vidspotai-shared 1.0.67 → 1.0.69-dev.0

package/lib/globals/aiModels/providers/runway.js

@@ -5,11 +5,24 @@ const types_1 = require("../../../services/aiGen/providers/runway/types");
 exports.runwayConfigs = {
     "runway-gen4.5": {
         modelId: "gen4.5",
+        // AG-16: gen4.5 supports both T2V and I2V via /text_to_video and
+        // /image_to_video respectively (Runway docs confirm — accepted on both
+        // endpoints). Aspect ratios broadened to Runway's full pixel-ratio set
+        // (T2V accepts "1280:720" + "720:1280"; I2V accepts the same 6-ratio set
+        // as gen4_turbo). The service path translates canonical "9:16"/"16:9"
+        // from the agent into pixel ratios at request time.
         type: ["text-to-video", "image-to-video"],
+        tier: "T3",
         fields: {
             prompt: { required: true },
-            aspectRatio: { allowedValues: ["1280x720", "720x1280"] },
-            duration: { allowedValues: [2, 3, 4, 5, 6, 7, 8, 9, 10] }, // in seconds
+            inputImageUrl: {}, // optional — present → I2V path, absent → T2V path
+            aspectRatio: {
+                allowedValues: ["1280:720", "720:1280", "1104:832", "960:960", "832:1104", "1584:672"],
+            },
+            duration: { allowedValues: [2, 3, 4, 5, 6, 7, 8, 9, 10] },
+            seed: {},
+            publicFigureThreshold: { allowedValues: ["auto", "low"] },
+            cameraControl: {},
         },
         cost: {
             perSecond: 0.12,
@@ -18,6 +31,7 @@ exports.runwayConfigs = {
     "runway-gen4-turbo": {
         modelId: "gen4_turbo",
         type: ["image-to-video"],
+        tier: "T2",
         fields: {
             prompt: { required: true },
             inputImageUrl: { required: true },
@@ -27,6 +41,10 @@ exports.runwayConfigs = {
             duration: {
                 allowedValues: types_1.GEN4_TURBO_DURATIONS,
             }, // in seconds
+            lastFrameImageUrl: {}, // promptImage position:"last" for first-last interpolation
+            seed: {},
+            publicFigureThreshold: { allowedValues: ["auto", "low"] },
+            cameraControl: {},
         },
         // parallelRequests: 1,
         cost: {
@@ -36,6 +54,7 @@ exports.runwayConfigs = {
     "runway-gen4-aleph": {
         modelId: "gen4_aleph",
         type: ["video-to-video"],
+        tier: "T3",
         fields: {
             prompt: { required: true },
             inputVideoUrl: { required: true },
@@ -43,10 +62,53 @@ exports.runwayConfigs = {
                 allowedValues: types_1.GEN4_ALEPH_RATIOS,
             },
             duration: { allowedValues: types_1.GEN4_ALEPH_DURATIONS }, // in seconds
+            // Aleph references[] — image refs for character/style consistency. Runway
+            // caps Gen4 references at 3 per call.
+            referenceImageUrls: { max: 3 },
+            seed: {},
+            publicFigureThreshold: { allowedValues: ["auto", "low"] },
+            cameraControl: {},
         },
         // parallelRequests: 1,
         cost: {
             perSecond: 0.15,
         },
     },
+    // --- Image models ---
+    "runwayml-gen4-image": {
+        modelId: "gen4_image",
+        type: ["text-to-image", "image-to-image"],
+        tier: "T2",
+        fields: {
+            prompt: { required: true },
+            aspectRatio: {
+                allowedValues: ["1920:1080", "1080:1920", "1024:1024", "1360:768", "768:1360"],
+            },
+            inputImageUrls: { max: 3 }, // referenceImages[] — up to 3 tagged refs
+            seed: {},
+        },
+        // Runway: $0.05 @ 720p tier, $0.08 @ 1080p tier. Square = 720p tier.
+        cost: {
+            perAspect: {
+                "1920:1080": 0.08, "1080:1920": 0.08,
+                "1360:768": 0.05, "768:1360": 0.05,
+                "1024:1024": 0.05,
+            },
+            fixed: 0.05, // fallback when aspect not supplied
+        },
+    },
+    "runwayml-gen4-image-turbo": {
+        modelId: "gen4_image_turbo",
+        type: ["text-to-image", "image-to-image"],
+        tier: "T1",
+        fields: {
+            prompt: { required: true },
+            aspectRatio: {
+                allowedValues: ["1920:1080", "1080:1920", "1024:1024", "1360:768", "768:1360"],
+            },
+            inputImageUrls: { max: 3 }, // referenceImages[] — up to 3 tagged refs
+            seed: {},
+        },
+        cost: { fixed: 0.02 },
+    },
 };

