vidgen 0.0.1

package/dist/lib/stt/index.js

@@ -0,0 +1,50 @@
+import { createClient } from '@deepgram/sdk';
+import { deepgramToCombo } from './deepgram-to-combo.js';
+/**
+ * Transcribe audio from a URL using Deepgram
+ *
+ * @param options - Transcription options
+ * @returns Parsed transcription result in Combo format
+ *
+ * @example
+ * ```typescript
+ * const result = await transcribe({
+ *   url: "https://example.com/audio.mp3",
+ *   language: "en"
+ * });
+ * ```
+ */
+export async function transcribe(options) {
+    const { url, apiKey = process.env.DEEPGRAM_API_KEY, language, model = 'nova-3', smartFormat = true, paragraphs = true } = options;
+    if (!url) {
+        throw new Error('Audio URL is required');
+    }
+    if (!apiKey) {
+        throw new Error('Deepgram API key is required');
+    }
+    // Create Deepgram client
+    const deepgram = createClient(apiKey);
+    // Build Deepgram options
+    const deepgramOptions = {
+        model,
+        smart_format: smartFormat,
+        paragraphs,
+        detect_language: true,
+    };
+    // Add language if provided
+    if (language && language !== 'auto') {
+        deepgramOptions.language = language;
+    }
+    // Transcribe audio
+    const { result, error } = await deepgram.listen.prerecorded.transcribeUrl({ url }, deepgramOptions);
+    if (error) {
+        throw new Error(error.message || 'Failed to transcribe audio');
+    }
+    // Convert Deepgram result to Combo format
+    const parsed = await deepgramToCombo(result);
+    return parsed;
+}
+// Export types
+export * from './types.js';
+export { deepgramToCombo } from './deepgram-to-combo.js';
+export { detectLanguage } from './detect-language.js';

package/dist/lib/stt/types.d.ts

@@ -0,0 +1,65 @@
+export type Transcript = {
+    id?: string;
+    userId: string;
+    sourceUrl: string;
+    originalLanguage: string;
+    targetLanguage: string;
+    originalTranscript: any;
+    translatedTranscript: any;
+    status: 'pending' | 'completed' | 'failed';
+    createdAt: Date;
+    updatedAt: Date;
+};
+export interface TranscriptObject {
+    id: string;
+    sourceUrl: string;
+    duration: number;
+    results: {
+        main: {
+            language: LanguageDetectionResultsEntry;
+            text: string;
+            words: Word[];
+            paragraphs: Paragraph[];
+        };
+        translation?: {
+            language: LanguageDetectionResultsEntry;
+            text: string;
+            words: Word[];
+            paragraphs: Paragraph[];
+        };
+    };
+    createdAt: Date;
+}
+export interface Word {
+    word: string;
+    start: number;
+    end: number;
+    confidence: number;
+}
+export type LanguageDetectionResults = LanguageDetectionResultsEntry[];
+export interface LanguageDetectionResultsEntry {
+    language: string;
+    languageName: string;
+    confidence?: number;
+}
+export interface Sentence {
+    text: string;
+    start: number;
+    end: number;
+}
+export interface Paragraph {
+    sentences: Sentence[];
+    numWords: number;
+    start: number;
+    end: number;
+}
+interface WordSegment {
+    word: string;
+    start_time: number;
+    end_time: number;
+}
+export interface OpenAITranscription {
+    transcript: WordSegment[];
+    duration: number;
+}
+export {};

package/dist/lib/stt/types.js

@@ -0,0 +1 @@
+export {};

package/dist/lib/types.d.ts

@@ -0,0 +1,131 @@
+export interface VoiceConfig {
+    id?: string;
+    name?: string;
+    speed?: number;
+    url?: string;
+}
+export interface VisualsConfig {
+    type: string;
+    style: string;
+}
+export interface MusicConfig {
+    id: string;
+    url: string;
+}
+export interface CaptionsConfig {
+    id: string;
+    name: string;
+    position: 'top' | 'middle' | 'bottom';
+    size: 'small' | 'medium' | 'large';
+}
+export interface Clip {
+    id: string;
+    refId?: string;
+    type: string;
+    src?: string[];
+    url?: string;
+    previewSrc?: string;
+    duration?: number;
+    display?: {
+        from: number;
+        to: number;
+    };
+    trim?: {
+        from: number;
+        to: number;
+    };
+    active?: boolean;
+    prompt?: string;
+    status?: 'generating' | 'completed' | 'failed';
+    taskId?: string;
+    error?: string;
+    createdAt?: number;
+    isVisualOnly?: boolean;
+    isCloned?: boolean;
+}
+export interface TextToSpeech {
+    refId: string;
+    src: string;
+    duration: number;
+}
+export interface VisualShot {
+    id?: string;
+    type: 'product' | 'generic' | 'lifestyle' | 'medical_cgi' | 'metaphor' | 'b-roll';
+    category: string;
+    firstFrame: string;
+    videoPrompt: string;
+    scenePrompt: string;
+    words?: string;
+    segmentId?: string;
+    prompt?: string;
+    triggerWords?: string;
+    video?: string;
+    lastFrame?: string;
+    hasProductInteraction?: boolean;
+    productSizing?: string;
+}
+export interface VisualBroll {
+    time?: number;
+    url?: string;
+    duration?: number;
+    type: 'video' | 'image';
+    firstFrame?: string;
+    videoPrompt?: string;
+    scenePrompt?: string;
+    words?: string;
+    productSizing?: string;
+}
+export interface SpeechToText {
+    refId: string;
+    src: string;
+}
+export interface Segment {
+    id: string;
+    title: string;
+    text: string;
+    description: string;
+    searchQuery: string;
+    tags: string[];
+    shots?: VisualShot[];
+    bRolls?: VisualBroll[];
+    clips: Clip[];
+    duration: number;
+    assets?: Clip[];
+    textToSpeech?: TextToSpeech;
+    speechToText?: SpeechToText;
+    mergeWithNext?: boolean;
+    estimatedDuration?: number;
+    isContinuation?: boolean;
+}
+export interface Schema {
+    id?: string;
+    title?: string;
+    description?: string;
+    tags?: string[];
+    segments?: Segment[];
+    voice: VoiceConfig;
+    visuals: VisualsConfig;
+    music?: MusicConfig;
+    caption: CaptionsConfig;
+    aspectRatio: '1:1' | '16:9' | '9:16' | '11';
+    product?: {
+        name?: string;
+        description?: string;
+    };
+    assets?: {
+        id: string;
+        name: string;
+        url: string;
+        type: 'image' | 'video';
+    }[];
+    avatar?: {
+        id: string;
+        name: string;
+        url: string;
+    };
+    type?: 'narrative-video' | 'product-video-ad' | 'ugc-video-ad' | string;
+    script?: string;
+    pacing?: 'fast' | 'slow' | 'regular' | 'dynamic' | 'relaxed';
+    quality?: 'regular' | 'high';
+    duration?: number;
+}

package/dist/lib/types.js

@@ -0,0 +1 @@
+export {};