ET3KDO4.cjs} +15 -338
- package/dist/chunk-6ET3KDO4.cjs.map +1 -0
- package/dist/{chunk-ACZZHHFZ.cjs → chunk-6OSIOVF7.cjs} +44 -22
- package/dist/chunk-6OSIOVF7.cjs.map +1 -0
- package/dist/{chunk-OBOOHG3N.js → chunk-6X5PBAC4.js} +3 -3
- package/dist/{chunk-OBOOHG3N.js.map → chunk-6X5PBAC4.js.map} +1 -1
- package/dist/chunk-7BN4EDIC.cjs +185 -0
- package/dist/chunk-7BN4EDIC.cjs.map +1 -0
- package/dist/{chunk-IYSOVS7O.js → chunk-7OWXHW7Z.js} +31 -13
- package/dist/chunk-7OWXHW7Z.js.map +1 -0
- package/dist/{chunk-ZI3QDUHO.js → chunk-AT53ZQUC.js} +3 -3
- package/dist/{chunk-ZI3QDUHO.js.map → chunk-AT53ZQUC.js.map} +1 -1
- package/dist/{chunk-SHH2T4CR.js → chunk-BSTAAGLY.js} +3 -3
- package/dist/{chunk-SHH2T4CR.js.map → chunk-BSTAAGLY.js.map} +1 -1
- package/dist/{chunk-DZ4IOP3Q.js → chunk-CW6MMQNO.js} +277 -3
- package/dist/chunk-CW6MMQNO.js.map +1 -0
- package/dist/{chunk-Q2Y4JHF7.js → chunk-D5Y3EYRM.js} +5 -5
- package/dist/{chunk-Q2Y4JHF7.js.map → chunk-D5Y3EYRM.js.map} +1 -1
- package/dist/{chunk-EXEVT6D2.cjs → chunk-DQ2PARNH.cjs} +5 -5
- package/dist/{chunk-EXEVT6D2.cjs.map → chunk-DQ2PARNH.cjs.map} +1 -1
- package/dist/{chunk-OOZW2MM6.cjs → chunk-E66HSC2E.cjs} +9 -9
- package/dist/{chunk-OOZW2MM6.cjs.map → chunk-E66HSC2E.cjs.map} +1 -1
- package/dist/{chunk-NIAFR5EE.js → chunk-FGBGKBML.js} +32 -3
- package/dist/chunk-FGBGKBML.js.map +1 -0
- package/dist/{chunk-5OBTGNUQ.cjs → chunk-FI3HJZWM.cjs} +294 -2
- package/dist/chunk-FI3HJZWM.cjs.map +1 -0
- package/dist/{chunk-SPAHI7DY.cjs → chunk-FJZTQTF4.cjs} +4 -4
- package/dist/{chunk-SPAHI7DY.cjs.map → chunk-FJZTQTF4.cjs.map} +1 -1
- package/dist/{chunk-J6B7J73J.cjs → chunk-GJ3D5X4X.cjs} +49 -49
- package/dist/{chunk-J6B7J73J.cjs.map → chunk-GJ3D5X4X.cjs.map} +1 -1
- package/dist/{chunk-KB2SQAR3.js → chunk-H745DQ6V.js} +3 -3
- package/dist/{chunk-KB2SQAR3.js.map → chunk-H745DQ6V.js.map} +1 -1
- package/dist/{chunk-7KQ6W5DS.js → chunk-HAH2F7LT.js} +4 -4
- package/dist/{chunk-7KQ6W5DS.js.map → chunk-HAH2F7LT.js.map} +1 -1
- package/dist/{chunk-HDPJNNMF.cjs → chunk-JQMFJAYU.cjs} +21 -21
- package/dist/{chunk-HDPJNNMF.cjs.map → chunk-JQMFJAYU.cjs.map} +1 -1
- package/dist/{chunk-D4KWNDPH.cjs → chunk-JYU5EIRM.cjs} +12 -12
- package/dist/{chunk-D4KWNDPH.cjs.map → chunk-JYU5EIRM.cjs.map} +1 -1
- package/dist/{chunk-AZNZ5WNZ.cjs → chunk-K6RIN3LW.cjs} +142 -45
- package/dist/chunk-K6RIN3LW.cjs.map +1 -0
- package/dist/{chunk-YQV53HRR.cjs → chunk-KTCEKGBJ.cjs} +7 -7
- package/dist/{chunk-YQV53HRR.cjs.map → chunk-KTCEKGBJ.cjs.map} +1 -1
- package/dist/{chunk-W6VJAUGM.cjs → chunk-KWYS4IIM.cjs} +83 -37
- package/dist/chunk-KWYS4IIM.cjs.map +1 -0
- package/dist/{chunk-GTWUS6MQ.cjs → chunk-MIT2USTS.cjs} +61 -61
- package/dist/{chunk-GTWUS6MQ.cjs.map → chunk-MIT2USTS.cjs.map} +1 -1
- package/dist/{chunk-Y4GCJJPF.cjs → chunk-NYZQOIXJ.cjs} +35 -2
- package/dist/chunk-NYZQOIXJ.cjs.map +1 -0
- package/dist/{chunk-FHGRKQVZ.cjs → chunk-PGDR7CO3.cjs} +8 -8
- package/dist/{chunk-FHGRKQVZ.cjs.map → chunk-PGDR7CO3.cjs.map} +1 -1
- package/dist/{chunk-W7KHHNZR.js → chunk-PL7AQQV7.js} +53 -9
- package/dist/chunk-PL7AQQV7.js.map +1 -0
- package/dist/chunk-Q4YOLT4M.js +164 -0
- package/dist/chunk-Q4YOLT4M.js.map +1 -0
- package/dist/{chunk-CHS2SIP6.cjs → chunk-R5ILCNCC.cjs} +90 -76
- package/dist/chunk-R5ILCNCC.cjs.map +1 -0
- package/dist/chunk-RP4KQUNU.cjs +168 -0
- package/dist/chunk-RP4KQUNU.cjs.map +1 -0
- package/dist/{chunk-CHTZZRAL.cjs → chunk-SPU7MUWE.cjs} +6 -6
- package/dist/{chunk-CHTZZRAL.cjs.map → chunk-SPU7MUWE.cjs.map} +1 -1
- package/dist/{chunk-VPTA4Z7G.js → chunk-ST7OBB5J.js} +5 -5
- package/dist/{chunk-VPTA4Z7G.js.map → chunk-ST7OBB5J.js.map} +1 -1
- package/dist/{chunk-TPQJKS7D.js → chunk-V5NLQ4L5.js} +21 -7
- package/dist/{chunk-CHS2SIP6.cjs.map → chunk-V5NLQ4L5.js.map} +1 -1
- package/dist/{chunk-QYU7J5EQ.js → chunk-V6OXSXIX.js} +29 -18
- package/dist/chunk-V6OXSXIX.js.map +1 -0
- package/dist/{chunk-T2OVDF2M.cjs → chunk-WQMPIURW.cjs} +6 -6
- package/dist/{chunk-T2OVDF2M.cjs.map → chunk-WQMPIURW.cjs.map} +1 -1
- package/dist/{chunk-67CMMTLN.cjs → chunk-WZG6GZIL.cjs} +4 -4
- package/dist/{chunk-67CMMTLN.cjs.map → chunk-WZG6GZIL.cjs.map} +1 -1
- package/dist/{chunk-36VWLB4J.js → chunk-XJANKTIL.js} +3 -3
- package/dist/{chunk-36VWLB4J.js.map → chunk-XJANKTIL.js.map} +1 -1
- package/dist/{chunk-O7RZRUMZ.js → chunk-XJWED257.js} +14 -309
- package/dist/chunk-XJWED257.js.map +1 -0
- package/dist/chunk-XSSLNINL.js +346 -0
- package/dist/chunk-XSSLNINL.js.map +1 -0
- package/dist/{chunk-O4RPXBDZ.cjs → chunk-YD4XY74W.cjs} +8 -8
- package/dist/{chunk-O4RPXBDZ.cjs.map → chunk-YD4XY74W.cjs.map} +1 -1
- package/dist/{chunk-SXHHLZ22.js → chunk-YMZUE2IX.js} +9 -9
- package/dist/{chunk-SXHHLZ22.js.map → chunk-YMZUE2IX.js.map} +1 -1
- package/dist/{chunk-4IUE2JB7.js → chunk-Z77M7Q2R.js} +3 -3
- package/dist/{chunk-4IUE2JB7.js.map → chunk-Z77M7Q2R.js.map} +1 -1
- package/dist/{chunk-3KTWYYCN.js → chunk-ZSRXY5YT.js} +3 -3
- package/dist/{chunk-3KTWYYCN.js.map → chunk-ZSRXY5YT.js.map} +1 -1
