@nodaro/prompts 1.6.0 → 1.7.2

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 CHANGED
@@ -10017,6 +10017,9 @@ function renderLens(l) {
10017
10017
  if (l.aperture) bits.push(`f/${l.aperture}`);
10018
10018
  return bits.length > 0 ? `Lens: ${bits.join(", ")}.` : "";
10019
10019
  }
10020
+ function identityRefsSentence(firstOrdinal, lastOrdinal) {
10021
+ return firstOrdinal === lastOrdinal ? `${REF_BINDING.ordinal(firstOrdinal)} is an identity reference for this shot's subjects \u2014 match its subject's exact appearance; it is not a frame.` : `${REF_BINDING.ordinal(firstOrdinal)} through ${REF_BINDING.ordinal(lastOrdinal)} are identity references for this shot's subjects \u2014 match each subject's exact appearance; they are not frames.`;
10022
+ }
10020
10023
  var REF_BINDING = {
10021
10024
  image: (label, n) => `the ${label} from @image_${n}`,
10022
10025
  video: (label, n) => `the ${label} from @video_${n}`,
@@ -10595,11 +10598,12 @@ function promptBindsFirstFrame(prompt) {
10595
10598
  return /@image_\d+\s+as\s+the\s+(first|opening)\s*(\(first\))?\s*frame/i.test(prompt);
10596
10599
  }
10597
10600
  function resolveSeedance2Inputs(args) {
10601
+ const limits = args.limits ?? shared.SEEDANCE_2_REF_LIMITS;
10598
10602
  const firstFrameUrl = clean(args.firstFrameUrl);
10599
10603
  const lastFrameUrl = clean(args.lastFrameUrl);
10600
10604
  const refImages = cleanList(args.refImageUrls);
10601
- const refVideos = cleanList(args.refVideoUrls).slice(0, shared.SEEDANCE_2_REF_LIMITS.videos);
10602
- const refAudios = cleanList(args.refAudioUrls).slice(0, shared.SEEDANCE_2_REF_LIMITS.audio);
10605
+ const refVideos = cleanList(args.refVideoUrls).slice(0, limits.videos);
10606
+ const refAudios = cleanList(args.refAudioUrls).slice(0, limits.audio);
10603
10607
  const hasAnyReference = refImages.length > 0 || refVideos.length > 0 || refAudios.length > 0;
10604
10608
  const canUseStrictMode = !hasAnyReference && (Boolean(firstFrameUrl) || !lastFrameUrl);
10605
10609
  if (canUseStrictMode) {
@@ -10609,7 +10613,7 @@ function resolveSeedance2Inputs(args) {
10609
10613
  return { mode: "first-frame", firstFrameUrl, lastFrameUrl: void 0, referenceImageUrls: [], referenceVideoUrls: [], referenceAudioUrls: [], promptSuffix: "", droppedRefImages: 0 };
10610
10614
  }
10611
10615
  const frameCount = (firstFrameUrl ? 1 : 0) + (lastFrameUrl ? 1 : 0);
10612
- const userImageSlots = Math.max(0, shared.SEEDANCE_2_REF_LIMITS.images - frameCount);
10616
+ const userImageSlots = Math.max(0, limits.images - frameCount);
10613
10617
  const keptUserImages = refImages.slice(0, userImageSlots);
10614
10618
  const droppedRefImages = refImages.length - keptUserImages.length;
10615
10619
  const referenceImageUrls = [...keptUserImages];
@@ -10643,6 +10647,43 @@ function resolveSeedance2Inputs(args) {
10643
10647
  droppedRefImages
10644
10648
  };
10645
10649
  }
10650
+ var GEMINI_OMNI_INPUT_SLOTS = shared.VIDEO_REF_LIMITS_BY_PROVIDER["gemini-omni-video"]?.images ?? 7;
10651
+ function resolveGeminiOmniI2vInputs(args) {
10652
+ const refs = (args.refImageUrls ?? []).filter((u) => typeof u === "string" && u.length > 0);
10653
+ const slots = GEMINI_OMNI_INPUT_SLOTS - (args.videoConnected ? 2 : 0);
10654
+ const refSlots = Math.max(0, slots - 1);
10655
+ const kept = refs.slice(0, refSlots);
10656
+ const droppedRefImages = refs.length - kept.length;
10657
+ const imageUrls = [args.firstFrameUrl, ...kept];
10658
+ if (kept.length === 0) return { imageUrls, promptSuffix: "", droppedRefImages };
10659
+ const frameSentence = promptBindsFirstFrame(args.prompt) ? "" : REF_BINDING.frame(1, "opening");
10660
+ const promptSuffix = [frameSentence, identityRefsSentence(2, kept.length + 1)].filter(Boolean).join(" ");
10661
+ return { imageUrls, promptSuffix, droppedRefImages };
10662
+ }
10663
+
10664
+ // src/veo-i2v-inputs.ts
10665
+ var VEO_INGREDIENT_SLOTS = 3;
10666
+ function resolveVeoI2vInputs(args) {
10667
+ const refs = (args.refImageUrls ?? []).filter((u) => typeof u === "string" && u.length > 0);
10668
+ if (refs.length === 0) {
10669
+ return {
10670
+ imageUrls: args.endFrameUrl ? [args.firstFrameUrl, args.endFrameUrl] : [args.firstFrameUrl],
10671
+ promptSuffix: "",
10672
+ droppedRefImages: 0,
10673
+ droppedEndFrame: false
10674
+ };
10675
+ }
10676
+ const kept = refs.slice(0, VEO_INGREDIENT_SLOTS - 1);
10677
+ const droppedRefImages = refs.length - kept.length;
10678
+ const frameSentence = promptBindsFirstFrame(args.prompt) ? "" : REF_BINDING.frame(1, "opening");
10679
+ return {
10680
+ imageUrls: [args.firstFrameUrl, ...kept],
10681
+ generationType: "REFERENCE_2_VIDEO",
10682
+ promptSuffix: [frameSentence, identityRefsSentence(2, kept.length + 1)].filter(Boolean).join(" "),
10683
+ droppedRefImages,
10684
+ droppedEndFrame: Boolean(args.endFrameUrl)
10685
+ };
10686
+ }
10646
10687
  function toOptions(arr, categoryField) {
10647
10688
  return arr.map((e) => {
10648
10689
  const opt = {
@@ -25694,9 +25735,10 @@ function date4(params) {
25694
25735
 
25695
25736
  // ../../node_modules/zod/v4/classic/external.js
25696
25737
  config(en_default());
25697
-
25698
- // src/picker-analyzer-registry.ts
25699
25738
  var PERSON_EXCLUDED = /* @__PURE__ */ new Set(["age-custom"]);
25739
+ function tagDim(arr, dimension) {
25740
+ return arr.map((e) => ({ id: e.id, label: e.label, description: e.description, dimension }));
25741
+ }
25700
25742
  var personCleanup = (patch, mode) => {
25701
25743
  if (mode === "override") {
25702
25744
  patch.customAge = void 0;
@@ -25751,9 +25793,148 @@ var PICKER_ANALYZER_REGISTRY = {
25751
25793
  field: "cameraFormat",
25752
25794
  label: "Camera / Film Stock",
25753
25795
  entries: CAMERA_FORMATS
25796
+ },
25797
+ // ─── Text-to-picker expansion (Cine AI Fill): every remaining catalog ─────
25798
+ // Flat single-value pickers — field names match the picker wiring's
25799
+ // valueField (and the node-data shape the published-app input card writes).
