Skip to content

replit.md

Latest commit

 

History

History
116 lines (104 loc) · 5.94 KB

replit.md

File metadata and controls

116 lines (104 loc) · 5.94 KB

PodcastGen - AI Podcast Generator

Overview

PodcastGen is a mobile app that creates podcasts from search queries. Users enter a topic they want to learn about, and the app uses AI (OpenAI GPT-4 and TTS) to generate engaging podcast content. Features a conversational AI interface that asks clarifying questions before generating, and supports both single focused episodes and multi-episode series for broader topics.

Tech Stack

  • Framework: Expo (React Native)
  • Navigation: React Navigation 7 with bottom tabs
  • Storage: AsyncStorage for local data persistence
  • Audio: expo-av for audio playback
  • AI Integration: OpenAI API (GPT-4 for script generation, TTS for audio)
  • Authentication: expo-apple-authentication for Apple Sign-In (iOS)

Project Structure

├── App.tsx                 # Root component with navigation + AuthProvider + AudioPlayerProvider + MiniPlayer
├── contexts/
│   ├── AudioPlayerContext.tsx  # Global audio playback state management
│   └── AuthContext.tsx         # Authentication state with Apple Sign-In
├── navigation/
│   ├── MainTabNavigator.tsx    # 4-tab navigation (Library, Create, Play, Profile)
│   ├── LibraryStackNavigator.tsx
│   ├── CreateStackNavigator.tsx
│   ├── PlayStackNavigator.tsx  # New - Play tab navigator
│   └── ProfileStackNavigator.tsx
├── screens/
│   ├── CreateScreen.tsx        # Podcast creation entry point
│   ├── ChatCreationScreen.tsx  # Conversational AI for refining topic
│   ├── GeneratingScreen.tsx    # Progress modal during generation
│   ├── LibraryScreen.tsx       # Saved podcasts and series list
│   ├── PlayerScreen.tsx        # Legacy player (used within stacks)
│   ├── PlayScreen.tsx          # New - Main player in Play tab
│   ├── SeriesDetailScreen.tsx  # Series episodes viewer
│   └── ProfileScreen.tsx       # User settings
├── components/
│   ├── AnimatedWaveform.tsx    # Animated audio visualization
│   ├── WaveformPreview.tsx     # Static waveform for library cards
│   ├── MiniPlayer.tsx          # Persistent play bar at bottom
│   ├── LoginPrompt.tsx         # Authentication modal for sign-in
│   ├── Button.tsx              # Primary action button
│   ├── Card.tsx                # Elevated card component
│   └── ...
├── utils/
│   ├── storage.ts              # AsyncStorage utilities (podcasts + series)
│   ├── podcastGenerator.ts     # OpenAI integration + series generation
│   └── conversationFlow.ts     # Chat state machine + AI follow-ups
└── constants/
    └── theme.ts                # Design system colors/spacing

Key Features

  1. Authentication: Apple Sign-In (iOS), Google Sign-In, and Email/Password (all platforms)
  2. Conversational AI Interface: Chat-based creation with follow-up questions to refine topic scope, depth, and style
  3. Single Episode Mode: Create a focused episode on a specific topic
  4. Series Mode: Generate multi-episode series (3-5 episodes) for broader topics
  5. Episode Plan Approval: For series, review and approve episode titles before generation
  6. AI Script Generation: Uses GPT-4 to create engaging podcast scripts
  7. Text-to-Speech: Converts scripts to natural audio using OpenAI TTS
  8. Audio Player: Full playback with time-synced scrolling lyrics
  9. Source Transparency: "See Sources" modal shows research citations
  10. Library Management: Separate views for series and standalone episodes
  11. Voice Selection: 6 narrator options (onyx, alloy, echo, fable, nova, shimmer)
  12. Favorites: Mark podcasts and series as favorites

Data Model

Podcast (single episode or series episode)

  • id, topic, script, audioUri, duration, createdAt, sources
  • seriesId (if part of series), episodeNumber, episodeTitle
  • isFavorite, voiceUsed, category, style, depth

PodcastSeries

  • id, topic, description, episodeCount, totalDuration
  • createdAt, coverColor, isFavorite, style, depth

Environment Variables

  • EXPO_PUBLIC_OPENAI_API_KEY: Required for AI features (script + audio generation)

Test Mode

Set TEST_MODE = true in utils/podcastGenerator.ts to avoid API calls during development. Uses hardcoded scripts and dummy audio.

Design System

  • Primary Color: #6B4CE6 (Deep purple)
  • Secondary Color: #FF6B6B (Coral)
  • Style: iOS 26 liquid glass-inspired with clean, modern aesthetics

Running the App

npm run dev

Recent Changes

  • Enhanced authentication with multiple sign-in options:
    • Apple Sign-In (iOS only)
    • Google Sign-In (all platforms)
    • Email/Password sign-up and sign-in (all platforms)
  • Login prompt redesigned with sign-in/sign-up toggle for email auth
  • Profile screen shows auth method used (Apple, Google, Email)
  • AuthContext manages global auth state with AsyncStorage persistence
  • Email accounts stored locally with basic validation
  • Added "Play" tab with persistent mini player bar across all screens (like Spotify)
  • Mini player hides on Play screen, reappears when pressing play
  • AudioPlayerContext provides global playback state management
  • After podcast generation, automatically navigates to Play tab and starts playing
  • Library and Series screens now use global audio player for seamless playback
  • Unified conversation flow for both single episodes and series
  • AI asks 3 questions: format preference, depth, and style
  • Choosing "Single episode" creates one focused episode
  • Choosing "Multi-part series" triggers series flow with episode plan approval
  • Quick-reply chips for fast topic refinement

Notes

  • Email/password authentication stores accounts locally for demo purposes
  • For production, integrate with a backend auth service like Firebase
  • Google Sign-In requires OAuth client IDs from Google Cloud Console