- package/dist/entries/apps.cjs +42 -42
- package/dist/entries/apps.js +7 -7
- package/dist/entries/captions.d.cts +1 -1
- package/dist/entries/captions.d.ts +1 -1
- package/dist/entries/credits.cjs +230 -214
- package/dist/entries/credits.d.cts +51 -14
- package/dist/entries/credits.d.ts +51 -14
- package/dist/entries/credits.js +10 -10
- package/dist/entries/entities.cjs +202 -193
- package/dist/entries/entities.d.cts +44 -9
- package/dist/entries/entities.d.ts +44 -9
- package/dist/entries/entities.js +2 -1
- package/dist/entries/llm-models.cjs +86 -62
- package/dist/entries/llm-models.d.cts +46 -15
- package/dist/entries/llm-models.d.ts +46 -15
- package/dist/entries/llm-models.js +2 -2
- package/dist/entries/media.cjs +79 -63
- package/dist/entries/media.d.cts +154 -1
- package/dist/entries/media.d.ts +154 -1
- package/dist/entries/media.js +1 -1
- package/dist/entries/model-catalog.cjs +175 -119
- package/dist/entries/model-catalog.d.cts +4 -54
- package/dist/entries/model-catalog.d.ts +4 -54
- package/dist/entries/model-catalog.js +4 -4
- package/dist/entries/models.cjs +491 -358
- package/dist/entries/models.d.cts +86 -7
- package/dist/entries/models.d.ts +86 -7
- package/dist/entries/models.js +6 -5
- package/dist/entries/nodes.cjs +212 -212
- 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 +153 -153
- package/dist/entries/pipeline.d.cts +3 -3
- package/dist/entries/pipeline.d.ts +3 -3
- package/dist/entries/pipeline.js +3 -3
- package/dist/entries/social.cjs +177 -177
- package/dist/entries/social.js +2 -2
- package/dist/entries/ugc.cjs +71 -71
- package/dist/entries/ugc.js +4 -4
- package/dist/entries/video.cjs +174 -145
- package/dist/entries/video.d.cts +114 -2
- package/dist/entries/video.d.ts +114 -2
- package/dist/entries/video.js +7 -6
- package/dist/entries/workflow.d.cts +1 -1
- package/dist/entries/workflow.d.ts +1 -1
- package/dist/index.cjs +2074 -1793
- package/dist/index.d.cts +10 -9
- package/dist/index.d.ts +10 -9
- package/dist/index.js +24 -23
- package/dist/{model-catalog-Cf9fkEUp.d.ts → model-catalog-BoUKGFNl.d.cts} +167 -1
- package/dist/{model-catalog-Cf9fkEUp.d.cts → model-catalog-BoUKGFNl.d.ts} +167 -1
- package/dist/{model-constants-BxjiWVQ9.d.ts → model-constants-BElS3AeE.d.ts} +55 -6
- package/dist/{model-constants-C_Nyft_i.d.cts → model-constants-CpQz1bGa.d.cts} +55 -6
- package/dist/tts-capabilities-BtnDWK9K.d.ts +133 -0
- package/dist/tts-capabilities-C52UccTI.d.cts +133 -0
- package/package.json +1 -1
- package/src/__tests__/dialogue-capabilities.test.ts +167 -6
- package/src/__tests__/extend-providers.test.ts +87 -0
- package/src/__tests__/kling4-catalog.test.ts +1 -1
- package/src/__tests__/llm-models.test.ts +41 -4
- package/src/__tests__/speech-pricing.test.ts +62 -1
- package/src/__tests__/tts-capabilities.test.ts +160 -4
- package/src/__tests__/video-analysis-pricing.test.ts +13 -1
- package/src/__tests__/video-subjects.test.ts +192 -0
- package/src/__tests__/wired-character-names.test.ts +142 -0
- package/src/credit-estimators/index.ts +4 -0
- package/src/credit-estimators/speech.ts +57 -9
- package/src/credit-identifiers.ts +48 -13
- package/src/dialogue-capabilities.ts +220 -3
- package/src/dubbing-structure.ts +186 -0
- package/src/entries/entities.ts +2 -0
- package/src/entries/llm-models.ts +7 -0
- package/src/entries/media.ts +3 -1
- package/src/entries/model-catalog.ts +17 -0
- package/src/entries/models.ts +39 -0
- package/src/entries/video.ts +8 -1
- package/src/gemini-tts.ts +144 -0
- package/src/llm-models.ts +71 -21
- package/src/model-catalog.ts +259 -0
- package/src/model-constants.ts +101 -8
- package/src/tts-capabilities.ts +193 -1
- package/src/video-analysis-pricing.ts +11 -0
- package/src/video-character-refs.ts +160 -3
- package/src/wired-character-names.ts +134 -18
- package/src/wired-character-video-subjects.ts +63 -0
- package/dist/chunk-3AYM47GW.js.map +0 -1
- package/dist/chunk-5OBTGNUQ.cjs.map +0 -1
- package/dist/chunk-6YIV37V7.js +0 -31
- package/dist/chunk-6YIV37V7.js.map +0 -1
- package/dist/chunk-7AJGUI7L.cjs.map +0 -1
- package/dist/chunk-ACZZHHFZ.cjs.map +0 -1
- package/dist/chunk-AZNZ5WNZ.cjs.map +0 -1
- package/dist/chunk-DZ4IOP3Q.js.map +0 -1
- package/dist/chunk-IYSOVS7O.js.map +0 -1
- package/dist/chunk-KHH54GRP.cjs.map +0 -1
- package/dist/chunk-NIAFR5EE.js.map +0 -1
- package/dist/chunk-O7RZRUMZ.js.map +0 -1
- package/dist/chunk-QYU7J5EQ.js.map +0 -1
- package/dist/chunk-S3UGLZSF.js +0 -42
- package/dist/chunk-S3UGLZSF.js.map +0 -1
- package/dist/chunk-TPQJKS7D.js.map +0 -1
- package/dist/chunk-TQW3R44P.cjs +0 -38
- package/dist/chunk-TQW3R44P.cjs.map +0 -1
- package/dist/chunk-VZ6TE23U.cjs +0 -54
- package/dist/chunk-VZ6TE23U.cjs.map +0 -1
- package/dist/chunk-W6VJAUGM.cjs.map +0 -1
- package/dist/chunk-W7KHHNZR.js.map +0 -1
- package/dist/chunk-Y4GCJJPF.cjs.map +0 -1
|
@@ -18,6 +18,14 @@
|
|
|
18
18
|
* whole number of units times a charged unit price never under-quotes the
|
|
19
19
|
* charge (the same property the per-second video rows rely on).