25800
+ setting: { kind: "flat", toolName: "emit_setting", field: "setting", label: "Setting", entries: SETTINGS },
25801
+ atmosphere: { kind: "flat", toolName: "emit_atmosphere", field: "atmosphere", label: "Atmosphere", entries: ATMOSPHERES },
25802
+ style: { kind: "flat", toolName: "emit_style", field: "style", label: "Style", entries: STYLES },
25803
+ mood: { kind: "flat", toolName: "emit_mood", field: "mood", label: "Mood", entries: MOODS },
25804
+ "color-look": { kind: "flat", toolName: "emit_color_look", field: "colorLook", label: "Color / Look", entries: COLOR_LOOKS },
25805
+ photographer: { kind: "flat", toolName: "emit_photographer", field: "photographer", label: "Photographer / Artist", entries: PHOTOGRAPHERS },
25806
+ aesthetic: { kind: "flat", toolName: "emit_aesthetic", field: "aesthetic", label: "Aesthetic / Microtrend", entries: AESTHETICS },
25807
+ era: { kind: "flat", toolName: "emit_era", field: "era", label: "Era / Period", entries: ERAS },
25808
+ "photo-genre": { kind: "flat", toolName: "emit_photo_genre", field: "photoGenre", label: "Photo Genre", entries: PHOTO_GENRES },
25809
+ backdrop: { kind: "flat", toolName: "emit_backdrop", field: "backdrop", label: "Backdrop", entries: BACKDROPS },
25810
+ "render-quality": { kind: "flat", toolName: "emit_render_quality", field: "renderQuality", label: "Render Quality", entries: RENDER_QUALITIES },
25811
+ "composition-effects": { kind: "flat", toolName: "emit_composition_effects", field: "compositionEffect", label: "Composition Effect", entries: COMPOSITION_EFFECTS },
25812
+ "post-process-effects": { kind: "flat", toolName: "emit_post_process_effects", field: "postProcess", label: "Post-Process Effect", entries: POST_PROCESS_EFFECTS },
25813
+ "action-fx": { kind: "flat", toolName: "emit_action_fx", field: "actionFx", label: "Action FX", entries: ACTION_FX },
25814
+ "loop-subject": { kind: "flat", toolName: "emit_loop_subject", field: "loopSubject", label: "Loop Subject", entries: LOOP_SUBJECTS },
25815
+ transition: { kind: "flat", toolName: "emit_transition", field: "transition", label: "Transition", entries: TRANSITIONS },
25816
+ "character-fx": { kind: "flat", toolName: "emit_character_fx", field: "characterFx", label: "Character FX", entries: CHARACTER_FX },
25817
+ pose: { kind: "flat", toolName: "emit_pose", field: "pose", label: "Pose", entries: POSES },
25818
+ material: { kind: "flat", toolName: "emit_material", field: "material", label: "Material", entries: MATERIALS },
25819
+ "held-prop": { kind: "flat", toolName: "emit_held_prop", field: "heldProp", label: "Held Prop", entries: HELD_PROPS },
25820
+ "camera-motion": { kind: "flat", toolName: "emit_camera_motion", field: "cameraMotion", label: "Camera Motion", entries: CAMERA_MOTIONS },
25821
+ animal: { kind: "flat", toolName: "emit_animal", field: "animal", label: "Animal", entries: shared.ANIMALS },
25822
+ vehicle: { kind: "flat", toolName: "emit_vehicle", field: "vehicle", label: "Vehicle", entries: shared.VEHICLES },
25823
+ weapon: { kind: "flat", toolName: "emit_weapon", field: "weapon", label: "Weapon", entries: shared.WEAPONS },
25824
+ furniture: { kind: "flat", toolName: "emit_furniture", field: "furniture", label: "Furniture", entries: shared.FURNITURE },
25825
+ // Discriminated multi-dim pickers. lighting/temporal/exposure discriminate
25826
+ // on the single catalog's `category`; the sound/voice pickers span several
25827
+ // per-field catalogs, so their entries are synthesized with an explicit
25828
+ // `dimension` tag (tagDim below) — same wire shape either way.
25829
+ lighting: {
25830
+ kind: "discriminated",
25831
+ toolName: "emit_lighting",
25832
+ discriminator: "category",
25833
+ order: LIGHTING_CATEGORY_ORDER,
25834
+ fieldByKey: LIGHTING_FIELD_BY_CATEGORY,
25835
+ labels: LIGHTING_CATEGORY_LABELS,
25836
+ entries: LIGHTINGS,
25837
+ limitFn: () => 1
25838
+ },
25839
+ temporal: {
25840
+ kind: "discriminated",
25841
+ toolName: "emit_temporal",
25842
+ discriminator: "category",
25843
+ order: ["speed", "freeze", "direction", "shutter"],
25844
+ fieldByKey: { speed: "temporalSpeed", freeze: "temporalFreeze", direction: "temporalDirection", shutter: "temporalShutter" },
25845
+ labels: { speed: "Playback Speed", freeze: "Freeze", direction: "Direction", shutter: "Shutter" },
25846
+ entries: TEMPORALS,
25847
+ limitFn: () => 1
25848
+ },
25849
+ "exposure-settings": {
25850
+ kind: "discriminated",
25851
+ toolName: "emit_exposure_settings",
25852
+ discriminator: "category",
25853
+ order: ["aperture", "shutter-speed", "iso"],
25854
+ fieldByKey: { aperture: "aperture", "shutter-speed": "shutterSpeed", iso: "isoValue" },
25855
+ labels: { aperture: "Aperture", "shutter-speed": "Shutter Speed", iso: "ISO" },
25856
+ entries: EXPOSURE_SETTINGS,
25857
+ limitFn: () => 1
25858
+ },
25859
+ "music-genre": {
25860
+ kind: "discriminated",
25861
+ toolName: "emit_music_genre",
25862
+ discriminator: "dimension",
25863
+ order: ["genre", "subgenre", "era"],
25864
+ fieldByKey: { genre: "genre", subgenre: "subgenre", era: "era" },
25865
+ labels: { genre: "Genre", subgenre: "Subgenre", era: "Era" },
25866
+ entries: [
25867
+ ...tagDim(MUSIC_GENRES, "genre"),
25868
+ // Subgenres carry promptHint but no description — the hint doubles as
25869
+ // the legend text (it describes the sound well enough for matching).
25870
+ ...MUSIC_GENRES.flatMap(
25871
+ (g) => g.subgenres.map((s) => ({ id: s.id, label: s.label, description: s.promptHint, dimension: "subgenre" }))
25872
+ ),
25873
+ ...tagDim(MUSIC_ERAS, "era")
25874
+ ],
25875
+ limitFn: (k) => k === "genre" ? 2 : 1
25876
+ },
25877
+ "music-mood": {
25878
+ kind: "discriminated",
25879
+ toolName: "emit_music_mood",
25880
+ discriminator: "dimension",
25881
+ order: ["energy", "emotion", "vibe"],
25882
+ fieldByKey: { energy: "energy", emotion: "emotion", vibe: "vibe" },
25883
+ labels: { energy: "Energy", emotion: "Emotion", vibe: "Vibe" },
25884
+ entries: [...tagDim(MUSIC_ENERGIES, "energy"), ...tagDim(MUSIC_EMOTIONS, "emotion"), ...tagDim(MUSIC_VIBES, "vibe")],
25885
+ limitFn: (k) => k === "energy" ? 1 : 2
25886
+ },
25887
+ instrumentation: {
25888
+ kind: "discriminated",
25889
+ toolName: "emit_instrumentation",
25890
+ discriminator: "dimension",
25891
+ order: ["instruments", "production", "vocalPresence", "singingStyle"],
25892
+ fieldByKey: { instruments: "instruments", production: "production", vocalPresence: "vocalPresence", singingStyle: "singingStyle" },
25893
+ labels: { instruments: "Instruments", production: "Production Style", vocalPresence: "Vocal Presence", singingStyle: "Singing Style" },
25894
+ entries: [
25895
+ ...tagDim(INSTRUMENTS, "instruments"),
25896
+ ...tagDim(PRODUCTION_STYLES, "production"),
25897
+ ...tagDim(VOCAL_PRESENCE, "vocalPresence"),
25898
+ ...tagDim(SINGING_STYLES, "singingStyle")
25899
+ ],
25900
+ limitFn: (k) => k === "instruments" ? 3 : k === "production" ? 1 : 2
25901
+ },
25902
+ "voice-character": {
25903
+ kind: "discriminated",
25904
+ toolName: "emit_voice_character",
25905
+ discriminator: "dimension",
25906
+ order: ["age", "gender", "language", "accent", "timbre"],
25907
+ fieldByKey: { age: "age", gender: "gender", language: "language", accent: "accent", timbre: "timbre" },
25908
+ labels: { age: "Age", gender: "Gender", language: "Language", accent: "Accent", timbre: "Timbre" },
25909
+ entries: [
25910
+ ...tagDim(VOICE_AGES, "age"),
25911
+ ...tagDim(VOICE_GENDERS, "gender"),
25912
+ ...tagDim(VOICE_LANGUAGES, "language"),
25913
+ ...tagDim(VOICE_ACCENTS, "accent"),
25914
+ ...tagDim(VOICE_TIMBRES, "timbre")
25915
+ ],
25916
+ limitFn: (k) => k === "language" ? 2 : 1
25917
+ },
25918
+ "voice-delivery": {
25919
+ kind: "discriminated",
25920
+ toolName: "emit_voice_delivery",
25921
+ discriminator: "dimension",
25922
+ order: ["pace", "emotion", "archetype"],
25923
+ fieldByKey: { pace: "pace", emotion: "emotion", archetype: "archetype" },
25924
+ labels: { pace: "Pace", emotion: "Emotion", archetype: "Archetype" },
25925
+ entries: [...tagDim(VOICE_PACES, "pace"), ...tagDim(VOICE_EMOTIONS, "emotion"), ...tagDim(VOICE_ARCHETYPES, "archetype")],
25926
+ limitFn: () => 1
25754
25927
  }
25755
25928
  };
25756
25929
  var PICKER_TYPES = Object.keys(PICKER_ANALYZER_REGISTRY);
25930
+ var PICKER_ANALYZER_FAMILIES = {
25931
+ scene: ["setting", "atmosphere", "backdrop", "era", "temporal"],
25932
+ look: ["style", "color-look", "mood", "aesthetic", "photographer", "photo-genre", "render-quality", "composition-effects", "post-process-effects"],
25933
+ camera: ["framing", "camera-motion", "lens", "camera-format", "lighting", "exposure-settings"],
25934
+ character: ["person", "styling", "pose", "character-fx"],
25935
+ elements: ["animal", "vehicle", "weapon", "furniture", "held-prop", "material", "action-fx", "loop-subject", "transition"],
25936
+ audio: ["music-genre", "music-mood", "instrumentation", "voice-character", "voice-delivery"]
25937
+ };
25757
25938
  var ANALYZABLE_PICKER_TYPES = new Set(PICKER_TYPES);
25758
25939
  function isAnalyzablePicker(t) {
25759
25940
  return ANALYZABLE_PICKER_TYPES.has(t);
@@ -26068,7 +26249,8 @@ var SEEDANCE_2_DOCTRINE = {
26068
26249
  "Native multi-track audio \u2014 cue it inline: \uFF08background music\uFF09, <sound effects>, and quoted dialogue.",
26069
26250
  "References go by ordinal (@Image 1, Video 2) in attachment order; earlier = higher priority. Identity = ONE headshot + ONE full-body (multi-view sheets cause ID drift). 4-5 assets total beats maxing the 9/3/3 caps.",
26070
26251
  "No negative-prompt parameter \u2014 put constraints in the prompt: 'keep it subtitle-free, do not generate a watermark, do not generate a logo'.",
26071
- "seedance-2-5 only: one shot runs to 30s (the 2.0 SKUs stop at 15s), so storyboard a whole beat instead of planning a stitch. Ref caps are wider (30/10/10), but 4-5 assets still gives the best identity fidelity."