package/lib/globals/aiModels/tierHelpers.d.ts

@@ -0,0 +1,33 @@
+import type { Tier } from "../../schemas/project.schema";
+import type { GenerationType, IModelSchema } from "./types";
+/**
+ * Helpers for the agent planner to query models by tier and capability.
+ * Designed to be cheap and side-effect free — safe to call inside hot planner loops.
+ */
+export interface ModelMatch {
+    modelKey: string;
+    config: IModelSchema;
+}
+/** Return every model whose tier matches. Optionally filter by GenerationType. */
+export declare function getModelsByTier(tier: Tier, type?: GenerationType): ModelMatch[];
+/** Return the cheapest model satisfying (tier, type). Uses cost.fixed → cost.perSecond → 0. */
+export declare function getCheapestModel(tier: Tier, type: GenerationType): ModelMatch | undefined;
+/** Return all models supporting a generation type, regardless of tier. */
+export declare function getModelsByType(type: GenerationType): ModelMatch[];
+/**
+ * Build a compact, planner-readable summary of the model catalog grouped by
+ * tier. Injected into the planner system prompt so the LLM can pick a concrete
+ * `preferredModel` per scene with cost/quality awareness — e.g. cheap Kling
+ * for b-roll, premium Veo for hero shots.
+ *
+ * Token budget matters here: ~10–20 lines, no JSON, sorted cheapest-first per
+ * tier so the planner's default tendency aligns with cost minimization.
+ */
+export declare function summarizeModelsForPlanner(type?: GenerationType): string;
+/**
+ * AG-15: returns the allowed scene durations (seconds) for a given modelKey,
+ * or null if the model has no duration constraint declared. Used by the
+ * planner's post-plan capability validator.
+ */
+export declare function getModelAllowedDurations(modelKey: string): number[] | null;
+//# sourceMappingURL=tierHelpers.d.ts.map

package/lib/globals/aiModels/tierHelpers.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"tierHelpers.d.ts","sourceRoot":"","sources":["../../../src/globals/aiModels/tierHelpers.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,8BAA8B,CAAC;AAEzD,OAAO,KAAK,EAAE,cAAc,EAAE,YAAY,EAAE,MAAM,SAAS,CAAC;AAE5D;;;GAGG;AAEH,MAAM,WAAW,UAAU;IACzB,QAAQ,EAAE,MAAM,CAAC;IACjB,MAAM,EAAE,YAAY,CAAC;CACtB;AAED,kFAAkF;AAClF,wBAAgB,eAAe,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC,EAAE,cAAc,GAAG,UAAU,EAAE,CAQ/E;AAED,+FAA+F;AAC/F,wBAAgB,gBAAgB,CAC9B,IAAI,EAAE,IAAI,EACV,IAAI,EAAE,cAAc,GACnB,UAAU,GAAG,SAAS,CAIxB;AAED,0EAA0E;AAC1E,wBAAgB,eAAe,CAAC,IAAI,EAAE,cAAc,GAAG,UAAU,EAAE,CAMlE;AAQD;;;;;;;;GAQG;AACH,wBAAgB,yBAAyB,CACvC,IAAI,GAAE,cAAgC,GACrC,MAAM,CAeR;AA6BD;;;;GAIG;AACH,wBAAgB,wBAAwB,CAAC,QAAQ,EAAE,MAAM,GAAG,MAAM,EAAE,GAAG,IAAI,CAU1E"}