|
|
20
20
|
*
|
|
21
|
+
* A model whose rate is below one credit per 100 characters (a row holds
|
|
22
|
+
* whole credits) quotes its row per 1,000 instead — `<model>:per-1k-chars`,
|
|
23
|
+
* its sheet's `ratePer1kChars` (decided 2026-11-12). The count is the same
|
|
24
|
+
* (started 100s, at least 8), and a request costs its hundreds × the row ÷ 10,
|
|
25
|
+
* rounded up once (`speechRowCredits`). Its request quote keeps the one id
|
|
26
|
+
* shape, `<model>:per-100-chars:<N>u`, N in started hundreds, so a client
|
|
27
|
+
* that builds that id quotes every model alike.
|
|
28
|
+
*
|
|
21
29
|
* Decided 2026-10-06. Every name here is a published contract.
|
|
22
30
|
*/
|
|
23
31
|
import { canonicalTtsProvider, findTtsCapabilities } from "../tts-capabilities.js"
|
|
@@ -33,13 +41,39 @@ export const SPEECH_FLOOR_UNITS = 8
|
|
|
33
41
|
/** Appended to a speech model's id to name its per-unit price row. */
|
|
34
42
|
export const SPEECH_UNIT_CREDIT_SUFFIX = ":per-100-chars"
|
|
35
43
|
|
|
44
|
+
/** Appended to a speech model's id to name its rate row quoted per 1,000 characters (`TtsCapabilities.ratePer1kChars`). */
|
|
45
|
+
export const SPEECH_PER_1K_CREDIT_SUFFIX = ":per-1k-chars"
|
|
46
|
+
|
|
36
47
|
/**
|
|
37
|
-
*
|
|
48
|
+
* How many units (started 100 characters) one rate row of the model prices:
|
|
49
|
+
* 10 for a model whose row is quoted per 1,000 characters, else 1. The
|
|
50
|
+
* legacy alias resolved; an id with no speech sheet is 1.
|
|
51
|
+
*/
|
|
52
|
+
export function speechRowUnits(modelId: string): 1 | 10 {
|
|
53
|
+
const id = canonicalTtsProvider(modelId)
|
|
54
|
+
return (findTtsCapabilities(id) ?? findDialogueCapabilities(id))?.ratePer1kChars ? 10 : 1
|
|
55
|
+
}
|
|
56
|
+
|
|
57
|
+
/**
|
|
58
|
+
* The price-row id of the rate the model a request runs as is charged at:
|
|
59
|
+
* `<model>:per-100-chars` (one unit), or `<model>:per-1k-chars` for a model
|
|
60
|
+
* quoted per 1,000 characters ({@link speechRowUnits}). The legacy
|
|
38
61
|
* `elevenlabs` alias prices on turbo's row, as it runs on turbo; any other id
|
|
39
62
|
* (including the dialogue model) is used as given.
|
|
40
63
|
*/
|
|
41
64
|
export function speechUnitCreditId(modelId: string): string {
|
|
42
|
-
|
|
65
|
+
const id = canonicalTtsProvider(modelId)
|
|
66
|
+
return `${id}${speechRowUnits(id) === 10 ? SPEECH_PER_1K_CREDIT_SUFFIX : SPEECH_UNIT_CREDIT_SUFFIX}`
|
|
67
|
+
}
|
|
68
|
+
|
|
69
|
+
/**
|
|
70
|
+
* Base credits of `units` (started hundreds) at a rate row of `rowCredits`
|
|
71
|
+
* that prices `rowUnits` units: `ceil(units × row ÷ rowUnits)`, in whole
|
|
72
|
+
* numbers so it never rounds a hair high. With `rowUnits` 1 it is exactly
|
|
73
|
+
* units × the row (every per-100 row is a whole number).
|
|
74
|
+
*/
|
|
75
|
+
export function speechRowCredits(units: number, rowCredits: number, rowUnits: number = 1): number {
|
|
76
|
+
return rowUnits === 1 ? units * rowCredits : Math.ceil((units * rowCredits) / rowUnits)
|
|
43
77
|
}
|
|
44
78
|
|
|
45
79
|
/**
|
|
@@ -57,6 +91,16 @@ export function speechCredits(chars: number, perUnitCredits: number): number {
|
|
|
57
91
|
return speechPriceUnits(chars) * perUnitCredits
|
|
58
92
|
}
|
|
59
93
|
|
|
94
|
+
/**
|
|
95
|
+
* Base credits for `chars` characters on `modelId` at its rate row
|
|
96
|
+
* (`speechUnitCreditId(modelId)`, read as `rowCredits`): per-100 or per-1K,
|
|
97
|
+
* whichever the model is quoted in. What every seam that reserves speech, and
|
|
98
|
+
* every client that quotes it, computes.
|
|
99
|
+
*/
|
|
100
|
+
export function speechCreditsFor(modelId: string, chars: number, rowCredits: number): number {
|
|
101
|
+
return speechRowCredits(speechPriceUnits(chars), rowCredits, speechRowUnits(modelId))
|
|
102
|
+
}
|
|
103
|
+
|
|
60
104
|
/**
|
|
61
105
|
* The id of the exact charge of a whole request of `units` units on `modelId`:
|
|
62
106
|
* `<model>:per-100-chars:<N>u`. Never a row of its own — a price lookup prices
|
|
@@ -65,17 +109,20 @@ export function speechCredits(chars: number, perUnitCredits: number): number {
|
|
|
65
109
|
* row it is priced from.
|
|
66
110
|
*/
|
|
67
111
|
export function speechQuoteCreditId(modelId: string, units: number): string {
|
|
68
|
-
|
|
112
|
+
// Always the per-100 shape, whatever the model's rate row: N is started hundreds.
|
|
113
|
+
return unitQuoteCreditId(`${canonicalTtsProvider(modelId)}${SPEECH_UNIT_CREDIT_SUFFIX}`, units)
|
|
69
114
|
}
|
|
70
115
|
|
|
71
116
|
/** A `<model>:per-100-chars:<N>u` id, read back. */
|
|
72
117
|
export interface SpeechQuoteCreditIdParts {
|
|
73
118
|
/** The speech model (canonical id) the request runs as. */
|
|
74
119
|
model: string
|
|
75
|
-
/** Its `<model>:per-100-chars`
|
|
120
|
+
/** Its rate row: `<model>:per-100-chars`, or `<model>:per-1k-chars` for a model quoted per 1,000. */
|
|
76
121
|
unitId: string
|
|
77
122
|
/** Units priced: at least the floor, at most the model's per-request cap in started hundreds. */
|
|
78
123
|
units: number
|
|
124
|
+
/** How many units one `unitId` row prices (10 for a per-1,000 row, else 1): the request is `speechRowCredits(units, row, rowUnits)`. */
|
|
125
|
+
rowUnits: 1 | 10
|
|
79
126
|
}
|
|
80
127
|
|
|
81
128
|
const SPEECH_QUOTE_ID = /^(.+):per-100-chars:(\d{1,6})u$/
|
|
@@ -96,16 +143,17 @@ export function parseSpeechQuoteCreditId(id: string): SpeechQuoteCreditIdParts |
|
|
|
96
143
|
if (!caps) return undefined
|
|
97
144
|
const units = Number(m[2])
|
|
98
145
|
if (!Number.isInteger(units) || units < SPEECH_FLOOR_UNITS || units > speechPriceUnits(caps.maxChars)) return undefined
|
|
99
|
-
return { model, unitId: speechUnitCreditId(model), units }
|
|
146
|
+
return { model, unitId: speechUnitCreditId(model), units, rowUnits: speechRowUnits(model) }
|
|
100
147
|
}
|
|
101
148
|
|
|
102
149
|
/**
|
|
103
150
|
* Whether an id is a length price of speech — a `<model>:per-100-chars` unit
|
|
104
|
-
* row or a `<model>:per-100-chars:<N>u`
|
|
105
|
-
*
|
|
151
|
+
* row, a `<model>:per-1k-chars` rate row or a `<model>:per-100-chars:<N>u`
|
|
152
|
+
* request quote — by its shape alone. All are served only while the server
|
|
153
|
+
* prices speech by length.