26252
+ "seedance-2-5 only: one shot runs to 30s (the 2.0 SKUs stop at 15s), so storyboard a whole beat instead of planning a stitch. Ref caps are wider (30/10/10), but 4-5 assets still gives the best identity fidelity.",
26253
+ "Auto-path formula: Subject \u2192 Action \u2192 Environment \u2192 Camera \u2192 Style \u2192 Constraints in 60-100 words; ONE camera instruction (chain with 'then'); separate camera motion from subject motion; always add one lighting phrase."
26072
26254
  ],
26073
26255
  doctrine: `Prompt structure (front-load what matters most):
26074
26256
  precise subject \u2192 action details \u2192 scene/environment \u2192 lighting & color tone \u2192 camera movement \u2192 visual style \u2192 image quality \u2192 constraints.
@@ -26083,7 +26265,7 @@ precise subject \u2192 action details \u2192 scene/environment \u2192 lighting &
26083
26265
  **Generation differences (seedance-2-5 vs the 2.0 SKUs)**
26084
26266
  - A single 2.5 shot runs to 30s, where every 2.0 SKU stops at 15s. Plan a complete 4-6 shot beat inside ONE generation instead of splitting it into two clips and stitching \u2014 no seam to hide, and continuity holds because it never leaves the model.
26085
26267
  - 2.5 also takes far more reference material (30 images / 10 videos / 10 audio vs 9/3/3). Treat that as room for COVERAGE \u2014 more distinct characters, locations and props in one shot \u2014 not as licence to pile refs onto one identity. The "ONE headshot + ONE full-body, 4-5 assets total" rule above still produces the best likeness on 2.5.
26086
- - 2.5 renders at 480p/720p only: there is no 1080p or 4K tier, so route a job that needs one to seedance-2 (which has both) or upscale afterwards.
26268
+ - 2.5 renders at 480p/720p/1080p (1080p since 2026-08-17): there is no 4K tier, so route a job that needs 4K to seedance-2 (which has it) or upscale afterwards.
26087
26269
  - With a start frame, 2.5 always derives the output aspect from that frame \u2014 an explicit aspect ratio is rejected outright, so compose the frame at the ratio you want.
26088
26270
 
26089
26271
  **References (when reference media is attached)**
@@ -26106,11 +26288,34 @@ precise subject \u2192 action details \u2192 scene/environment \u2192 lighting &
26106
26288
  **Known weaknesses \u2192 workarounds**
26107
26289
  - Text rendering is weak: keep on-screen text to short common words; for exact text or logos, attach the artwork as a reference image and instruct "the logo from Image N stays in the corner unchanged".
26108
26290
  - More than 4 referenced people gets unstable: group people into composite images of \u22644 first (image generation), then reference those composites.
26109
- - Repeated extension degrades quality: prefer high-definition reference assets and avoid stacking many continuations.`
26291
+ - Repeated extension degrades quality: prefer high-definition reference assets and avoid stacking many continuations.
26292
+
26293
+ **Auto-path formula (community-sourced enrichment \u2014 apiyi.com Seedance 2.0 prompt guide,
26294
+ higgsfield.ai 4K breakdown; captured 2026-08-09)**
26295
+ - Six steps IN ORDER, 60-100 words total (longer measurably degrades): Subject \u2192 Action \u2192 Environment \u2192 Camera \u2192 Style \u2192 Constraints.
26296
+ - ONE primary camera instruction per shot. Compound moves chain with "then": "camera slow tracking then subtle rise" \u2014 never two competing verbs. The 8 reliable camera types: push-in, pull-out, pan, tracking, orbit/arc, aerial, handheld, locked-off.
26297
+ - SEPARATE camera movement from subject movement \u2014 the single biggest quality lever: "The dancer spins slowly. Camera holds fixed framing." \u2014 never "spinning camera around a dancing person".
26298
+ - Pace with human words (slow / gentle / gradual / smooth / controlled) \u2014 never fps numbers or f-stops in the basic path.
26299
+ - ALWAYS add one lighting phrase (highest-impact single addition): golden hour / rim light / neon glow / backlit / overcast.
26300
+ - Bake stability constraints in: "avoid jitter and bent limbs", "avoid temporal flicker", "avoid identity drift".
26301
+ - Ban vague adjectives standing alone ("epic", "amazing", "beautiful", bare "cinematic") \u2014 every adjective needs a concrete noun.
26302
+ - Mode notes: i2v \u2014 skip subject description (the frame has it), focus on motion, append "preserve composition and colors". v2v \u2014 describe the style TRANSFORM, keep motion + identity.
26303
+ - Advanced (pro path): focal angles in degrees ("47\xB0 normal", "29\xB0 telephoto", "107\xB0 wide"); "180\xB0 shutter" for filmic motion blur; handheld texture as "organic shake, micro-drift, subtle dutch"; "white balance locked 5200K"; explicit POSITIVE LOCKS section + "100% matches the reference" for identity-critical shots.
26304
+
26305
+ **Camera-path control \u2014 the magenta-line method (STORYBOARD community technique; the
26306
+ manual pro path for precise trajectories, NOT the auto path)**
26307
+ 1. Duplicate the start frame; on the COPY draw a thick magenta line + arrowhead \u2014 the line is the camera's flight path, the arrow its end point. Keep the clean original.
26308
+ 2. Attach BOTH frames and declare the guide: "Image N contains a magenta line and arrow \u2014 a hidden camera trajectory guide, NOT part of the scene. Completely remove it: no line, no arrow, no paint, no trail, no reflection." Skipping the removal order RENDERS the line.
26309
+ 3. Command the path: "one continuous FPV drone glide following the S-shaped curve as closely as possible \u2014 do not shortcut. Camera motion is the priority." Lock the clean frame as first frame + scene reference; lock the destination frame if wired.
26310
+ 4. Pace with timing blocks ("[00:00-00:02] rise over the rooftop \u2026 [00:07-00:09] settle on the doorway") and keep any dialogue SHORT \u2014 long lines fight the move.
26311
+ 5. Assign image-input roles explicitly: first-frame/scene-ref \xB7 destination frame \xB7 path-guide \xB7 3-6 character-identity refs \u2014 and bind identities with @-mentions exactly like the platform's reference pills.`
26110
26312
  };
26111
26313
  var KLING_AUDIO_DOCTRINE = {
26112
- providers: ["kling", "kling-3.0", "kling-3-omni"],
26113
- heading: "Kling 2.6 / 3.0 / 3 Omni (kling, kling-3.0, kling-3-omni)",
26314
+ // kling-turbo (2.5 Turbo Pro) + kling-master (2.1 Master) are SILENT tiers of
26315
+ // the same engine: the structure/motion guidance applies, the Audio block
26316
+ // does not (variant note in the doctrine body).
26317
+ providers: ["kling", "kling-3.0", "kling-3-omni", "kling-turbo", "kling-master"],
26318
+ heading: "Kling 2.1 / 2.5 / 2.6 / 3.0 / 3 Omni (kling, kling-3.0, kling-3-omni, kling-turbo, kling-master)",
26114
26319
  tips: [
26115
26320
  'Kling speaks scripted dialogue natively with lip sync \u2014 quote the line and enable sound: [Anna: warm calm voice]: "We made it." On kling/kling-3.0 audio raises the credit cost; kling-3-omni includes it.',
26116
26321
  "Structure prompts as Scene \u2192 character/element \u2192 Motion \u2192 Audio \u2192 style. Put ALL sound in one 'Audio:' block: dialogue in quotes, then SFX and ambience described plainly ('rain tapping on glass, no music').",
@@ -26142,7 +26347,11 @@ var KLING_AUDIO_DOCTRINE = {
26142
26347
 
26143
26348
  **Limits**
26144
26349
  - Kling 2.6 prompts cap at 1000 characters \u2014 front-load scene + dialogue and trim style tails first. kling-3.0 accepts long prompts.
26145
- - Durations: 2.6 = 5/10s; 3.0/omni = 3-15s. A spoken line needs roughly 1s per 2-3 words \u2014 don't script more dialogue than the clip can hold.`
26350
+ - Durations: 2.6 = 5/10s; 3.0/omni = 3-15s. A spoken line needs roughly 1s per 2-3 words \u2014 don't script more dialogue than the clip can hold.
26351
+
26352
+ **Variant note \u2014 kling-turbo (2.5 Turbo Pro) & kling-master (2.1 Master)**
26353
+ - SILENT tiers: no audio parameter, so the entire Audio block above does not apply \u2014 skip dialogue/SFX cues; the Scene \u2192 Character \u2192 Motion \u2192 Style structure and motion guidance carry over unchanged.