package/lib/globals/aiModels/tierHelpers.js

@@ -0,0 +1,109 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.getModelsByTier = getModelsByTier;
+exports.getCheapestModel = getCheapestModel;
+exports.getModelsByType = getModelsByType;
+exports.summarizeModelsForPlanner = summarizeModelsForPlanner;
+exports.getModelAllowedDurations = getModelAllowedDurations;
+const index_1 = require("./index");
+/** Return every model whose tier matches. Optionally filter by GenerationType. */
+function getModelsByTier(tier, type) {
+    const matches = [];
+    for (const [modelKey, config] of Object.entries(index_1.aiModelConfigs)) {
+        if (config.tier !== tier)
+            continue;
+        if (type && !config.type.includes(type))
+            continue;
+        matches.push({ modelKey, config });
+    }
+    return matches;
+}
+/** Return the cheapest model satisfying (tier, type). Uses cost.fixed → cost.perSecond → 0. */
+function getCheapestModel(tier, type) {
+    const candidates = getModelsByTier(tier, type);
+    if (candidates.length === 0)
+        return undefined;
+    return candidates.sort((a, b) => costScore(a.config) - costScore(b.config))[0];
+}
+/** Return all models supporting a generation type, regardless of tier. */
+function getModelsByType(type) {
+    const matches = [];
+    for (const [modelKey, config] of Object.entries(index_1.aiModelConfigs)) {
+        if (config.type.includes(type))
+            matches.push({ modelKey, config });
+    }
+    return matches;
+}
+function costScore(c) {
+    if (c.cost?.fixed != null)
+        return c.cost.fixed;
+    if (c.cost?.perSecond != null)
+        return c.cost.perSecond;
+    return Number.POSITIVE_INFINITY;
+}
+/**
+ * Build a compact, planner-readable summary of the model catalog grouped by
+ * tier. Injected into the planner system prompt so the LLM can pick a concrete
+ * `preferredModel` per scene with cost/quality awareness — e.g. cheap Kling
+ * for b-roll, premium Veo for hero shots.
+ *
+ * Token budget matters here: ~10–20 lines, no JSON, sorted cheapest-first per
+ * tier so the planner's default tendency aligns with cost minimization.
+ */
+function summarizeModelsForPlanner(type = "text-to-video") {
+    const lines = [];
+    for (const tier of ["T0", "T1", "T2", "T3"]) {
+        const matches = getModelsByTier(tier, type);
+        if (matches.length === 0)
+            continue;
+        matches.sort((a, b) => costScore(a.config) - costScore(b.config));
+        lines.push(`  ${tier}:`);
+        for (const m of matches) {
+            const cost = formatCost(m.config);
+            const dur = formatDurations(m.config);
+            const meta = [cost, dur].filter(Boolean).join(", ");
+            lines.push(`    - ${m.modelKey}${meta ? ` (${meta})` : ""}`);
+        }
+    }
+    return lines.length ? lines.join("\n") : "";
+}
+function formatCost(c) {
+    if (c.cost?.fixed != null)
+        return `$${c.cost.fixed.toFixed(2)} fixed`;
+    if (c.cost?.perSecond != null)
+        return `$${c.cost.perSecond.toFixed(2)}/s`;
+    return "";
+}
+/**
+ * AG-15: emit each model's allowed scene durations into the planner catalog so
+ * the LLM can match scene.durationMs to a model that actually accepts it.
+ * Contiguous integer ranges collapse to "dur:2-12s"; sparse sets stay literal
+ * "dur:[6,10]s". The planner uses this + the duration-capability validator
+ * to avoid the Test-1 failure mode (1.7s scenes routed to Hailuo, which only
+ * supports 6 or 10s, → fallback chain exhausted, $0 of useful work).
+ */
+function formatDurations(c) {
+    const allowed = c.fields?.duration?.allowedValues;
+    if (!allowed?.length || !allowed.every((v) => typeof v === "number"))
+        return "";
+    const sorted = [...allowed].sort((a, b) => a - b);
+    const first = sorted[0];
+    const last = sorted[sorted.length - 1];
+    const contiguous = sorted.length >= 3 &&
+        sorted.every((v, i) => i === 0 || v === sorted[i - 1] + 1);
+    if (contiguous)
+        return `dur:${first}-${last}s`;
+    return `dur:[${sorted.join(",")}]s`;
+}
+/**
+ * AG-15: returns the allowed scene durations (seconds) for a given modelKey,
+ * or null if the model has no duration constraint declared. Used by the
+ * planner's post-plan capability validator.
+ */
+function getModelAllowedDurations(modelKey) {
+    const cfg = index_1.aiModelConfigs[modelKey];
+    const allowed = cfg?.fields?.duration?.allowedValues;
+    if (!allowed?.length || !allowed.every((v) => typeof v === "number"))
+        return null;
+    return [...allowed].sort((a, b) => a - b);
+}

package/lib/globals/aiModels/types.d.ts

@@ -1,5 +1,6 @@
 import { TtsProvider } from "../ttsModels/types";
-export type GenerationType = "text-to-video" | "image-to-video" | "video-to-video" | "avatar-video" | "text-to-text" | "image-to-text";
+import type { Tier } from "../../schemas/project.schema";
+export type GenerationType = "text-to-video" | "image-to-video" | "video-to-video" | "audio-to-video" | "avatar-video" | "motion-control" | "lip-sync" | "text-to-text" | "image-to-text" | "text-to-image" | "image-to-image" | "stock-video" | "stock-image" | "stock-music" | "text-to-music" | "sound-effects";
 /**
  * Config block present on every avatar-video type model.
  * Encapsulates all avatar-specific constraints and provider defaults
@@ -43,6 +44,14 @@ export interface ISchemaField {
     required?: boolean;
     allowedValues?: string[] | number[] | boolean[];
     default?: string | number | boolean;
+    /**
+     * Maximum item count when the field value is an array (e.g. URL lists like
+     * `referenceImageUrls`, `inputImageUrls`, `referenceVideoUrls`). Enforced
+     * centrally in `validateParams`. Leave unset to mean "no declared cap"
+     * (validation passes regardless of length). Frontend reads the same cap to
+     * gate uploads — keep this in sync with the provider's hard limit.
+     */
+    max?: number;
 }
 export interface ICostTable {
     [key: string]: {
@@ -52,6 +61,16 @@ export interface ICostTable {
 export interface IModelSchema {
     modelId: string;
     type: GenerationType[];
+    /**
+     * Generation tier for the agent's cost-aware planning.
+     *   T0 — stock (~free)
+     *   T1 — AI image + motion (~$0.02–0.05/scene)
+     *   T2 — image-to-video on key frames (~$0.10–0.40/scene)
+     *   T3 — pure text-to-video (~$0.50–2.00/scene)
+     *
+     * Optional during rollout; planner falls back to `cost` when absent.
+     */
+    tier?: Tier;
     fields: Record<string, ISchemaField>;
     resolutionRules?: Record<string, string[]>;
     durationRules?: Record<number, string[]>;