|
|
106
154
|
*/
|
|
107
155
|
export function isSpeechLengthPriceId(id: string): boolean {
|
|
108
|
-
return typeof id === "string" &&
|
|
156
|
+
return typeof id === "string" && /(?::per-100-chars(?::\d+u)?|:per-1k-chars)$/.test(id)
|
|
109
157
|
}
|
|
110
158
|
|
|
111
159
|
/**
|
|
@@ -115,5 +163,5 @@ export function isSpeechLengthPriceId(id: string): boolean {
|
|
|
115
163
|
*/
|
|
116
164
|
export function speechLengthPriceModel(id: string): string | undefined {
|
|
117
165
|
if (!isSpeechLengthPriceId(id)) return undefined
|
|
118
|
-
return id.replace(
|
|
166
|
+
return id.replace(/(?::per-100-chars(?::\d+u)?|:per-1k-chars)$/, "")
|
|
119
167
|
}
|
|
@@ -1009,19 +1009,56 @@ export function grokExtendCreditId(duration?: unknown, sourceResolution?: unknow
|
|
|
1009
1009
|
/** The video node types whose run Grok Extend can extend, set to a Grok Imagine model. */
|
|
1010
1010
|
const GROK_EXTEND_SOURCE_NODE_TYPES: ReadonlySet<string> = new Set(["generate-video", "text-to-video", "image-to-video"])
|
|
1011
1011
|
|
|
1012
|
+
/**
|
|
1013
|
+
* The resolution a Grok Extend is quoted at from the clip a Grok Imagine node
|
|
1014
|
+
* already made — the clip the run extends: the take its task (`kieTaskId`, the
|
|
1015
|
+
* task the node hands on) names. The charge reads the resolution the source's
|
|
1016
|
+
* JOB recorded (`input_data.resolution`; with none, the KIE config's pin, else
|
|
1017
|
+
* Grok's own 480p default — `sourceResolution` in grok-extend-source.ts), and
|
|
1018
|
+
* every lane that lands a take records that value on the result row beside
|
|
1019
|
+
* the task (`recordedResolution`), so the quote reads the same fact: `480p`
|
|
1020
|
+
* (any case) → `480p`; anything else, or none → `720p`, never below the charge
|
|
1021
|
+
* — a take with no recorded value may have been charged at a pinned 720p. Also
|
|
1022
|
+
* `720p` when no result names the node's task (a result landed before results
|
|
1023
|
+
* recorded it, or the take was deleted while the node kept its task).
|
|
1024
|
+
* Undefined when the node holds neither a task nor a result (it has not run).
|
|
1025
|
+
*/
|
|
1026
|
+
function grokMadeClipResolution(data: Record<string, unknown>): GrokExtendSourceResolution | undefined {
|
|
1027
|
+
const rows = (Array.isArray(data.generatedResults) ? (data.generatedResults as ReadonlyArray<unknown>) : []).filter(
|
|
1028
|
+
(r): r is Record<string, unknown> =>
|
|
1029
|
+
typeof r === "object" && r !== null && typeof (r as { url?: unknown }).url === "string" && (r as { url: string }).url.trim() !== "",
|
|
1030
|
+
)
|
|
1031
|
+
const task = typeof data.kieTaskId === "string" && data.kieTaskId.trim() !== "" ? data.kieTaskId.trim() : undefined
|
|
1032
|
+
if (task === undefined && rows.length === 0) return undefined
|
|
1033
|
+
const named = task !== undefined ? rows.find((r) => typeof r.kieTaskId === "string" && r.kieTaskId.trim() === task) : undefined
|
|
1034
|
+
const recorded = typeof named?.recordedResolution === "string" ? named.recordedResolution.trim().toLowerCase() : undefined
|
|
1035
|
+
return recorded === "480p" ? "480p" : "720p"
|
|
1036
|
+
}
|
|
1037
|
+
|
|
1012
1038
|
/**
|
|
1013
1039
|
* The source resolution a Grok Extend node is QUOTED at (its estimate, its
|
|
1014
|
-
* editor price), read off the node wired into its `video` input
|
|
1015
|
-
* when
|
|
1016
|
-
*
|
|
1017
|
-
*
|
|
1018
|
-
*
|
|
1019
|
-
*
|
|
1040
|
+
* editor price), read off the node wired into its `video` input — `480p` only
|
|
1041
|
+
* when the clip the extension continues is surely 480p; `720p`, the dearer
|
|
1042
|
+
* row, for anything else (nothing wired, another model, a clip whose
|
|
1043
|
+
* resolution is not known):
|
|
1044
|
+
*
|
|
1045
|
+
* - a Grok Imagine video node (text- or image-to-video) that already made a
|
|
1046
|
+
* clip and does not make a new one in the quoted run: THAT clip's
|
|
1047
|
+
* resolution ({@link grokMadeClipResolution}) — the run extends it, whatever
|
|
1048
|
+
* the node is set to now;
|
|
1049
|
+
* - one that has not run yet, or runs again in the quoted run (it is in
|
|
1050
|
+
* `rerunIds` and not frozen): its setting — `480p` when set so, which every
|
|
1051
|
+
* mode sends as set; unset, Grok may render it at either.
|
|
1052
|
+
*
|
|
1053
|
+
* `rerunIds`: the nodes the quoted run executes — empty for one node's own
|
|
1054
|
+
* price (its pill, its Run button), where nothing upstream runs again. The
|
|
1055
|
+
* run is held and charged at the source's own row, read off its job.
|
|
1020
1056
|
*/
|
|
1021
1057
|
export function grokExtendWiredSourceResolution(
|
|
1022
1058
|
nodeId: string | undefined,
|
|
1023
1059
|
nodes: ReadonlyArray<{ readonly id?: string; readonly type?: string; readonly data?: unknown }>,
|
|
1024
1060
|
edges: ReadonlyArray<{ readonly source?: string; readonly target?: string; readonly targetHandle?: string | null }> | undefined,
|
|
1061
|
+
rerunIds: ReadonlySet<string> = new Set(),
|
|
1025
1062
|
): GrokExtendSourceResolution {
|
|
1026
1063
|
if (nodeId === undefined || !edges) return "720p"
|
|
1027
1064
|
const sources = edges
|
|
@@ -1030,13 +1067,11 @@ export function grokExtendWiredSourceResolution(
|
|
|
1030
1067
|
if (sources.length === 0) return "720p"
|
|
1031
1068
|
const surely480p = sources.every((n) => {
|
|
1032
1069
|
const data = (n?.data ?? {}) as Record<string, unknown>
|
|
1033
|
-
return
|
|
1034
|
-
|
|
1035
|
-
|
|
1036
|
-
|
|
1037
|
-
|
|
1038
|
-
data.resolution.trim().toLowerCase() === "480p"
|
|
1039
|
-
)
|
|
1070
|
+
if (n?.type === undefined || !GROK_EXTEND_SOURCE_NODE_TYPES.has(n.type) || !isGrokExtendSourceModel(data.provider)) return false
|
|
1071
|
+
const runsAgain = n.id !== undefined && rerunIds.has(n.id) && data.skipped !== true
|
|
1072
|
+
const made = runsAgain ? undefined : grokMadeClipResolution(data)
|
|
1073
|
+
if (made !== undefined) return made === "480p"
|
|
1074
|
+
return typeof data.resolution === "string" && data.resolution.trim().toLowerCase() === "480p"
|
|
1040
1075
|
})
|
|
1041
1076
|
return surely480p ? "480p" : "720p"
|
|
1042
1077
|
}
|
|
@@ -4,13 +4,15 @@
|
|
|
4
4
|
*
|
|
5
5
|
* The dialogue models carry a `tts` sheet in `MODEL_CATALOG` like the
|
|
6
6
|
* text-to-speech models, but each lane answers only for its own models: the
|
|
7
|
-
* text-to-speech lookups (`tts-capabilities.ts`) treat a dialogue id as
|
|
8
|
-
* and these treat a text-to-speech id as unknown
|
|
7
|
+
* text-to-speech lookups (`tts-capabilities.ts`) treat a dialogue-only id as
|
|
8
|
+
* unknown, and these treat a text-to-speech-only id as unknown (a model in both
|
|
9
|
+
* lanes — Gemini — has one sheet, read by either). An unknown id here runs as
|
|
9
10
|
* {@link DEFAULT_DIALOGUE_PROVIDER} (v3 dialogue) — not as the text-to-speech
|
|
10
11
|
* fallback (turbo), which has no dialogue form.