26354
+ - Durations 5/10s; kling-turbo takes an end frame (tail_image_url); kling-master is single-image i2v.`
26146
26355
  };
26147
26356
  var MINIMAX_H3_DOCTRINE = {
26148
26357
  providers: ["minimax-h3"],
@@ -26179,10 +26388,199 @@ precise subject \u2192 action details \u2192 scene/environment \u2192 lighting &
26179
26388
  **Constraints**
26180
26389
  - There is NO negative-prompt parameter \u2014 all constraints belong in the prompt text itself: "keep it subtitle-free, do not generate a watermark, do not generate a logo, stable picture".`
26181
26390
  };
26391
+ var VEO_31_DOCTRINE = {
26392
+ providers: ["veo3", "veo3.1", "veo3_lite", "veo-1080p", "veo-4k", "veo-extend"],
26393
+ heading: "VEO 3.1 \u2014 Quality / Fast / Lite (veo3, veo3.1, veo3_lite)",
26394
+ tips: [
26395
+ "Structure prompts as [Cinematography] + [Subject] + [Action] + [Context] + [Style & Ambiance] \u2014 lead with the camera, not the subject (Google's official formula).",
26396
+ 'Dialogue: quote the exact line with attribution \u2014 A woman says, "We have to leave now." (no subtitles). Cue sound as separate lines: SFX: thunder cracks; Ambient noise: quiet hum of a starship bridge.',
26397
+ "Multi-shot pacing via timestamp blocks: [00:00-00:02] medium shot\u2026 [00:02-00:04] reverse shot\u2026 \u2014 VEO honors per-window actions inside one 8s generation.",
26398
+ "Negative prompting is positive phrasing: not 'no buildings' but 'a desolate landscape with no buildings or roads'. Keep prompts under ~175 words \u2014 longer overloads the generation.",
26399
+ "Start+end frame: pass both and describe the transition move ('smooth 180-degree arc ending on the POV behind her'). References (ingredients) keep characters/objects consistent and DO generate audio."
26400
+ ],
26401
+ doctrine: `Prompt structure (Google's official Veo 3.1 formula \u2014 lead with the camera):
26402
+ [Cinematography] + [Subject] + [Action] + [Context] + [Style & Ambiance].
26403
+ Example: "Medium shot, a tired corporate worker, rubbing his temples in exhaustion, in front of a bulky 1980s computer in a cluttered office late at night, lit by harsh fluorescents and the green monitor glow. Retro aesthetic, 1980s color film, slightly grainy."
26404
+
26405
+ **Camera vocabulary (use the exact terms)**
26406
+ - Movement: dolly shot, tracking shot, crane shot, aerial view, slow pan, POV shot, 180-degree arc shot.
26407
+ - Composition: wide shot, medium shot, close-up, extreme close-up, two-shot, low angle, high angle.
26408
+ - Lens/focus: shallow depth of field, deep focus, wide-angle lens, macro lens, soft focus.
26409
+
26410
+ **Audio (native, multi-track \u2014 dialogue / SFX / ambience)**
26411
+ - Dialogue: quote the exact line with attribution: The detective says in a weary voice, "Of all the offices in this town, you had to walk into mine." Append "(no subtitles)" \u2014 VEO otherwise tends to burn captions in.
26412
+ - Sound effects on their own line: "SFX: a crystal wine glass shatters on the marble floor". Ambient bed: "Ambient noise: rain against the window, distant traffic".
26413
+ - Sound can drive the visual ("the sound reverberating through the empty ballroom") \u2014 VEO syncs audio-visual timing.
26414
+
26415
+ **Multi-shot timestamp prompting (inside one generation)**
26416
+ - Split the clip into [mm:ss-mm:ss] windows, one action per window:
26417
+ [00:00-00:02] Medium shot from behind a young explorer walking toward a clearing.
26418
+ [00:02-00:04] Reverse shot of her freckled face, eyes widening.
26419
+ [00:04-00:08] Wide, high-angle crane shot revealing the ruins below.
26420
+ - 4 / 6 / 8 second clips; budget ~2s per window.
26421
+
26422
+ **Frames & references**
26423
+ - Start + end frame: wire both (imageUrls [start, end]) and describe the camera path between them \u2014 "a smooth 180-degree arc shot, starting front-facing and circling to end on the POV from behind her".
26424
+ - Reference images (ingredients): attach character/object/scene refs and name them in the prompt ("using the provided images for the detective and the office, \u2026"). Reference runs DO generate audio.
26425
+
26426
+ **Constraints**
26427
+ - Negative prompting works by positive description: write "a desolate landscape with no buildings or roads", not "no buildings".
26428
+ - Keep prompts \u2264 ~175 words \u2014 beyond that instructions conflict and adherence drops. Resolution 720p/1080p; aspect 16:9 / 9:16.
26429
+
26430
+ Sources: Google Cloud "Ultimate prompting guide for Veo 3.1"
26431
+ (cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1),
26432
+ KIE VEO API docs (docs.kie.ai/veo3-api/generate-veo-3-video). Captured 2026-08-09.`
26433
+ };
26434
+ var GEMINI_OMNI_DOCTRINE = {
26435
+ providers: ["gemini-omni-video"],
26436
+ heading: "Gemini Omni Video (gemini-omni-video)",
26437
+ tips: [
26438
+ "Multimodal Google video with native audio: text-to-video, image-to-video, and video-edit through the same prompt surface. 4/6/8/10s; 720p/1080p or 4K tier.",
26439
+ "Structure like the platform default: subject \u2192 action \u2192 scene \u2192 lighting \u2192 camera \u2192 style. Quote dialogue lines to have them spoken; describe SFX/ambience plainly in the prompt.",
26440
+ "Text-to-video REQUIRES a concrete aspect ratio (the API hard-rejects a missing one); image runs infer aspect from the input.",
26441
+ "Reference images ride along as additional imageUrls \u2014 bind them in the prompt ('the woman from the first image'). Video-edit: wire a source clip and describe the change, not the whole scene."
26442
+ ],
26443
+ doctrine: `Prompt structure (no public Google prompt guide exists for the Omni video endpoint \u2014
26444
+ the API contract is the doctrine source, like MiniMax H3; structure guidance mirrors the
26445
+ platform's ordinal-reference conventions):
26446
+ subject \u2192 action \u2192 scene/environment \u2192 lighting \u2192 camera movement \u2192 style \u2192 constraints.
26447
+
26448
+ **Modes (picked from the wired inputs)**
26449
+ - Nothing visual \u2192 text-to-video. A concrete aspect ratio is REQUIRED \u2014 the API hard-rejects a missing one (Nodaro sends the node's ratio; there is no adaptive).
26450
+ - Image(s) wired \u2192 image-to-video: the first image anchors the scene; extra images are references \u2014 bind each in the prompt ("the woman from the first image", "the interior from the second image").
26451
+ - Source video wired \u2192 video-edit (served through the same handle): describe the CHANGE ("replace the daylight with dusk, keep the motion and framing"), not a full re-description.
26452
+
26453
+ **Audio (native)**
26454
+ - Audio is generated with the clip. Quote dialogue to have it spoken; describe SFX and ambience plainly ("rain on glass, low synth bed"). State exclusions ("no music") or a bed may be invented.
26455
+
26456
+ **Duration & tiers**
26457
+ - 4 / 6 / 8 / 10 seconds. 720p/1080p tier or the pricier 4K tier \u2014 pick 4K only when the deliverable needs it (nearly 2\xD7 the credits).
26458
+
26459
+ Source: KIE gemini-omni-video market contract (parameters + live behavior probed for the
26460
+ aspect-ratio hard-reject, see providers/kie/video.ts). Captured 2026-08-09.`
26461
+ };
26462
+ var GROK_IMAGINE_DOCTRINE = {
26463
+ providers: ["grok-i2v", "grok-imagine-video-1.5"],
26464
+ heading: "Grok Imagine (grok-i2v, grok-imagine-video-1.5)",
26465
+ tips: [
26466
+ "Keep prompts simple and direct \u2014 Subject + Action + Setting + Camera + Mood. Grok expands the prompt itself; over-specification fights the expander.",
26467
+ "Image-to-video: the input image IS the first frame (composition, identity, and style are preserved) \u2014 describe the MOTION, don't re-describe the still.",
26468
+ "Video 1.5: 1-15s (default 8), 480p/720p/1080p, up to 7 input images (1080p allows only one). Native audio incl. music, SFX, and lip-synced dialogue \u2014 quote the line to have it spoken.",
26469
+ "Aspect ratio applies to text runs (1:1/16:9/9:16/3:2/2:3/auto); a single input image locks the output to the image's own aspect."
26470
+ ],
26471
+ doctrine: `Prompt structure (xAI's guidance is minimal by design \u2014 the model auto-expands prompts):
26472
+ Subject + Action + Setting + Camera + Lighting/Mood, written simply and directly. Reduce
26473
+ descriptions of static/unchanged parts \u2014 spend the words on what MOVES.
26474
+
26475
+ **Image-to-video (the primary mode)**
26476
+ - The input image is the FIRST FRAME, not a loose reference: composition, subject identity, and visual style carry over. Describe motion and camera only ("she turns toward the window as the camera slowly pushes in"); re-describing the still wastes adherence.
26477
+ - grok-imagine-video-1.5 accepts up to 7 images (identity/scene references beyond the first frame); at 1080p only ONE image is allowed.
26478
+
26479
+ **Audio (video-1.5)**
26480
+ - Native audio generates with the clip \u2014 background music, SFX, and lip-synced dialogue. Quote the spoken line; describe the music/SFX plainly. There is no audio toggle on the KIE contract \u2014 cue (or exclude) sound in the prompt text.