package/lib/globals/aiModels/types.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../../src/globals/aiModels/types.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,oBAAoB,CAAC;AAEjD,MAAM,MAAM,cAAc,GACtB,eAAe,GACf,gBAAgB,GAChB,gBAAgB,GAChB,cAAc,GACd,cAAc,GACd,eAAe,CAAC;AAEpB;;;;GAIG;AACH,MAAM,WAAW,kBAAkB;IACjC,wFAAwF;IACxF,gBAAgB,EAAE,MAAM,CAAC;IACzB,gDAAgD;IAChD,gBAAgB,EAAE,MAAM,CAAC;IACzB,kEAAkE;IAClE,WAAW,EAAE,WAAW,CAAC;IACzB;;;;OAIG;IACH,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB;;;;;OAKG;IACH,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB;;;;;OAKG;IACH,sBAAsB,CAAC,EAAE,MAAM,CAAC;IAChC;;;;OAIG;IACH,kBAAkB,CAAC,EAAE,MAAM,CAAC;CAC7B;AAED,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,EAAE,OAAO,CAAC;IACnB,aAAa,CAAC,EAAE,MAAM,EAAE,GAAG,MAAM,EAAE,GAAG,OAAO,EAAE,CAAC;IAChD,OAAO,CAAC,EAAE,MAAM,GAAG,MAAM,GAAG,OAAO,CAAC;CACrC;AAED,MAAM,WAAW,UAAU;IACzB,CAAC,GAAG,EAAE,MAAM,GAAG;QACb,CAAC,GAAG,EAAE,MAAM,GAAG,MAAM,CAAC;KACvB,CAAC;CACH;AAED,MAAM,WAAW,YAAY;IAC3B,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,cAAc,EAAE,CAAC;IACvB,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,YAAY,CAAC,CAAC;IAErC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC;IAC3C,aAAa,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC;IACzC,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB;;;OAGG;IACH,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB;;;OAGG;IACH,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,IAAI,CAAC,EAAE;QACL,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,WAAW,CAAC,EAAE,MAAM,CAAC;QACrB,KAAK,CAAC,EAAE,UAAU,CAAC;QACnB,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,CAAC,GAAG,EAAE,MAAM,GAAG,GAAG,CAAC;KACpB,CAAC;IACF,4FAA4F;IAC5F,MAAM,CAAC,EAAE,kBAAkB,CAAC;IAC5B,kHAAkH;IAClH,oBAAoB,CAAC,EAAE,MAAM,EAAE,CAAC;CACjC"}
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../../src/globals/aiModels/types.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,oBAAoB,CAAC;AACjD,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,8BAA8B,CAAC;AAEzD,MAAM,MAAM,cAAc,GACtB,eAAe,GACf,gBAAgB,GAChB,gBAAgB,GAChB,gBAAgB,GAChB,cAAc,GACd,gBAAgB,GAChB,UAAU,GACV,cAAc,GACd,eAAe,GACf,eAAe,GACf,gBAAgB,GAChB,aAAa,GACb,aAAa,GACb,aAAa,GACb,eAAe,GACf,eAAe,CAAC;AAEpB;;;;GAIG;AACH,MAAM,WAAW,kBAAkB;IACjC,wFAAwF;IACxF,gBAAgB,EAAE,MAAM,CAAC;IACzB,gDAAgD;IAChD,gBAAgB,EAAE,MAAM,CAAC;IACzB,kEAAkE;IAClE,WAAW,EAAE,WAAW,CAAC;IACzB;;;;OAIG;IACH,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB;;;;;OAKG;IACH,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB;;;;;OAKG;IACH,sBAAsB,CAAC,EAAE,MAAM,CAAC;IAChC;;;;OAIG;IACH,kBAAkB,CAAC,EAAE,MAAM,CAAC;CAC7B;AAED,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,EAAE,OAAO,CAAC;IACnB,aAAa,CAAC,EAAE,MAAM,EAAE,GAAG,MAAM,EAAE,GAAG,OAAO,EAAE,CAAC;IAChD,OAAO,CAAC,EAAE,MAAM,GAAG,MAAM,GAAG,OAAO,CAAC;IACpC;;;;;;OAMG;IACH,GAAG,CAAC,EAAE,MAAM,CAAC;CACd;AAED,MAAM,WAAW,UAAU;IACzB,CAAC,GAAG,EAAE,MAAM,GAAG;QACb,CAAC,GAAG,EAAE,MAAM,GAAG,MAAM,CAAC;KACvB,CAAC;CACH;AAED,MAAM,WAAW,YAAY;IAC3B,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,cAAc,EAAE,CAAC;IACvB;;;;;;;;OAQG;IACH,IAAI,CAAC,EAAE,IAAI,CAAC;IACZ,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,YAAY,CAAC,CAAC;IAErC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC;IAC3C,aAAa,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC;IACzC,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB;;;OAGG;IACH,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB;;;OAGG;IACH,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,IAAI,CAAC,EAAE;QACL,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,WAAW,CAAC,EAAE,MAAM,CAAC;QACrB,KAAK,CAAC,EAAE,UAAU,CAAC;QACnB,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,CAAC,GAAG,EAAE,MAAM,GAAG,GAAG,CAAC;KACpB,CAAC;IACF,4FAA4F;IAC5F,MAAM,CAAC,EAAE,kBAAkB,CAAC;IAC5B,kHAAkH;IAClH,oBAAoB,CAAC,EAAE,MAAM,EAAE,CAAC;CACjC"}