|
|
11
12
|
*/
|
|
12
13
|
import { MODEL_CATALOG, type TtsCapabilities, type TtsSettingLever } from "./model-catalog.js"
|
|
13
|
-
import { DEFAULT_DIALOGUE_PROVIDER, type DialogueProvider } from "./model-constants.js"
|
|
14
|
+
import { DEFAULT_DIALOGUE_PROVIDER, DEFAULT_TTS_OUTPUT_FORMAT, TTS_OUTPUT_FORMATS, type DialogueProvider, type TtsOutputFormat } from "./model-constants.js"
|
|
15
|
+
import { replicaOf, ttsDirectionFor, ttsDirectionProblem, ttsReplicaProblem, ttsTakesReplica, ttsVoiceForModel, type TtsDirectionValues } from "./tts-capabilities.js"
|
|
14
16
|
|
|
15
17
|
/** The sheet of a dialogue model, or `undefined` when the id is not one. Own properties only; a non-string is unknown. */
|
|
16
18
|
export function findDialogueCapabilities(provider: unknown): TtsCapabilities | undefined {
|
|
@@ -46,3 +48,218 @@ export function dialogueStabilityAccepted(provider: unknown, value: number): boo
|
|
|
46
48
|
const steps = getDialogueCapabilities(provider).stabilitySteps
|
|
47
49
|
return steps ? steps.includes(value) : true
|
|
48
50
|
}
|
|
51
|
+
|
|
52
|
+
/** The ElevenLabs dialogue lane's own limit on distinct voices in one script (probed 2026-08-30: an 11th → 400). */
|
|
53
|
+
export const ELEVENLABS_DIALOGUE_MAX_VOICES = 10
|
|
54
|
+
|
|
55
|
+
/** The most speakers (distinct voices) one script on the dialogue model may have: its sheet's, else the ElevenLabs limit. */
|
|
56
|
+
export function dialogueMaxSpeakers(provider: unknown): number {
|
|
57
|
+
return getDialogueCapabilities(provider).dialogueSpeakers ?? ELEVENLABS_DIALOGUE_MAX_VOICES
|
|
58
|
+
}
|
|
59
|
+
|
|
60
|
+
/** The dialogue model adds filler words between two speakers on request (`TtsCapabilities.fillerWords`). */
|
|
61
|
+
export function dialogueTakesFillerWords(provider: unknown): boolean {
|
|
62
|
+
return getDialogueCapabilities(provider).fillerWords === true
|
|
63
|
+
}
|
|
64
|
+
|
|
65
|
+
/**
|
|
66
|
+
* Each line's voice on the dialogue model, when the model and the voices may
|
|
67
|
+
* have been chosen apart (a published app's Model card, a field mapping, the
|
|
68
|
+
* editor's model switch): the text-to-speech lane's `ttsVoiceForModel` rule,
|
|
69
|
+
* per speaker. A voice the model speaks stays. On a model with named voices
|
|
70
|
+
* (Gemini) every other voice moves onto one of its voices that the script does
|
|
71
|
+
* not use yet, its default first — so two speakers are never merged into one
|
|
72
|
+
* voice by the move. On an ElevenLabs model a voice only a named-voice model
|
|
73
|
+
* speaks becomes "" (the deployment's default voice). Returns the voices in
|
|
74
|
+
* line order.
|
|
75
|
+
*/
|
|
76
|
+
export function dialogueVoicesForModel(provider: unknown, voices: readonly unknown[], speakers?: unknown): string[] {
|
|
77
|
+
const sheet = getDialogueCapabilities(provider)
|
|
78
|
+
const id = dialogueProviderOf(provider)
|
|
79
|
+
const named = sheet.voices
|
|
80
|
+
if (!named) return voices.map((v) => ttsVoiceForModel(id, v) ?? "")
|
|
81
|
+
// A speaker cloned for this request (`replica`) is the model's own voice too, on a model that clones;
|
|
82
|
+
// on any other model it is a foreign voice and moves (decided 2026-11-13: one rule, `dialogueReplicaVoices`).
|
|
83
|
+
const clones = dialogueReplicaVoices(id, speakers)
|
|
84
|
+
const own = (v: unknown): v is string => typeof v === "string" && (named.some((n) => n.id === v) || clones.has(v))
|
|
85
|
+
const used = new Set(voices.filter(own))
|
|
86
|
+
const pool = [sheet.defaultVoice, ...named.map((n) => n.id)].filter((v): v is string => typeof v === "string")
|
|
87
|
+
const moved = new Map<string, string>()
|
|
88
|
+
return voices.map((v) => {
|
|
89
|
+
if (own(v)) return v
|
|
90
|
+
const key = typeof v === "string" ? v : ""
|
|
91
|
+
const kept = moved.get(key)
|
|
92
|
+
if (kept !== undefined) return kept
|
|
93
|
+
const next = pool.find((p) => !used.has(p)) ?? sheet.defaultVoice ?? named[0]!.id
|
|
94
|
+
used.add(next)
|
|
95
|
+
moved.set(key, next)
|
|
96
|
+
return next
|
|
97
|
+
})
|
|
98
|
+
}
|
|
99
|
+
|
|
100
|
+
/**
|
|
101
|
+
* The most characters of the free-text style ONE LINE takes on the dialogue
|
|
102
|
+
* model (Gemini 3.8: `direction.styleMaxChars`, a style per turn), or
|
|
103
|
+
* undefined on a model whose style is a speaker's choice (3.1 / 2.5) or that
|
|
104
|
+
* takes none (ElevenLabs).
|
|
105
|
+
*/
|
|
106
|
+
export function dialogueLineStyleMaxChars(provider: unknown): number | undefined {
|
|
107
|
+
const d = getDialogueCapabilities(provider).direction
|
|
108
|
+
return d && d.styles === undefined ? d.styleMaxChars : undefined
|
|
109
|
+
}
|
|
110
|
+
|
|
111
|
+
/** A line's style as the dialogue model is sent it: trimmed, when the model takes one per line and it fits; else undefined. */
|
|
112
|
+
export function dialogueLineStyleFor(provider: unknown, value: unknown): string | undefined {
|
|
113
|
+
const max = dialogueLineStyleMaxChars(provider)
|
|
114
|
+
if (max === undefined || typeof value !== "string") return undefined
|
|
115
|
+
const trimmed = value.trim()
|
|
116
|
+
return trimmed && trimmed.length <= max ? trimmed : undefined
|
|
117
|
+
}
|
|
118
|
+
|
|
119
|
+
/** Whether the dialogue model takes direction per SPEAKER (Gemini 3.1 / 2.5: a style, pace, accent and audio profile per voice). */
|
|
120
|
+
export function dialogueTakesSpeakerDirection(provider: unknown): boolean {
|
|
121
|
+
const d = getDialogueCapabilities(provider).direction
|
|
122
|
+
return d !== undefined && (d.styles !== undefined || d.paces !== undefined || d.accents !== undefined || d.audioProfileMaxChars !== undefined)
|
|
123
|
+
}
|
|
124
|
+
|
|
125
|
+
/** One speaker's direction as the dialogue model is sent it (`ttsDirectionFor`), or {} on a model that takes none per speaker. */
|
|
126
|
+
export function dialogueSpeakerDirectionFor(provider: unknown, values: Readonly<Record<string, unknown>>): TtsDirectionValues {
|
|
127
|
+
return dialogueTakesSpeakerDirection(provider) ? ttsDirectionFor(dialogueProviderOf(provider), values) : {}
|
|
128
|
+
}
|
|
129
|
+
|
|
130
|
+
/**
|
|
131
|
+
* The files the dialogue model delivers: every speech file format on a model
|
|
132
|
+
* that renders a WAV the platform keeps or encodes (Gemini,
|
|
133
|
+
* `dialogueAnyOutputFormat`); the default MP3 only on the ElevenLabs dialogue
|
|
134
|
+
* models, as the platform calls them.