26481
+
26482
+ **Durations / tiers**
26483
+ - grok-i2v: 6 or 10 seconds. grok-imagine-video-1.5: 1-15 seconds in 1s steps (default 8), 480p (default) / 720p / 1080p. Prompt cap 4096 chars \u2014 but shorter is better here.
26484
+
26485
+ Sources: KIE Grok Imagine contracts (docs.kie.ai/market/grok-imagine/image-to-video,
26486
+ docs.kie.ai/market/grok-imagine/1-5-preview), xAI Grok Imagine 1.5 release notes
26487
+ (x.ai/news/grok-imagine-1-5). Captured 2026-08-09.`
26488
+ };
26489
+ var WAN_DOCTRINE = {
26490
+ providers: ["wan", "wan-i2v", "wan-turbo", "wan-flash", "wan-2.7", "wan-2.7-i2v", "wan-2.7-t2v", "wan-2.7-pro", "wan-videoedit"],
26491
+ heading: "Wan 2.x (wan, wan-i2v, wan-turbo, wan-2.7 family)",
26492
+ tips: [
26493
+ "Alibaba's official formula: Entity + Scene + Motion (basic) \u2192 add Aesthetic control + Stylization (advanced). Image-to-video: Motion + Camera only \u2014 the image already defines entity and scene.",
26494
+ "Sound (2.5+): append a sound description block \u2014 voice / sound effects / background music. Avoid scripting EXACT lip-synced lines (official anti-pattern); describe the voice and intent instead.",
26495
+ "Multi-shot (2.6/2.7): Overall description + shot number + timestamp + per-shot content. For ONE continuous take write 'Generate single shot' (the shot_type parameter is gone in 2.7).",
26496
+ "References go by 'Image 1' / 'Video 1' (capitalized, with a space). Anti-patterns: naming real people, demanding exact legible text, rapid scene changes in one clip, very long choreography.",
26497
+ "Style words are strong levers: cyberpunk, claymation, pixel style, felt style, tilt-shift, time-lapse. wan-videoedit: describe the transform, keep motion + identity."
26498
+ ],
26499
+ doctrine: `Prompt structure (Alibaba Model Studio's official formulas):
26500
+ - Basic: Entity + Scene + Motion.
26501
+ - Advanced: Entity (description) + Scene (description) + Motion (description) + Aesthetic control + Stylization.
26502
+ - Image-to-video: Motion + Camera movement ONLY \u2014 the wired image already defines entity and scene; re-describing it fights the frame.
26503
+ - Sound (2.5/2.6/2.7): \u2026 + Sound description (voice / sound effects / background music).
26504
+ - Multi-shot (2.6/2.7): Overall description + Shot number + Timestamp + Shot content.
26505
+ - Reference-to-video (2.6/2.7): Reference identifier + Action + Scene + optional Lines + optional BGM.
26506
+
26507
+ **Camera vocabulary**
26508
+ push-in (intimacy/tension), pull-out (scale/isolation), tracking shot, orbit, fixed camera, and compound movements chained sequentially for epic scale.
26509
+
26510
+ **Single-shot control (2.7)**
26511
+ - The shot_type parameter no longer exists \u2014 write "Generate single shot" in the prompt to force one continuous take; otherwise 2.7's planner may cut.
26512
+
26513
+ **References**
26514
+ - English format is "Image 1" / "Video 1" (capitalized, space-separated) \u2014 bind every wired asset by that name or it may be ignored.
26515
+
26516
+ **Official anti-patterns (from Alibaba's guide)**
26517
+ - Do NOT name specific real people.
26518
+ - Do NOT script exact lip-synced dialogue \u2014 describe the voice and intent ("she murmurs a reassurance, warm and low") instead of demanding word-perfect lips.
26519
+ - Avoid rapid scene changes inside a single clip, very long choreographed sequences, and demands for exactly legible on-screen text.
26520
+
26521
+ **Stylization**
26522
+ - Style words are strong levers: cyberpunk, line-art illustration, felt style, 3D cartoon, pixel style, puppet animation, claymation, black-and-white animation, tilt-shift, time-lapse.
26523
+
26524
+ Source: Alibaba Cloud Model Studio \u2014 "Text-to-video / image-to-video prompt guide"
26525
+ (alibabacloud.com/help/en/model-studio/text-to-video-prompt). Captured 2026-08-09.`
26526
+ };
26527
+ var HAPPYHORSE_DOCTRINE = {
26528
+ providers: ["happyhorse", "happyhorse-i2v", "happyhorse-ref2v", "happyhorse-edit"],
26529
+ heading: "HappyHorse 1.1 (happyhorse, happyhorse-i2v, happyhorse-ref2v)",
26530
+ tips: [
26531
+ "Any-language prompts up to 5000 chars (2500 Chinese) \u2014 excess is silently truncated, so front-load subject \u2192 action \u2192 scene \u2192 camera \u2192 style.",
26532
+ "3-15 seconds per second of billing; 720p or 1080p; ratios 16:9 / 9:16 / 1:1 / 4:3 / 3:4. Pick the shortest duration that serves the shot.",
26533
+ "ref2v is one of the few true REFERENCE modes on the roster: wired refs keep identity across the clip \u2014 bind each reference explicitly in the prompt.",
26534
+ "No published vendor style guide \u2014 the platform's standard structure applies; keep one camera move per shot and quantify motion physically."
26535
+ ],
26536
+ doctrine: `Prompt structure (no public HappyHorse prompt guide exists \u2014 the KIE API contract is the
26537
+ doctrine source; platform-standard structure applies):
26538
+ subject \u2192 action \u2192 scene/environment \u2192 lighting \u2192 camera movement \u2192 style \u2192 constraints.
26539
+
26540
+ **Contract facts (KIE, per-mode pages)**
26541
+ - Prompts: any language, up to 5000 non-Chinese / 2500 Chinese characters \u2014 excess is TRUNCATED silently, so put the load-bearing content first.
26542
+ - Duration 3-15s (default 5), billed per second. Resolution 720p / 1080p (default). Aspect 16:9 (default) / 9:16 / 1:1 / 4:3 / 3:4.
26543
+ - Modes: text-to-video (happyhorse), image-to-video (happyhorse-i2v), reference-to-video (happyhorse-ref2v) \u2014 ref2v preserves wired identities; name each reference in the prompt so the binding is explicit.
26544
+
26545
+ **Style guidance (platform-standard, honestly generic)**
26546
+ - One camera movement per shot; physical, quantified action ("slowly raises a hand") over abstract emotion words; state exclusions ("no on-screen text, no watermark") in the prompt.
26547
+
26548
+ Source: KIE HappyHorse 1.1 contracts (docs.kie.ai/market/happyhorse/text-to-video,
26549
+ \u2026/happyhorse-1-1/image-to-video, \u2026/happyhorse-1-1/reference-to-video). Captured 2026-08-09.`
26550
+ };
26551
+ var RUNWAY_KIE_DOCTRINE = {
26552
+ providers: ["runway-kie", "runway-extend", "runway-aleph"],
26553
+ heading: "Runway via KIE (runway-kie)",
26554
+ tips: [
26555
+ "Prompt cap is 1800 chars; KIE's own guidance: be specific about subject, action, style, and setting. No native audio \u2014 plan sound as a separate pass.",
26556
+ "Durations 5 or 10s with a hard trade-off: 10s cannot be 1080p, 1080p cannot exceed 5s \u2014 pick per deliverable.",
26557
+ "Text runs REQUIRE an aspect ratio (16:9/4:3/1:1/3:4/9:16); image runs IGNORE it \u2014 the input image dictates output dimensions.",
26558
+ "Image-to-video treats the image as the anchor frame: describe motion and camera, not the still."
26559
+ ],
26560
+ doctrine: `Prompt structure (KIE contract guidance): "be specific about subject, action, style, and
26561
+ setting" \u2014 subject \u2192 action \u2192 scene \u2192 camera \u2192 style, within the 1800-character cap.
26562
+
26563
+ **Contract facts (KIE Runway endpoint)**
26564
+ - Duration 5 or 10 seconds; quality 720p or 1080p \u2014 10s@1080p does NOT exist (10s forces 720p; 1080p forces 5s). Choose by deliverable: crisp hero shot \u2192 5s/1080p; longer beat \u2192 10s/720p.
26565
+ - Text-to-video REQUIRES aspectRatio (16:9 / 4:3 / 1:1 / 3:4 / 9:16). Image-to-video IGNORES aspectRatio \u2014 the input image dictates output dimensions.
26566
+ - No audio is generated \u2014 score/SFX are a separate pass (merge-video-audio / video-sfx downstream).
26567
+
26568
+ **Style guidance**
26569
+ - The image input anchors composition and identity \u2014 describe the motion ("she pushes the door open as the camera tracks left"), not the still.
26570
+ - Keep one continuous camera idea per clip; front-load the subject and action.