package/lib/globals/ttsModels/index.d.ts

@@ -2,6 +2,8 @@ export * from "./types";
 export * from "./voices";
 import { TtsProvider, ITtsProviderConfig } from "./types";
 export { elevenlabsConfig } from "./providers/elevenlabs";
+export { minimaxTtsConfig } from "./providers/minimax";
+export { openaiTtsConfig } from "./providers/openai";
 export declare const ttsProviderConfigs: Record<TtsProvider, ITtsProviderConfig>;
 /** Fallback TTS provider used when a model config does not specify one. */
 export declare const DEFAULT_TTS_PROVIDER: TtsProvider;

package/lib/globals/ttsModels/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/globals/ttsModels/index.ts"],"names":[],"mappings":"AAAA,cAAc,SAAS,CAAC;AACxB,cAAc,UAAU,CAAC;AAEzB,OAAO,EAAE,WAAW,EAAE,kBAAkB,EAAE,MAAM,SAAS,CAAC;AAG1D,OAAO,EAAE,gBAAgB,EAAE,MAAM,wBAAwB,CAAC;AAE1D,eAAO,MAAM,kBAAkB,EAAE,MAAM,CAAC,WAAW,EAAE,kBAAkB,CAEtE,CAAC;AAEF,2EAA2E;AAC3E,eAAO,MAAM,oBAAoB,EAAE,WAA0B,CAAC"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/globals/ttsModels/index.ts"],"names":[],"mappings":"AAAA,cAAc,SAAS,CAAC;AACxB,cAAc,UAAU,CAAC;AAEzB,OAAO,EAAE,WAAW,EAAE,kBAAkB,EAAE,MAAM,SAAS,CAAC;AAK1D,OAAO,EAAE,gBAAgB,EAAE,MAAM,wBAAwB,CAAC;AAC1D,OAAO,EAAE,gBAAgB,EAAE,MAAM,qBAAqB,CAAC;AACvD,OAAO,EAAE,eAAe,EAAE,MAAM,oBAAoB,CAAC;AAErD,eAAO,MAAM,kBAAkB,EAAE,MAAM,CAAC,WAAW,EAAE,kBAAkB,CAItE,CAAC;AAEF,2EAA2E;AAC3E,eAAO,MAAM,oBAAoB,EAAE,WAA0B,CAAC"}

package/lib/globals/ttsModels/index.js

@@ -14,14 +14,22 @@ var __exportStar = (this && this.__exportStar) || function(m, exports) {
     for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.DEFAULT_TTS_PROVIDER = exports.ttsProviderConfigs = exports.elevenlabsConfig = void 0;
+exports.DEFAULT_TTS_PROVIDER = exports.ttsProviderConfigs = exports.openaiTtsConfig = exports.minimaxTtsConfig = exports.elevenlabsConfig = void 0;
 __exportStar(require("./types"), exports);
 __exportStar(require("./voices"), exports);
 const elevenlabs_1 = require("./providers/elevenlabs");
+const minimax_1 = require("./providers/minimax");
+const openai_1 = require("./providers/openai");
 var elevenlabs_2 = require("./providers/elevenlabs");
 Object.defineProperty(exports, "elevenlabsConfig", { enumerable: true, get: function () { return elevenlabs_2.elevenlabsConfig; } });
+var minimax_2 = require("./providers/minimax");
+Object.defineProperty(exports, "minimaxTtsConfig", { enumerable: true, get: function () { return minimax_2.minimaxTtsConfig; } });
+var openai_2 = require("./providers/openai");
+Object.defineProperty(exports, "openaiTtsConfig", { enumerable: true, get: function () { return openai_2.openaiTtsConfig; } });
 exports.ttsProviderConfigs = {
     elevenlabs: elevenlabs_1.elevenlabsConfig,
+    minimax: minimax_1.minimaxTtsConfig,
+    openai: openai_1.openaiTtsConfig,
 };
 /** Fallback TTS provider used when a model config does not specify one. */
 exports.DEFAULT_TTS_PROVIDER = "elevenlabs";

