Back to Generator

Technology Stack

All the tools, models, and services powering this application.

🧠Language Models

GPT-5.2

OpenAIgpt-5.2

Flagship LLM for high-quality script generation and content analysis.

Used for:

  • Generating pitch scripts (linear & branched)
  • Creating email content
  • Landing page copy generation
  • Parsing transcripts into sections
Docs →OPENAI_API_KEY

GPT-4.1

OpenAIgpt-4.1

Fast, capable model for Scout conversations with function calling.

Used for:

  • Scout AI assistant responses (traditional mode)
  • Resource suggestion tool calls
Docs →OPENAI_API_KEY

GPT-4.1 Mini

OpenAIgpt-4.1-mini

Faster, lighter model for low-latency Scout responses.

Used for:

  • Scout conversations (when SCOUT_FAST_MODE=true)
Docs →SCOUT_FAST_MODE

🎤Speech Recognition

Whisper

OpenAIwhisper-1

Automatic speech recognition for transcribing user voice input.

Used for:

  • Transcribing Scout voice questions
  • Converting user audio to text for processing
Docs →OPENAI_API_KEY

Web Speech API

Browser

Browser-native speech recognition for real-time voice commands.

Used for:

  • Voice commands during branching video choice points
  • Real-time transcript display

🔊Text-to-Speech

ElevenLabs TTS

ElevenLabseleven_multilingual_v2

High-quality text-to-speech for pitch video narration.

Used for:

  • Generating pitch audio (linear videos)
  • Generating segment audio (branching videos)
Docs →ELEVENLABS_API_KEY

ElevenLabs Turbo

ElevenLabseleven_turbo_v2_5

Low-latency TTS for real-time conversational responses.

Used for:

  • Scout voice responses (optimized for speed)
Docs →ELEVENLABS_API_KEY

ElevenLabs Sales Agent

ElevenLabs

Conversational AI agent for generating personalized sales pitch videos.

Used for:

  • Voice First pitch generation (improvised sales monologue)
Docs →ELEVENLABS_AGENT_ID

ElevenLabs Scout Agent

ElevenLabs

Conversational AI agent for landing page Q&A with knowledge base and resource suggestions.

Used for:

  • Scout Agent mode (WebSocket - voice-to-voice)
  • Scout WebRTC mode (best echo cancellation & noise removal)
  • Knowledge base RAG for accurate answers
  • Resource suggestions via client tools
Docs →ELEVENLABS_SCOUT_AGENT_ID

🎬Video Generation

Creatomate

Creatomate

Programmatic video rendering with audio sync and Ken Burns effects.

Used for:

  • Generating MP4 pitch videos
  • Creating animated GIFs for email
  • Branching video segment generation
Docs →CREATOMATE_API_KEY

💾Storage

Vercel Blob

Vercel

Edge-optimized object storage for media and configuration.

Used for:

  • Storing generated audio files
  • Storing landing page configurations
  • Public URL generation for Creatomate
Docs →BLOB_READ_WRITE_TOKEN

🗣️Voices

Rachel

ElevenLabs21m00Tcm4TlvDq8ikWAM

Professional female voice.

Used for:

  • Pitch narration option

Adam

ElevenLabspNInz6obpgDQGcFmaJgB

Professional male voice.

Used for:

  • Pitch narration option

Antoni

ElevenLabsErXwobaYiN019PkySvjV

Friendly male voice.

Used for:

  • Pitch narration option
  • Scout AI assistant

Bella

ElevenLabsEXAVITQu4vr4xnSDxMaL

Warm female voice.

Used for:

  • Pitch narration option

🏗️Framework & Infrastructure

Next.js 16

Vercel

React framework with App Router for full-stack development.

Used for:

  • Server-side rendering
  • API routes
  • Static page generation

Tailwind CSS

Tailwind Labs

Utility-first CSS framework for styling.

Used for:

  • All UI styling

TypeScript

Microsoft

Type-safe JavaScript for better developer experience.

Used for:

  • All application code

🔑Required Environment Variables

Add these to your .env.local file:

OPENAI_API_KEY=your_openai_api_key_here
SCOUT_FAST_MODE=your_scout_fast_mode_here
ELEVENLABS_API_KEY=your_elevenlabs_api_key_here
ELEVENLABS_AGENT_ID=your_elevenlabs_agent_id_here
ELEVENLABS_SCOUT_AGENT_ID=your_elevenlabs_scout_agent_id_here
CREATOMATE_API_KEY=your_creatomate_api_key_here
BLOB_READ_WRITE_TOKEN=your_blob_read_write_token_here

🔄Generation Flows

Script FirstDefault mode

AI writes the script first, then text-to-speech reads it aloud.

1

Script Generation

GPT-4o writes pitch script with sections

GPT-4o
2

Text-to-Speech

ElevenLabs converts script to audio

ElevenLabs TTS
3

Video Render

Combine audio + images

Creatomate
Voice FirstDirect mode

AI agent improvises naturally, then we transcribe and structure.

1

Agent Prompt

GPT-4o creates natural prompt for agent

GPT-4o
2

Improvised Speech

Sales agent delivers personalized pitch

Sales Agent
3

Parse & Structure

GPT-4o sections transcript, writes email

GPT-4o
4

Video Render

Combine audio + images

Creatomate
Branching VideoInteractive mode

Generates 4 video segments (intro, revenue path, retention path, outro) for choose-your-path experience.

GPT-4o4 scripts
ElevenLabs4 audio files
Creatomate4 videos (parallel)
Web Speech APIVoice commands
Landing Page & ScoutCommon to all modes

Landing Page Generation

GPT-4oPage content
Vercel BlobStore config

Scout Conversation

Traditional3-step pipeline
WhisperTranscribe
GPT-4oRespond
ElevenLabs TurboSpeak
Agentlower latency, knowledge base
Scout AgentVoice + RAG + Tools