26571
+
26572
+ Source: KIE Runway contract (docs.kie.ai/runway-api/generate-ai-video). Captured 2026-08-09.`
26573
+ };
26182
26574
  var PROVIDER_PROMPT_DOCTRINES = [
26183
26575
  SEEDANCE_2_DOCTRINE,
26184
26576
  KLING_AUDIO_DOCTRINE,
26185
- MINIMAX_H3_DOCTRINE
26577
+ MINIMAX_H3_DOCTRINE,
26578
+ VEO_31_DOCTRINE,
26579
+ GEMINI_OMNI_DOCTRINE,
26580
+ GROK_IMAGINE_DOCTRINE,
26581
+ WAN_DOCTRINE,
26582
+ HAPPYHORSE_DOCTRINE,
26583
+ RUNWAY_KIE_DOCTRINE
26186
26584
  ];
26187
26585
  var DOCTRINE_BY_PROVIDER = new Map(
26188
26586
  PROVIDER_PROMPT_DOCTRINES.flatMap((d) => d.providers.map((p) => [p, d]))
@@ -26286,6 +26684,7 @@ var PROVIDER_CAPABILITIES = {
26286
26684
  "gpt-image": "Creative concepts, illustration, variable quality tiers",
26287
26685
  "gpt-image-2": "Latest GPT Image \u2014 sharp text, photorealism, 1K/2K/4K resolution",
26288
26686
  "grok": "General purpose, good text understanding",
26687
+ "grok-2": "Grok Imagine 2 \u2014 expressive, high-contrast, stylized output",
26289
26688
  "imagen4": "Google's latest, strong photorealism and text rendering",
26290
26689
  "imagen4-fast": "Faster Imagen 4 variant",
26291
26690
  "imagen4-ultra": "Highest quality Imagen 4",
@@ -26362,7 +26761,7 @@ var PROVIDER_CAPABILITIES = {
26362
26761
  "seedance-2": "Seedance 2.0 \u2014 multimodal refs (9 images / 3 videos / 3 audio), native multi-track audio, multi-shot storytelling, 4-15s",
26363
26762
  "seedance-2-fast": "Seedance 2.0 Fast \u2014 same multimodal + audio capabilities, cheaper and quicker",
26364
26763
  "seedance-2-mini": "Seedance 2.0 Mini \u2014 same multimodal + audio capabilities, budget tier, 480p/720p, 4-15s",
26365
- "seedance-2-5": "Seedance 2.5 \u2014 up to 30s in ONE shot (no stitching), wider multimodal refs (30 images / 10 videos / 10 audio), native audio, 480p/720p",
26764
+ "seedance-2-5": "Seedance 2.5 \u2014 up to 30s in ONE shot (no stitching), wider multimodal refs (30 images / 10 videos / 10 audio), native audio, 480p/720p/1080p",
26366
26765
  "minimax-h3": "MiniMax Hailuo 3 \u2014 premium multimodal refs (9 images / 3 videos / 3 audio), always-on audio, 2K or 768P, 4-15s per-second pricing",
26367
26766
  "wan": "Versatile, good for animations and transformations",
26368
26767
  "wan-turbo": "Faster Wan generation",
@@ -26390,7 +26789,7 @@ var PROVIDER_CAPABILITIES = {
26390
26789
  "seedance-2": "Seedance 2.0 \u2014 start/end frame + multimodal refs, native audio, 4-15s",
26391
26790
  "seedance-2-fast": "Seedance 2.0 Fast \u2014 same capabilities, cheaper and quicker",
26392
26791
  "seedance-2-mini": "Seedance 2.0 Mini \u2014 same capabilities, budget tier, 480p/720p",
26393
- "seedance-2-5": "Seedance 2.5 \u2014 start/end frame + wide multimodal refs, native audio, up to 30s, 480p/720p",
26792
+ "seedance-2-5": "Seedance 2.5 \u2014 start/end frame + wide multimodal refs, native audio, up to 30s, 480p/720p/1080p",
26394
26793
  "minimax-h3": "MiniMax Hailuo 3 \u2014 first/last frame + multimodal refs, always-on audio, 2K or 768P, 4-15s",
26395
26794
  "hailuo-2.3-pro": "Premium Hailuo animation",
26396
26795
  "hailuo-2.3": "Standard Hailuo animation",
@@ -26503,7 +26902,7 @@ var NODE_PROMPT_CANDIDATE_FIELDS = {
26503
26902
  function computeNodePrompt(nodeType, data, { override, wired, refMap, appendWired }) {
26504
26903
  let typed;
26505
26904
  if (nodeType === "text-to-speech") {
26506
- typed = data.textSource === "direct" ? [data.directText] : [];
26905
+ typed = data.textSource === "direct" || !present(wired) ? [data.directText] : [];
26507
26906
  } else {
26508
26907
  const fields = NODE_PROMPT_CANDIDATE_FIELDS[nodeType] ?? ["prompt"];
26509
26908
  typed = fields.map((f) => data[f]);
@@ -31511,6 +31910,212 @@ function getFactorySnippets(target, media) {
31511
31910
  (s) => s.target === target && s.media.includes(media)
31512
31911
  );
31513
31912
  }
31913
+ function mapCat(arr, groupKey) {
31914
+ return arr.map((e) => ({
31915
+ id: e.id,
31916
+ label: e.label,
31917
+ description: e.description,
31918
+ group: groupKey ? e[groupKey] : void 0
31919
+ }));
31920
+ }
31921
+ function flatCat(arr) {
31922
+ return arr.map((e) => ({ id: e.id, label: e.label }));
31923
+ }
31924
+ function optionsByDiscriminator(arr, keyOf, fieldByKey) {
31925
+ const out = {};
31926
+ for (const e of arr) {
31927
+ const field = fieldByKey[keyOf(e)];
31928
+ if (!field) continue;
31929
+ (out[field] ??= []).push({ id: e.id, label: e.label });
31930
+ }
31931
+ return out;
31932
+ }
31933
+ var SINGLE_PICKER_WIRING = [
31934
+ // -------- "Look" family --------
31935
+ { kind: "single", nodeType: "setting", label: "Setting", valueField: "setting", defaultValue: "forest", catalogId: "setting", entries: mapCat(SETTINGS, "category"), groupOrder: ["indoor", "urban", "nature", "fantastical"], groupLabels: SETTING_CATEGORY_LABELS },
31936
+ { kind: "single", nodeType: "atmosphere", label: "Atmosphere", valueField: "atmosphere", defaultValue: "clear", catalogId: "atmosphere", entries: mapCat(ATMOSPHERES) },
31937
+ { kind: "single", nodeType: "style", label: "Style", valueField: "style", defaultValue: "cinematic", catalogId: "style", entries: mapCat(STYLES) },
31938
+ { kind: "single", nodeType: "color-look", label: "Color / Look", valueField: "colorLook", defaultValue: "warm", catalogId: "color-look", entries: mapCat(COLOR_LOOKS, "category"), groupOrder: COLOR_LOOK_CATEGORY_ORDER, groupLabels: COLOR_LOOK_CATEGORY_LABELS },
31939
+ { kind: "single", nodeType: "mood", label: "Mood", valueField: "mood", defaultValue: "calm", catalogId: "mood", entries: mapCat(MOODS, "category"), groupOrder: MOOD_CATEGORY_ORDER, groupLabels: MOOD_CATEGORY_LABELS },
31940
+ { kind: "single", nodeType: "photographer", label: "Photographer / Artist", valueField: "photographer", defaultValue: "tim-walker", catalogId: "photographer", entries: mapCat(PHOTOGRAPHERS, "category"), groupOrder: PHOTOGRAPHER_CATEGORY_ORDER, groupLabels: PHOTOGRAPHER_CATEGORY_LABELS },
31941
+ { kind: "single", nodeType: "aesthetic", label: "Aesthetic / Microtrend", valueField: "aesthetic", defaultValue: "y2k", catalogId: "aesthetic", entries: mapCat(AESTHETICS, "category"), groupOrder: AESTHETIC_CATEGORY_ORDER, groupLabels: AESTHETIC_CATEGORY_LABELS },
31942
+ { kind: "single", nodeType: "era", label: "Era / Period", valueField: "era", defaultValue: "1990s-mall", catalogId: "era", entries: mapCat(ERAS, "category"), groupOrder: ERA_CATEGORY_ORDER, groupLabels: ERA_CATEGORY_LABELS },
31943
+ { kind: "single", nodeType: "photo-genre", label: "Photo Genre", valueField: "photoGenre", defaultValue: "fashion-editorial", catalogId: "photo-genre", entries: mapCat(PHOTO_GENRES, "category"), groupOrder: PHOTO_GENRE_CATEGORY_ORDER, groupLabels: PHOTO_GENRE_CATEGORY_LABELS },
31944
+ { kind: "single", nodeType: "backdrop", label: "Backdrop", valueField: "backdrop", defaultValue: "white-seamless", catalogId: "backdrop", entries: mapCat(BACKDROPS, "category"), groupOrder: BACKDROP_CATEGORY_ORDER, groupLabels: BACKDROP_CATEGORY_LABELS },
31945
+ { kind: "single", nodeType: "render-quality", label: "Render Quality", valueField: "renderQuality", defaultValue: "raytracing", catalogId: "render-quality", entries: mapCat(RENDER_QUALITIES) },