package/lib/globals/ttsModels/providers/minimax.d.ts

@@ -0,0 +1,8 @@
+import { ITtsProviderConfig } from "../types";
+/**
+ * MiniMax Speech-02-HD — high-fidelity multilingual TTS.
+ * API: POST https://api.minimax.io/v1/t2a_v2 returns hex-encoded audio inline.
+ * Pricing: roughly $0.10 per 1K chars (T2A HD tier).
+ */
+export declare const minimaxTtsConfig: ITtsProviderConfig;
+//# sourceMappingURL=minimax.d.ts.map

package/lib/globals/ttsModels/providers/minimax.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"minimax.d.ts","sourceRoot":"","sources":["../../../../src/globals/ttsModels/providers/minimax.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,MAAM,UAAU,CAAC;AAE9C;;;;GAIG;AACH,eAAO,MAAM,gBAAgB,EAAE,kBAS9B,CAAC"}

package/lib/globals/ttsModels/providers/minimax.js

@@ -0,0 +1,18 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.minimaxTtsConfig = void 0;
+/**
+ * MiniMax Speech-02-HD — high-fidelity multilingual TTS.
+ * API: POST https://api.minimax.io/v1/t2a_v2 returns hex-encoded audio inline.
+ * Pricing: roughly $0.10 per 1K chars (T2A HD tier).
+ */
+exports.minimaxTtsConfig = {
+    name: "MiniMax Speech-02 HD",
+    costPer1KChars: 0.1,
+    defaultVoiceId: "male-qn-qingse",
+    modelId: "speech-02-hd",
+    wordsPerSecond: 2.5,
+    maxCharsPerRequest: 5000,
+    durationSafetyBufferSeconds: 100,
+    concurrentRequests: 5,
+};

package/lib/globals/ttsModels/providers/openai.d.ts

@@ -0,0 +1,12 @@
+import { ITtsProviderConfig } from "../types";
+/**
+ * OpenAI gpt-4o-mini-tts — cheap, fast TTS with optional `instructions` styling.
+ * API: POST https://api.openai.com/v1/audio/speech returns mp3 binary directly.
+ *
+ * Pricing (2026): token-based — $0.60/1M text + $12/1M audio out, effectively
+ * $0.015 per minute of generated audio. We convert to per-1K-chars for the
+ * shared `costPer1KChars` field at ~2.5 wps × ~5 chars/word ≈ 12.5 chars/sec
+ * ≈ 750 chars/min → $0.020/1K chars. Stored at that converted rate.
+ */
+export declare const openaiTtsConfig: ITtsProviderConfig;
+//# sourceMappingURL=openai.d.ts.map

package/lib/globals/ttsModels/providers/openai.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"openai.d.ts","sourceRoot":"","sources":["../../../../src/globals/ttsModels/providers/openai.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,kBAAkB,EAAE,MAAM,UAAU,CAAC;AAE9C;;;;;;;;GAQG;AACH,eAAO,MAAM,eAAe,EAAE,kBAS7B,CAAC"}

package/lib/globals/ttsModels/providers/openai.js

@@ -0,0 +1,22 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.openaiTtsConfig = void 0;
+/**
+ * OpenAI gpt-4o-mini-tts — cheap, fast TTS with optional `instructions` styling.
+ * API: POST https://api.openai.com/v1/audio/speech returns mp3 binary directly.
+ *
+ * Pricing (2026): token-based — $0.60/1M text + $12/1M audio out, effectively
+ * $0.015 per minute of generated audio. We convert to per-1K-chars for the
+ * shared `costPer1KChars` field at ~2.5 wps × ~5 chars/word ≈ 12.5 chars/sec
+ * ≈ 750 chars/min → $0.020/1K chars. Stored at that converted rate.
+ */
+exports.openaiTtsConfig = {
+    name: "OpenAI gpt-4o-mini-tts",
+    costPer1KChars: 0.02,
+    defaultVoiceId: "alloy",
+    modelId: "gpt-4o-mini-tts",
+    wordsPerSecond: 2.5,
+    maxCharsPerRequest: 4096,
+    durationSafetyBufferSeconds: 60,
+    concurrentRequests: 5,
+};

package/lib/globals/ttsModels/types.d.ts

@@ -1,5 +1,5 @@
 /** Union of all supported TTS provider keys. Extend as new providers are added. */
-export type TtsProvider = "elevenlabs";
+export type TtsProvider = "elevenlabs" | "minimax" | "openai";
 export interface ITtsProviderConfig {
     /** Human-readable provider name */
     name: string;