|
|
135
|
+
*/
|
|
136
|
+
export function dialogueOutputFormats(provider: unknown): readonly TtsOutputFormat[] {
|
|
137
|
+
return getDialogueCapabilities(provider).dialogueAnyOutputFormat ? TTS_OUTPUT_FORMATS : [DEFAULT_TTS_OUTPUT_FORMAT]
|
|
138
|
+
}
|
|
139
|
+
|
|
140
|
+
/** The file a dialogue request on `provider` asks for, when it is one the model delivers; else undefined (the default MP3). */
|
|
141
|
+
export function dialogueOutputFormatFor(provider: unknown, value: unknown): TtsOutputFormat | undefined {
|
|
142
|
+
return typeof value === "string" && (dialogueOutputFormats(provider) as readonly string[]).includes(value) ? (value as TtsOutputFormat) : undefined
|
|
143
|
+
}
|
|
144
|
+
|
|
145
|
+
/** A dialogue request as every lane reads it (the route's body, a node's data, the MCP verb's args mapped). */
|
|
146
|
+
export interface DialogueRequestShape {
|
|
147
|
+
readonly dialogue?: unknown
|
|
148
|
+
readonly speakers?: unknown
|
|
149
|
+
/** The file asked for; judged against `dialogueOutputFormats` when given. */
|
|
150
|
+
readonly outputFormat?: unknown
|
|
151
|
+
}
|
|
152
|
+
|
|
153
|
+
/** Why a dialogue request cannot run on its model: where, and the sentence a caller is answered with. */
|
|
154
|
+
export interface DialogueProblem {
|
|
155
|
+
readonly path: ReadonlyArray<string | number>
|
|
156
|
+
readonly message: string
|
|
157
|
+
}
|
|
158
|
+
|
|
159
|
+
/**
|
|
160
|
+
* The speakers a dialogue clones for this request: the voice labels whose
|
|
161
|
+
* `speakers` entry carries a `replica` (two https links), on a model that
|
|
162
|
+
* clones (`TtsCapabilities.replica`); none on any other. The ONE predicate
|
|
163
|
+
* every dialogue seam reads — the voice move, the check, the worker — so a
|
|
164
|
+
* cloned speaker is kept on a model that clones and never silently voiced by
|
|
165
|
+
* a stock voice in its place.
|
|
166
|
+
*/
|
|
167
|
+
export function dialogueReplicaVoices(provider: unknown, speakers: unknown): Set<string> {
|
|
168
|
+
const id = dialogueProviderOf(provider)
|
|
169
|
+
if (!ttsTakesReplica(id) || !Array.isArray(speakers)) return new Set()
|
|
170
|
+
const out = new Set<string>()
|
|
171
|
+
for (const entry of speakers) {
|
|
172
|
+
const s = entry as { voice?: unknown; replica?: unknown } | null
|
|
173
|
+
if (s && typeof s.voice === "string" && s.voice && replicaOf(s.replica)) out.add(s.voice)
|
|
174
|
+
}
|
|
175
|
+
return out
|
|
176
|
+
}
|
|
177
|
+
|
|
178
|
+
/** A cloned speaker's recordings in a dialogue request, by voice label (`dialogueReplicaVoices`' speakers). */
|
|
179
|
+
export function dialogueReplicaOf(provider: unknown, speakers: unknown, voice: string): ReturnType<typeof replicaOf> {
|
|
180
|
+
if (!dialogueReplicaVoices(provider, speakers).has(voice) || !Array.isArray(speakers)) return undefined
|
|
181
|
+
const entry = speakers.find((s) => (s as { voice?: unknown } | null)?.voice === voice) as { replica?: unknown } | undefined
|
|
182
|
+
return replicaOf(entry?.replica)
|
|
183
|
+
}
|
|
184
|
+
|
|
185
|
+
/** The distinct voices of a script, in the order they first speak. */
|
|
186
|
+
export function dialogueVoicesOf(lines: unknown): string[] {
|
|
187
|
+
if (!Array.isArray(lines)) return []
|
|
188
|
+
const out: string[] = []
|
|
189
|
+
for (const line of lines) {
|
|
190
|
+
const voice = (line as { voice?: unknown } | null)?.voice
|
|
191
|
+
if (typeof voice === "string" && !out.includes(voice)) out.push(voice)
|
|
192
|
+
}
|
|
193
|
+
return out
|
|
194
|
+
}
|
|
195
|
+
|
|
196
|
+
/**
|
|
197
|
+
* The first reason a dialogue request cannot run on `provider`, or undefined.
|
|
198
|
+
* The ONE rule the route's schema, the workflow engine and the MCP verb read,
|
|
199
|
+
* so a script is refused the same way at every door and before any hold:
|
|
200
|
+
* at most `dialogueMaxSpeakers` distinct voices; on a model with named voices
|
|
201
|
+
* (Gemini) every voice one of them; a line's style within what the model takes
|
|
202
|
+
* per line; each speaker's direction within what the model takes per speaker.
|
|
203
|
+
* A field the model does not take is never sent, so it is no problem (as on
|
|
204
|
+
* one speaker: `ttsDirectionProblem`). The script's length is the caller's to
|
|
205
|
+
* judge (`dialogueTextBeforeCap`, @nodaro/prompts).
|
|
206
|
+
*/
|
|
207
|
+
export function dialogueProblem(provider: unknown, request: DialogueRequestShape): DialogueProblem | undefined {
|
|
208
|
+
const id = dialogueProviderOf(provider)
|
|
209
|
+
const sheet = getDialogueCapabilities(id)
|
|
210
|
+
const lines = Array.isArray(request.dialogue) ? request.dialogue : []
|
|
211
|
+
// One entry per speaker: two entries for one voice would leave which one applies to chance.
|
|
212
|
+
if (Array.isArray(request.speakers)) {
|
|
213
|
+
const seen = new Set<string>()
|
|
214
|
+
for (let i = 0; i < request.speakers.length; i++) {
|
|
215
|
+
const voice = (request.speakers[i] as { voice?: unknown } | null)?.voice
|
|
216
|
+
if (typeof voice !== "string") continue
|
|
217
|
+
if (seen.has(voice)) return { path: ["speakers", i, "voice"], message: `speakers lists the voice "${voice}" more than once; give each speaker one entry` }
|
|
218
|
+
seen.add(voice)
|
|
219
|
+
}
|
|
220
|
+
}
|
|
221
|
+
const voices = dialogueVoicesOf(lines)
|
|
222
|
+
const max = dialogueMaxSpeakers(id)
|
|
223
|
+
if (voices.length > max) {
|
|
224
|
+
return { path: ["dialogue"], message: `A dialogue can use at most ${max} unique voices — reuse voices across lines or split the script into two nodes` }
|
|
225
|
+
}
|
|
226
|
+
// A one-off clone (`replica`) on a speaker: refused on a model that does not clone, never dropped.