31946
+ { kind: "single", nodeType: "composition-effects", label: "Composition Effect", valueField: "compositionEffect", defaultValue: "bursting-through-frame", catalogId: "composition-effects", entries: mapCat(COMPOSITION_EFFECTS) },
31947
+ { kind: "single", nodeType: "action-fx", label: "Action FX", valueField: "actionFx", defaultValue: "earthquake-tremor", catalogId: "action-fx", entries: mapCat(ACTION_FX, "category"), groupOrder: ACTION_FX_CATEGORY_ORDER, groupLabels: ACTION_FX_CATEGORY_LABELS },
31948
+ { kind: "single", nodeType: "loop-subject", label: "Loop Subject", valueField: "loopSubject", defaultValue: "tunnel", catalogId: "loop-subject", entries: mapCat(LOOP_SUBJECTS, "category"), groupOrder: LOOP_SUBJECT_CATEGORY_ORDER, groupLabels: LOOP_SUBJECT_CATEGORY_LABELS },
31949
+ { kind: "single", nodeType: "post-process-effects", label: "Post-Process Effect", valueField: "postProcess", defaultValue: "vignette-soft", catalogId: "post-process-effects", entries: mapCat(POST_PROCESS_EFFECTS) },
31950
+ // -------- "Camera" family --------
31951
+ { kind: "single", nodeType: "camera-motion", label: "Camera Motion", valueField: "cameraMotion", defaultValue: "static", catalogId: "camera-motions", entries: mapCat(CAMERA_MOTIONS, "category"), groupOrder: CAMERA_MOTION_CATEGORY_ORDER, groupLabels: CAMERA_MOTION_CATEGORY_LABELS },
31952
+ { kind: "single", nodeType: "lens", label: "Lens", valueField: "lens", defaultValue: "normal-50mm", catalogId: "lens", entries: mapCat(LENSES) },
31953
+ { kind: "single", nodeType: "camera-format", label: "Camera / Film", valueField: "cameraFormat", defaultValue: "35mm-film", catalogId: "camera-format", entries: mapCat(CAMERA_FORMATS) },
31954
+ { kind: "single", nodeType: "transition", label: "Transition", valueField: "transition", defaultValue: "auto", catalogId: "transitions", entries: mapCat(TRANSITIONS, "category"), groupOrder: TRANSITION_CATEGORY_ORDER, groupLabels: TRANSITION_CATEGORY_LABELS },
31955
+ { kind: "single", nodeType: "character-fx", label: "Character FX", valueField: "characterFx", defaultValue: "auto", catalogId: "character-fx", entries: mapCat(CHARACTER_FX, "category"), groupOrder: CHARACTER_FX_CATEGORY_ORDER, groupLabels: CHARACTER_FX_CATEGORY_LABELS },
31956
+ // -------- "Subject / Object" family --------
31957
+ { kind: "single", nodeType: "pose", label: "Pose", valueField: "pose", defaultValue: "standing-upright", catalogId: "pose", entries: mapCat(POSES, "category"), groupOrder: POSE_CATEGORY_ORDER, groupLabels: POSE_CATEGORY_LABELS },
31958
+ { kind: "single", nodeType: "material", label: "Material", valueField: "material", defaultValue: "silk", catalogId: "materials", entries: mapCat(MATERIALS, "category"), groupOrder: MATERIAL_CATEGORY_ORDER, groupLabels: MATERIAL_CATEGORY_LABELS },
31959
+ { kind: "single", nodeType: "animal", label: "Animal", valueField: "animal", defaultValue: "dog-golden-retriever", catalogId: "animals", entries: mapCat(shared.ANIMALS, "subcategory"), groupOrder: shared.ANIMAL_SUBCATEGORY_ORDER, groupLabels: shared.ANIMAL_SUBCATEGORY_LABELS },
31960
+ { kind: "single", nodeType: "vehicle", label: "Vehicle", valueField: "vehicle", defaultValue: "sedan", catalogId: "vehicles", entries: mapCat(shared.VEHICLES, "subcategory"), groupOrder: shared.VEHICLE_SUBCATEGORY_ORDER, groupLabels: shared.VEHICLE_SUBCATEGORY_LABELS },
31961
+ { kind: "single", nodeType: "weapon", label: "Weapon", valueField: "weapon", defaultValue: "katana", catalogId: "weapons", entries: mapCat(shared.WEAPONS, "subcategory"), groupOrder: shared.WEAPON_SUBCATEGORY_ORDER, groupLabels: shared.WEAPON_SUBCATEGORY_LABELS },
31962
+ { kind: "single", nodeType: "furniture", label: "Furniture", valueField: "furniture", defaultValue: "sofa", catalogId: "furniture", entries: mapCat(shared.FURNITURE, "subcategory"), groupOrder: shared.FURNITURE_SUBCATEGORY_ORDER, groupLabels: shared.FURNITURE_SUBCATEGORY_LABELS },
31963
+ { kind: "single", nodeType: "held-prop", label: "Held Prop", valueField: "heldProp", defaultValue: "smartphone", catalogId: "held-prop", entries: mapCat(HELD_PROPS, "category"), groupOrder: HELD_PROP_CATEGORY_ORDER, groupLabels: HELD_PROP_CATEGORY_LABELS }
31964
+ ];
31965
+ var STYLING_FIELDS2 = STYLING_DIMENSION_ORDER.map((d) => STYLING_FIELD_BY_DIMENSION[d]);
31966
+ var PERSON_FIELDS2 = PERSON_DIMENSION_ORDER.map((d) => PERSON_FIELD_BY_DIMENSION[d]);
31967
+ var LIGHTING_FIELDS2 = LIGHTING_CATEGORY_ORDER.map((c) => LIGHTING_FIELD_BY_CATEGORY[c]);
31968
+ var TEMPORAL_FIELD_BY_CATEGORY2 = {
31969
+ speed: "temporalSpeed",
31970
+ freeze: "temporalFreeze",
31971
+ direction: "temporalDirection",
31972
+ shutter: "temporalShutter"
31973
+ };
31974
+ var EXPOSURE_FIELD_BY_CATEGORY2 = {
31975
+ aperture: "aperture",
31976
+ "shutter-speed": "shutterSpeed",
31977
+ iso: "isoValue"
31978
+ };
31979
+ var MULTI_PICKER_WIRING = [
31980
+ {
31981
+ kind: "multi",
31982
+ nodeType: "framing",
31983
+ label: "Framing",
31984
+ fields: ["shotSize", "angle", "coverage", "composition", "vantage"],
31985
+ catalogId: "framing",
31986
+ catalogEntries: flatCat(FRAMINGS),
31987
+ fieldOptions: optionsByDiscriminator(FRAMINGS, (e) => e.category, FRAMING_FIELD_BY_CATEGORY)
31988
+ },
31989
+ {
31990
+ kind: "multi",
31991
+ nodeType: "lighting",
31992
+ label: "Lighting",
31993
+ fields: LIGHTING_FIELDS2,
31994
+ catalogId: "lighting",
31995
+ catalogEntries: flatCat(LIGHTINGS),
31996
+ fieldOptions: optionsByDiscriminator(LIGHTINGS, (e) => e.category, LIGHTING_FIELD_BY_CATEGORY)
31997
+ },
31998
+ {
31999
+ kind: "multi",
32000
+ nodeType: "person",
32001
+ label: "Person",
32002
+ fields: PERSON_FIELDS2,
32003
+ catalogId: "person",
32004
+ catalogEntries: flatCat(PEOPLE),
32005
+ fieldOptions: optionsByDiscriminator(PEOPLE, (e) => e.dimension, PERSON_FIELD_BY_DIMENSION)
32006
+ },
32007
+ {
32008
+ kind: "multi",
32009
+ nodeType: "styling",
32010
+ label: "Styling",
32011
+ fields: STYLING_FIELDS2,
32012
+ catalogId: "styling",
32013
+ catalogEntries: flatCat(STYLINGS),
32014
+ fieldOptions: optionsByDiscriminator(STYLINGS, (e) => e.dimension, STYLING_FIELD_BY_DIMENSION)
32015
+ },
32016
+ {
32017
+ kind: "multi",
32018
+ nodeType: "temporal",
32019
+ label: "Temporal",
32020
+ fields: ["temporalSpeed", "temporalFreeze", "temporalDirection", "temporalShutter"],
32021
+ catalogId: "temporal",
32022
+ catalogEntries: flatCat(TEMPORALS),
32023
+ fieldOptions: optionsByDiscriminator(TEMPORALS, (e) => e.category, TEMPORAL_FIELD_BY_CATEGORY2)
32024
+ },
32025
+ {
32026
+ kind: "multi",
32027
+ nodeType: "exposure-settings",
32028
+ label: "Exposure Settings",
32029
+ fields: ["aperture", "shutterSpeed", "isoValue"],
32030
+ catalogId: "exposure-settings",
32031
+ catalogEntries: flatCat(EXPOSURE_SETTINGS),
32032
+ fieldOptions: optionsByDiscriminator(EXPOSURE_SETTINGS, (e) => e.category, EXPOSURE_FIELD_BY_CATEGORY2)
32033
+ },
32034
+ // -------- "Sound" family --------
32035
+ // Music Genre catalog is hierarchical: flatten genres + every subgenre +
32036
+ // eras so summary chips can resolve any selected id back to a human label.