package/lib/globals/ttsModels/types.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../../src/globals/ttsModels/types.ts"],"names":[],"mappings":"AAAA,mFAAmF;AACnF,MAAM,MAAM,WAAW,GAAG,YAAY,CAAC;AAEvC,MAAM,WAAW,kBAAkB;IACjC,mCAAmC;IACnC,IAAI,EAAE,MAAM,CAAC;IACb,4DAA4D;IAC5D,cAAc,EAAE,MAAM,CAAC;IACvB,gDAAgD;IAChD,cAAc,EAAE,MAAM,CAAC;IACvB,mDAAmD;IACnD,OAAO,EAAE,MAAM,CAAC;IAChB,mFAAmF;IACnF,cAAc,EAAE,MAAM,CAAC;IACvB;;;OAGG;IACH,kBAAkB,EAAE,MAAM,CAAC;IAC3B;;;;;OAKG;IACH,2BAA2B,EAAE,MAAM,CAAC;IACpC,+EAA+E;IAC/E,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,iEAAiE;IACjE,aAAa,CAAC,EAAE,MAAM,CAAC;CACxB"}
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../../src/globals/ttsModels/types.ts"],"names":[],"mappings":"AAAA,mFAAmF;AACnF,MAAM,MAAM,WAAW,GAAG,YAAY,GAAG,SAAS,GAAG,QAAQ,CAAC;AAE9D,MAAM,WAAW,kBAAkB;IACjC,mCAAmC;IACnC,IAAI,EAAE,MAAM,CAAC;IACb,4DAA4D;IAC5D,cAAc,EAAE,MAAM,CAAC;IACvB,gDAAgD;IAChD,cAAc,EAAE,MAAM,CAAC;IACvB,mDAAmD;IACnD,OAAO,EAAE,MAAM,CAAC;IAChB,mFAAmF;IACnF,cAAc,EAAE,MAAM,CAAC;IACvB;;;OAGG;IACH,kBAAkB,EAAE,MAAM,CAAC;IAC3B;;;;;OAKG;IACH,2BAA2B,EAAE,MAAM,CAAC;IACpC,+EAA+E;IAC/E,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,iEAAiE;IACjE,aAAa,CAAC,EAAE,MAAM,CAAC;CACxB"}

package/lib/globals/ttsModels/voices.d.ts

@@ -3,22 +3,57 @@ export interface ITtsVoice {
     name: string;
     gender: "male" | "female";
     accent: string;
-    /** Public preview MP3 hosted by ElevenLabs */
+    /** Public preview MP3 hosted by ElevenLabs (or empty when provider doesn't expose one) */
     previewUrl: string;
+    /** Apparent age bucket — agents pick voices to match brief demographics */
+    age?: "young" | "middle_aged" | "old";
+    /** One-line vibe label the agent reasons about ("warm narrator", "energetic hype", etc.) */
+    persona?: string;
+    /** Free-text descriptor surfaced in tool descriptions / UI tooltips */
+    description?: string;
+    /** Suggested use cases — narrative, advertisement, conversational, characters, audiobook, news */
+    useCases?: Array<"advertisement" | "narrative" | "conversational" | "characters" | "audiobook" | "news" | "social_media" | "informative_educational">;
+    /** ElevenLabs categorisation — premade is free for any plan; professional may be PVC/IVC restricted */
+    category?: "premade" | "professional";
+    /** ElevenLabs model_ids this voice has been QA'd against by the team */
+    recommendedModels?: string[];
 }
 /**
- * Curated list of ElevenLabs voices compatible with eleven_multilingual_v3.
- * Any voice can speak any supported language via the language_code parameter.
+ * Curated catalog of ElevenLabs voices we surface to the agent + UI.
+ *
+ * All voice IDs verified against the ElevenLabs public voice library on
+ * 2026-05-16. Metadata (persona, age, useCases) is drawn from the library
+ * card on https://elevenlabs.io/app/voice-library so the agent can pick a
+ * voice that matches the brief without sampling every preview.
+ *
+ * The list mixes premade voices that work on every plan (free + paid) with
+ * a couple of the team's go-to professional voices. Premade entries are
+ * safe defaults; professional entries may require a higher-tier plan and
+ * will surface a clear 401/403 from ElevenLabs at request time if not.
+ *
+ * Any voice can speak any v3 / multilingual_v2 supported language when the
+ * `language_code` parameter is passed — accent is the voice's *native*
+ * delivery, not a hard locale restriction.
  */
 export declare const ELEVENLABS_VOICES: ITtsVoice[];
-/** Default voice ID — Sarah (female, American) */
+/** Default voice ID — Sarah (female, American). Picked for neutrality / safe-default UX. */
 export declare const DEFAULT_ELEVENLABS_VOICE_ID = "EXAVITQu4vr4xnSDxMaL";
+/**
+ * O(1) lookup by voiceId. Built lazily; kept here so the
+ * agent + controllers can validate or hydrate metadata fast.
+ */
+export declare const ELEVENLABS_VOICES_BY_ID: Record<string, ITtsVoice>;
+/** Lowercased name → voice for resolver tools that take "george" / "Sarah" etc. */
+export declare const ELEVENLABS_VOICES_BY_NAME: Record<string, ITtsVoice>;
 /**
  * Maps frontend language keys → ElevenLabs language_code values supported by eleven_multilingual_v3.
  * Omitting a language lets the model auto-detect it (safe fallback).
  * Reference: https://help.elevenlabs.io/hc/en-us/articles/13313366263441
  */
 export declare const LANG_TO_ELEVENLABS_CODE: Record<string, string>;
+export declare const MINIMAX_VOICES: ITtsVoice[];
+export declare const LANG_TO_MINIMAX_CODE: Record<string, string>;
+export declare const OPENAI_VOICES: ITtsVoice[];
 /**
  * ElevenLabs `eleven_turbo_v2_5` supported languages (32 total, per docs:
  * https://elevenlabs.io/docs/models#supported-languages). Used for the