|
|
227
|
+
if (Array.isArray(request.speakers)) {
|
|
228
|
+
for (let i = 0; i < request.speakers.length; i++) {
|
|
229
|
+
const replica = (request.speakers[i] as { replica?: unknown } | null)?.replica
|
|
230
|
+
const problem = ttsReplicaProblem(id, replica)
|
|
231
|
+
if (problem) return { path: ["speakers", i, "replica"], message: problem }
|
|
232
|
+
}
|
|
233
|
+
}
|
|
234
|
+
const clones = dialogueReplicaVoices(id, request.speakers)
|
|
235
|
+
const named = sheet.voices
|
|
236
|
+
if (named) {
|
|
237
|
+
for (let i = 0; i < lines.length; i++) {
|
|
238
|
+
const voice = (lines[i] as { voice?: unknown } | null)?.voice
|
|
239
|
+
if (typeof voice === "string" && !clones.has(voice) && !named.some((v) => v.id === voice)) {
|
|
240
|
+
return { path: ["dialogue", i, "voice"], message: `${voice} is not a voice of ${id}. Its voices: ${named.map((v) => v.id).join(", ")}` }
|
|
241
|
+
}
|
|
242
|
+
}
|
|
243
|
+
}
|
|
244
|
+
const styleMax = dialogueLineStyleMaxChars(id)
|
|
245
|
+
if (styleMax !== undefined) {
|
|
246
|
+
for (let i = 0; i < lines.length; i++) {
|
|
247
|
+
const style = (lines[i] as { styleDirection?: unknown } | null)?.styleDirection
|
|
248
|
+
if (style === undefined) continue
|
|
249
|
+
if (typeof style !== "string") return { path: ["dialogue", i, "styleDirection"], message: "styleDirection must be a string" }
|
|
250
|
+
if (style.trim().length > styleMax) return { path: ["dialogue", i, "styleDirection"], message: `styleDirection must be at most ${styleMax} characters` }
|
|
251
|
+
}
|
|
252
|
+
}
|
|
253
|
+
if (request.outputFormat !== undefined && dialogueOutputFormatFor(id, request.outputFormat) === undefined) {
|
|
254
|
+
return { path: ["outputFormat"], message: `outputFormat on ${id} must be one of: ${dialogueOutputFormats(id).join(", ")}` }
|
|
255
|
+
}
|
|
256
|
+
if (dialogueTakesSpeakerDirection(id) && Array.isArray(request.speakers)) {
|
|
257
|
+
for (let i = 0; i < request.speakers.length; i++) {
|
|
258
|
+
const speaker = request.speakers[i]
|
|
259
|
+
if (!speaker || typeof speaker !== "object") continue
|
|
260
|
+
const problem = ttsDirectionProblem(id, speaker as Record<string, unknown>)
|
|
261
|
+
if (problem) return { path: ["speakers", i, problem.field], message: problem.message }
|
|
262
|
+
}
|
|
263
|
+
}
|
|
264
|
+
return undefined
|
|
265
|
+
}
|
|
@@ -0,0 +1,186 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Dubbing structure — the wire contract of a dub read back by its job id
|
|
3
|
+
* (`GET /v1/dubbing/:jobId`), and of the captions export beside it
|
|
4
|
+
* (`GET /v1/dubbing/:jobId/captions`).
|
|
5
|
+
*
|
|
6
|
+
* One shape whichever engine made the dub, so a client is written once. A
|
|
7
|
+
* field an engine does not tell is null, never absent:
|
|
8
|
+
* - `sourceSegmentId` names the source line a language's line translates.
|
|
9
|
+
* It is null where the dub's lines carry no link to the source lines; the
|
|
10
|
+
* two lists are then read side by side, by time.
|
|
11
|
+
* - `sourceText` is the original text a line translates, as the dub keeps
|
|
12
|
+
* it beside the line; null where the dub keeps none. Each of the two is
|
|
13
|
+
* told on its own: a line can carry `sourceText` and no `sourceSegmentId`.
|
|
14
|
+
* - a speaker's `name` and `voice` are null until they can be read.
|
|
15
|
+
*
|
|
16
|
+
* `capabilities` says what can be done to this dub now. A client offers an
|
|
17
|
+
* action only when its capability is true.
|
|
18
|
+
*/
|
|
19
|
+
|
|
20
|
+
/** The caption file formats a dub exports. */
|
|
21
|
+
export const DUBBING_CAPTION_FORMATS = ["srt", "vtt"] as const
|
|
22
|
+
export type DubbingCaptionFormat = (typeof DUBBING_CAPTION_FORMATS)[number]
|
|
23
|
+
|
|
24
|
+
/** The `language` value that names the original speech in a captions request. */
|
|
25
|
+
export const DUBBING_SOURCE_LANGUAGE = "source"
|
|
26
|
+
|
|
27
|
+
/**
|
|
28
|
+
* Where a language of a dub stands:
|
|
29
|
+
* - `processing`: being made; its lines are not readable yet.
|
|
30
|
+
* - `completed`: it is made; its lines and captions are readable where the dubbing service gives them.
|
|
31
|
+
* - `stale`: its lines changed after its media was last made.
|
|
32
|
+
* - `failed`: it was not made.
|
|
33
|
+
*/
|
|
34
|
+
export const DUBBING_LANGUAGE_STATUSES = ["processing", "completed", "stale", "failed"] as const
|
|
35
|
+
export type DubbingLanguageStatus = (typeof DUBBING_LANGUAGE_STATUSES)[number]
|
|
36
|
+
|
|
37
|
+
/** One line of the original speech. Times are seconds from the start of the media. */
|
|
38
|
+
export interface DubbingSourceSegment {
|
|
39
|
+
/** Unique within the source lines. */
|
|
40
|
+
id: string
|
|
41
|
+
speakerId: string | null
|
|
42
|
+
start: number
|
|
43
|
+
end: number
|
|
44
|
+
text: string
|
|
45
|
+
}
|
|
46
|
+
|
|
47
|
+
/** One line of a dubbed language. Times are seconds from the start of the media. */
|
|
48
|
+
export interface DubbingSegment {
|
|
49
|
+
/** Unique within this language's lines. */
|
|
50
|
+
id: string
|
|
51
|
+
speakerId: string | null
|
|
52
|
+
start: number
|
|
53
|
+
end: number
|
|
54
|
+
/**
|
|
55
|
+
* The `id` of the source line this one translates; null where the dub's
|
|
56
|
+
* lines carry no link to the source lines.
|
|
57
|
+
*/
|
|
58
|
+
sourceSegmentId: string | null
|
|
59
|
+
/**
|
|
60
|
+
* The original text this line translates, as the dub keeps it beside the
|
|
61
|
+
* line; null where it keeps none. It can be told where `sourceSegmentId` is
|
|
62
|
+
* null.
|
|
63
|
+
*/
|
|
64
|
+
sourceText: string | null
|
|
65
|
+
text: string
|
|
66
|
+
}
|
|
67
|
+
|
|
68
|
+
/** One language a dub speaks. */
|
|
69
|
+
export interface DubbingLanguage {
|
|
70
|
+
/** The language code the dub was asked for (`es`, `pt-BR`); unique within a dub. */
|
|
71
|
+
code: string
|
|
72
|
+
/**
|
|
73
|
+
* The job whose result is this language's dubbed media. null when the
|
|
74
|
+
* language was made but no job of the account holds its media: adding the
|
|
75
|
+
* language again delivers it, unless its status is `stale` (its lines
|
|
76
|
+
* changed after its media was made), when adding it is refused.