32037
+ {
32038
+ kind: "multi",
32039
+ nodeType: "music-genre",
32040
+ label: "Music Genre",
32041
+ fields: ["genre", "subgenre", "era"],
32042
+ catalogId: "music-genre",
32043
+ catalogEntries: [
32044
+ ...flatCat(MUSIC_GENRES),
32045
+ ...MUSIC_GENRES.flatMap((g) => g.subgenres.map((s) => ({ id: s.id, label: s.label }))),
32046
+ ...flatCat(MUSIC_ERAS)
32047
+ ],
32048
+ fieldOptions: {
32049
+ genre: flatCat(MUSIC_GENRES),
32050
+ subgenre: MUSIC_GENRES.flatMap((g) => g.subgenres.map((s) => ({ id: s.id, label: s.label }))),
32051
+ era: flatCat(MUSIC_ERAS)
32052
+ }
32053
+ },
32054
+ {
32055
+ kind: "multi",
32056
+ nodeType: "music-mood",
32057
+ label: "Music Mood",
32058
+ fields: ["energy", "emotion", "vibe"],
32059
+ catalogId: "music-mood",
32060
+ catalogEntries: [...flatCat(MUSIC_ENERGIES), ...flatCat(MUSIC_EMOTIONS), ...flatCat(MUSIC_VIBES)],
32061
+ fieldOptions: {
32062
+ energy: flatCat(MUSIC_ENERGIES),
32063
+ emotion: flatCat(MUSIC_EMOTIONS),
32064
+ vibe: flatCat(MUSIC_VIBES)
32065
+ }
32066
+ },
32067
+ {
32068
+ kind: "multi",
32069
+ nodeType: "instrumentation",
32070
+ label: "Instrumentation",
32071
+ fields: ["instruments", "production", "vocalPresence", "singingStyle"],
32072
+ catalogId: "instrumentation",
32073
+ catalogEntries: [...flatCat(INSTRUMENTS), ...flatCat(PRODUCTION_STYLES), ...flatCat(VOCAL_PRESENCE), ...flatCat(SINGING_STYLES)],
32074
+ fieldOptions: {
32075
+ instruments: flatCat(INSTRUMENTS),
32076
+ production: flatCat(PRODUCTION_STYLES),
32077
+ vocalPresence: flatCat(VOCAL_PRESENCE),
32078
+ singingStyle: flatCat(SINGING_STYLES)
32079
+ }
32080
+ },
32081
+ {
32082
+ kind: "multi",
32083
+ nodeType: "voice-character",
32084
+ label: "Voice Character",
32085
+ fields: ["age", "gender", "language", "accent", "timbre"],
32086
+ catalogId: "voice-character",
32087
+ catalogEntries: [...flatCat(VOICE_AGES), ...flatCat(VOICE_GENDERS), ...flatCat(VOICE_LANGUAGES), ...flatCat(VOICE_ACCENTS), ...flatCat(VOICE_TIMBRES)],
32088
+ fieldOptions: {
32089
+ age: flatCat(VOICE_AGES),
32090
+ gender: flatCat(VOICE_GENDERS),
32091
+ language: flatCat(VOICE_LANGUAGES),
32092
+ accent: flatCat(VOICE_ACCENTS),
32093
+ timbre: flatCat(VOICE_TIMBRES)
32094
+ }
32095
+ },
32096
+ {
32097
+ kind: "multi",
32098
+ nodeType: "voice-delivery",
32099
+ label: "Voice Delivery",
32100
+ fields: ["pace", "emotion", "archetype"],
32101
+ catalogId: "voice-delivery",
32102
+ catalogEntries: [...flatCat(VOICE_PACES), ...flatCat(VOICE_EMOTIONS), ...flatCat(VOICE_ARCHETYPES)],
32103
+ fieldOptions: {
32104
+ pace: flatCat(VOICE_PACES),
32105
+ emotion: flatCat(VOICE_EMOTIONS),
32106
+ archetype: flatCat(VOICE_ARCHETYPES)
32107
+ }
32108
+ }
32109
+ ];
32110
+ var ALL_PICKER_WIRING = [
32111
+ ...SINGLE_PICKER_WIRING,
32112
+ ...MULTI_PICKER_WIRING
32113
+ ];
32114
+ var WIRING_MAP = new Map(ALL_PICKER_WIRING.map((w) => [w.nodeType, w]));
32115
+ function getPickerWiring(nodeType) {
32116
+ if (!nodeType) return void 0;
32117
+ return WIRING_MAP.get(nodeType);
32118
+ }
31514
32119
 
31515
32120
  exports.ACTION_FX = ACTION_FX;
31516
32121
  exports.ACTION_FX_CATEGORY_LABELS = ACTION_FX_CATEGORY_LABELS;
@@ -31520,6 +32125,7 @@ exports.AESTHETICS = AESTHETICS;
31520
32125
  exports.AESTHETIC_CATEGORY_LABELS = AESTHETIC_CATEGORY_LABELS;
31521
32126
  exports.AESTHETIC_CATEGORY_ORDER = AESTHETIC_CATEGORY_ORDER;
31522
32127
  exports.AESTHETIC_IDS = AESTHETIC_IDS;
32128
+ exports.ALL_PICKER_WIRING = ALL_PICKER_WIRING;
31523
32129
  exports.ANALYZABLE_PICKER_TYPES = ANALYZABLE_PICKER_TYPES;
31524
32130
  exports.ANGLE_LABELS = ANGLE_LABELS;
31525
32131
  exports.ASPECT_RATIO_LABELS = ASPECT_RATIO_LABELS;
@@ -31602,6 +32208,7 @@ exports.MOOD_CATEGORY_LABELS = MOOD_CATEGORY_LABELS;
31602
32208
  exports.MOOD_CATEGORY_ORDER = MOOD_CATEGORY_ORDER;
31603
32209
  exports.MOOD_IDS = MOOD_IDS;
31604
32210
  exports.MOVEMENT_LABELS = MOVEMENT_LABELS;
32211
+ exports.MULTI_PICKER_WIRING = MULTI_PICKER_WIRING;
31605
32212
  exports.MUSIC_EMOTIONS = MUSIC_EMOTIONS;
31606
32213
  exports.MUSIC_ENERGIES = MUSIC_ENERGIES;
31607
32214
  exports.MUSIC_ERAS = MUSIC_ERAS;
@@ -31635,6 +32242,7 @@ exports.PHOTO_GENRES = PHOTO_GENRES;
31635
32242
  exports.PHOTO_GENRE_CATEGORY_LABELS = PHOTO_GENRE_CATEGORY_LABELS;
31636
32243
  exports.PHOTO_GENRE_CATEGORY_ORDER = PHOTO_GENRE_CATEGORY_ORDER;
31637
32244
  exports.PHOTO_GENRE_IDS = PHOTO_GENRE_IDS;
32245
+ exports.PICKER_ANALYZER_FAMILIES = PICKER_ANALYZER_FAMILIES;
31638
32246
  exports.PICKER_ANALYZER_REGISTRY = PICKER_ANALYZER_REGISTRY;
31639
32247
  exports.PICKER_CATALOGS = PICKER_CATALOGS;
31640
32248
  exports.PICKER_TYPES = PICKER_TYPES;
@@ -31660,6 +32268,7 @@ exports.SETTING_CATEGORY_LABELS = SETTING_CATEGORY_LABELS;
31660
32268
  exports.SETTING_IDS = SETTING_IDS;
31661
32269
  exports.SHOT_LABELS = SHOT_LABELS;
31662
32270
  exports.SINGING_STYLES = SINGING_STYLES;
32271
+ exports.SINGLE_PICKER_WIRING = SINGLE_PICKER_WIRING;
31663
32272
  exports.SNIPPET_MEDIA_VALUES = SNIPPET_MEDIA_VALUES;
31664
32273
  exports.STYLES = STYLES;
31665
32274
  exports.STYLE_IDS = STYLE_IDS;
@@ -31837,6 +32446,7 @@ exports.getPhotographerLabel = getPhotographerLabel;
31837
32446
  exports.getPhotographerPromptHint = getPhotographerPromptHint;
31838
32447
  exports.getPickerAnalyzer = getPickerAnalyzer;
31839
32448
  exports.getPickerCatalog = getPickerCatalog;
32449
+ exports.getPickerWiring = getPickerWiring;
31840
32450
  exports.getPose = getPose;
31841
32451
  exports.getPoseLabel = getPoseLabel;
31842
32452
  exports.getPosePromptHint = getPosePromptHint;
@@ -31881,6 +32491,7 @@ exports.getWardrobeEntry = getWardrobeEntry;
31881
32491
  exports.getWardrobePromptHint = getWardrobePromptHint;
31882
32492
  exports.groupFactoryPresets = groupFactoryPresets;
31883
32493
  exports.hasUpstreamCharacter = hasUpstreamCharacter;
32494
+ exports.identityRefsSentence = identityRefsSentence;
31884
32495
  exports.isAnalyzablePicker = isAnalyzablePicker;
31885
32496
  exports.isInstrumentalVocal = isInstrumentalVocal;
31886
32497
  exports.isVantageFraming = isVantageFraming;
@@ -31894,11 +32505,13 @@ exports.referenceRulesBlock = referenceRulesBlock;
31894
32505
  exports.renderStructuredFields = renderStructuredFields;
31895
32506
  exports.resolveBrandInput = resolveBrandInput;
31896
32507
  exports.resolveCharacterMentions = resolveCharacterMentions;
32508
+ exports.resolveGeminiOmniI2vInputs = resolveGeminiOmniI2vInputs;
31897
32509
  exports.resolveLocationMentions = resolveLocationMentions;
31898
32510
  exports.resolvePrompt = resolvePrompt;
31899
32511
  exports.resolveReferenceTokens = resolveReferenceTokens;
31900
32512
  exports.resolveSeedance2Inputs = resolveSeedance2Inputs;
31901
32513
  exports.resolveTemplate = resolveTemplate;
32514
+ exports.resolveVeoI2vInputs = resolveVeoI2vInputs;
31902
32515
  exports.resolveVideoReferenceCore = resolveVideoReferenceCore;
31903
32516
  exports.summarizePickerCatalogs = summarizePickerCatalogs;
31904
32517
  exports.toIdentityLockMode = toIdentityLockMode;