package/lib/globals/ttsModels/voices.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"voices.d.ts","sourceRoot":"","sources":["../../../src/globals/ttsModels/voices.ts"],"names":[],"mappings":"AAAA,MAAM,WAAW,SAAS;IACxB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,MAAM,EAAE,MAAM,GAAG,QAAQ,CAAC;IAC1B,MAAM,EAAE,MAAM,CAAC;IACf,8CAA8C;IAC9C,UAAU,EAAE,MAAM,CAAC;CACpB;AAED;;;GAGG;AACH,eAAO,MAAM,iBAAiB,EAAE,SAAS,EAuFxC,CAAC;AAEF,kDAAkD;AAClD,eAAO,MAAM,2BAA2B,yBAAyB,CAAC;AAElE;;;;GAIG;AACH,eAAO,MAAM,uBAAuB,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAuC1D,CAAC;AAEF;;;;;;;;;;GAUG;AACH,eAAO,MAAM,uBAAuB,EAAE,WAAW,CAAC,MAAM,CAKtD,CAAC;AAEH;;;;GAIG;AACH,wBAAgB,2BAA2B,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,GAAG,MAAM,CAI9E"}
+{"version":3,"file":"voices.d.ts","sourceRoot":"","sources":["../../../src/globals/ttsModels/voices.ts"],"names":[],"mappings":"AAAA,MAAM,WAAW,SAAS;IACxB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,MAAM,EAAE,MAAM,GAAG,QAAQ,CAAC;IAC1B,MAAM,EAAE,MAAM,CAAC;IACf,0FAA0F;IAC1F,UAAU,EAAE,MAAM,CAAC;IACnB,2EAA2E;IAC3E,GAAG,CAAC,EAAE,OAAO,GAAG,aAAa,GAAG,KAAK,CAAC;IACtC,4FAA4F;IAC5F,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,uEAAuE;IACvE,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,kGAAkG;IAClG,QAAQ,CAAC,EAAE,KAAK,CACZ,eAAe,GACf,WAAW,GACX,gBAAgB,GAChB,YAAY,GACZ,WAAW,GACX,MAAM,GACN,cAAc,GACd,yBAAyB,CAC5B,CAAC;IACF,uGAAuG;IACvG,QAAQ,CAAC,EAAE,SAAS,GAAG,cAAc,CAAC;IACtC,wEAAwE;IACxE,iBAAiB,CAAC,EAAE,MAAM,EAAE,CAAC;CAC9B;AAED;;;;;;;;;;;;;;;;GAgBG;AACH,eAAO,MAAM,iBAAiB,EAAE,SAAS,EA+RxC,CAAC;AAEF,4FAA4F;AAC5F,eAAO,MAAM,2BAA2B,yBAAyB,CAAC;AAElE;;;GAGG;AACH,eAAO,MAAM,uBAAuB,EAAE,MAAM,CAAC,MAAM,EAAE,SAAS,CAE7D,CAAC;AAEF,mFAAmF;AACnF,eAAO,MAAM,yBAAyB,EAAE,MAAM,CAAC,MAAM,EAAE,SAAS,CAE/D,CAAC;AAEF;;;;GAIG;AACH,eAAO,MAAM,uBAAuB,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAuC1D,CAAC;AAIF,eAAO,MAAM,cAAc,EAAE,SAAS,EAYrC,CAAC;AAGF,eAAO,MAAM,oBAAoB,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAiBvD,CAAC;AAIF,eAAO,MAAM,aAAa,EAAE,SAAS,EAWpC,CAAC;AAEF;;;;;;;;;;GAUG;AACH,eAAO,MAAM,uBAAuB,EAAE,WAAW,CAAC,MAAM,CAKtD,CAAC;AAEH;;;;GAIG;AACH,wBAAgB,2BAA2B,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,GAAG,MAAM,CAI9E"}