|
|
77
|
+
*/
|
|
78
|
+
jobId: string | null
|
|
79
|
+
status: DubbingLanguageStatus
|
|
80
|
+
/** Empty unless `status` is `completed` or `stale`, and where the dubbing service does not give the language's lines. */
|
|
81
|
+
segments: DubbingSegment[]
|
|
82
|
+
}
|
|
83
|
+
|
|
84
|
+
/** The voice a speaker is dubbed in. */
|
|
85
|
+
export interface DubbingSpeakerVoice {
|
|
86
|
+
id: string
|
|
87
|
+
name: string | null
|
|
88
|
+
previewUrl: string | null
|
|
89
|
+
}
|
|
90
|
+
|
|
91
|
+
/** One speaker the dub tells apart. */
|
|
92
|
+
export interface DubbingSpeaker {
|
|
93
|
+
/** The `speakerId` its lines carry. */
|
|
94
|
+
id: string
|
|
95
|
+
name: string | null
|
|
96
|
+
voice: DubbingSpeakerVoice | null
|
|
97
|
+
/** How many source lines the speaker has (lines of the first language when the source is not readable). */
|
|
98
|
+
segmentCount: number
|
|
99
|
+
}
|
|
100
|
+
|
|
101
|
+
/** What can be done to a dub now. */
|
|
102
|
+
export interface DubbingCapabilities {
|
|
103
|
+
/** Change a line's text. */
|
|
104
|
+
editSegments: boolean
|
|
105
|
+
/** Make chosen lines again. */
|
|
106
|
+
regenerate: boolean
|
|
107
|
+
/** Replace a speaker's voice. */
|
|
108
|
+
setSpeakerVoice: boolean
|
|
109
|
+
/** Render the media again after edits. */
|
|
110
|
+
render: boolean
|
|
111
|
+
/** Add a target language (`POST /v1/dubbing/:jobId/languages`). */
|
|
112
|
+
addLanguage: boolean
|
|
113
|
+
/** Export captions (`GET /v1/dubbing/:jobId/captions`). */
|
|
114
|
+
captions: boolean
|
|
115
|
+
}
|
|
116
|
+
|
|
117
|
+
/**
|
|
118
|
+
* What adding a language to this dub is charged. The price is the answer of
|
|
119
|
+
* `POST /v1/credits/model-costs` for `creditIdByLanguage[language] ?? creditId`.
|
|
120
|
+
*/
|
|
121
|
+
export interface DubbingAddLanguagePrice {
|
|
122
|
+
/** Whole minutes an added language is charged for; null when the length is only known once it runs. */
|
|
123
|
+
billableMinutes: number | null
|
|
124
|
+
/** The price id of an added language; null with `billableMinutes`. */
|
|
125
|
+
creditId: string | null
|
|
126
|
+
/** The languages priced by another id than `creditId`, by lowercase language code. */
|
|
127
|
+
creditIdByLanguage: Record<string, string>
|
|
128
|
+
}
|
|
129
|
+
|
|
130
|
+
/** `GET /v1/dubbing/:jobId`. */
|
|
131
|
+
export interface DubbingStructure {
|
|
132
|
+
/** The job the dub was created by. Languages added later hang off it. */
|
|
133
|
+
jobId: string
|
|
134
|
+
/** Seconds of media the dub covers; null when not known. */
|
|
135
|
+
durationSec: number | null
|
|
136
|
+
/** The original speech; null when its lines are not readable. */
|
|
137
|
+
source: { language: string | null; segments: DubbingSourceSegment[] } | null
|
|
138
|
+
languages: DubbingLanguage[]
|
|
139
|
+
speakers: DubbingSpeaker[]
|
|
140
|
+
capabilities: DubbingCapabilities
|
|
141
|
+
addLanguage: DubbingAddLanguagePrice
|
|
142
|
+
}
|
|
143
|
+
|
|
144
|
+
/** `POST /v1/dubbing/:jobId/languages` request. */
|
|
145
|
+
export interface DubbingAddLanguageInput {
|
|
146
|
+
/** The language to add (`es`, `pt-BR`). */
|
|
147
|
+
targetLanguage: string
|
|
148
|
+
/** The rights confirmation, when the dub was made from a link the deployment asks one for. */
|
|
149
|
+
confirmationId?: string
|
|
150
|
+
}
|
|
151
|
+
|
|
152
|
+
/** `POST /v1/dubbing/:jobId/languages` response: the job that delivers the added language. */
|
|
153
|
+
export interface DubbingAddLanguageResult {
|
|
154
|
+
jobId: string
|
|
155
|
+
}
|
|
156
|
+
|
|
157
|
+
/**
|
|
158
|
+
* The error codes the dubbing reads and the add-language call answer beside
|
|
159
|
+
* the platform's own (`not_found`, `unauthorized`, `validation_error`). Each
|
|
160
|
+
* is the `error.code` of the response body.
|
|
161
|
+
*/
|
|
162
|
+
export const DUBBING_ERROR_CODES = [
|
|
163
|
+
/** 422: the job is not a dub. */
|
|
164
|
+
"not_a_dubbing_job",
|
|
165
|
+
/** 409: the dub is still being made. */
|
|
166
|
+
"dubbing_not_ready",
|
|
167
|
+
/** 422: the dub was not made. */
|
|
168
|
+
"dubbing_failed",
|
|
169
|
+
/** 410: the dub is no longer held where it was made. */
|
|
170
|
+
"dubbing_unavailable",
|
|
171
|
+
/** 404: the dub has no such language. */
|
|
172
|
+
"dubbing_language_not_found",
|
|
173
|
+
/** 409: that language is still being made. */
|
|
174
|
+
"dubbing_language_processing",
|
|
175
|
+
/** 422: that language has no captions to export. */
|
|
176
|
+
"dubbing_captions_unavailable",
|
|
177
|
+
/** 409: the dub already speaks that language. */
|
|
178
|
+
"dubbing_language_exists",
|
|
179
|
+
/** 422: the length of the dub cannot be read, so a language cannot be priced. */
|
|
180
|
+
"unreadable_dubbing_source",
|
|
181
|
+
/** 413: the dub is longer than 30 minutes, so a language cannot be added to it. */
|
|
182
|
+
"media_duration_exceeds_limit",
|
|
183
|
+
/** 502: the dubbing service did not answer. */
|
|
184
|
+
"dubbing_read_failed",
|
|
185
|
+
] as const
|
|
186
|
+
export type DubbingErrorCode = (typeof DUBBING_ERROR_CODES)[number]
|
package/src/entries/entities.ts
CHANGED
|
@@ -50,7 +50,9 @@ export {
|
|
|
50
50
|
wiredCharacterNames,
|
|
51
51
|
wiredCharacterNamesOf,
|
|
52
52
|
wiredCharacterBySlug,
|
|
53
|
+
wiredCharacterSlug,
|
|
53
54
|
wiredCharacterKey,
|
|
55
|
+
wiredCharacterVideoUrls,
|
|
54
56
|
aliasRetiredCharacterMentions,
|
|
55
57
|
RETIRED_CHARACTER_MENTION_SLUGS,
|
|
56
58
|
} from "../wired-character-names.js"
|
|
@@ -45,6 +45,13 @@ export {
|
|
|
45
45
|
getLlmModel,
|
|
46
46
|
getLlmTier,
|
|
47
47
|
getLlmModalityCaps,
|
|
48
|
+
type LlmModalityCaps,
|
|
49
|
+
llmAcceptsFiles,
|
|
50
|
+
LLM_FILE_MODELS,
|
|
51
|
+
LLM_REFERENCE_FILES_MAX,
|
|
52
|
+
LLM_REFERENCE_FILES_MAX_TOTAL_BYTES,
|
|
53
|
+
LLM_REFERENCE_FILES_MAX_PAGES,
|
|
54
|
+
LLM_REFERENCE_FILE_MIME_TYPES,
|
|
48
55
|
buildLlmCreditIdentifier,
|
|
49
56
|
llmCreditRung,
|
|
50
57
|
resolveLlmCreditId,
|