Imported from jeremylongshore/tons-of-skills-marketplace (skills/.curated/speak-install-auth/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-install-auth. Copyright stays with the author (MIT).

Speak Install & Auth

Overview

Set up the Speak language learning platform integration. Speak uses OpenAI's GPT-4o and Realtime API for AI tutoring with real-time pronunciation feedback. Supports 14+ languages including Korean, Spanish, Japanese, French, and Mandarin.

Prerequisites

• Node.js 18+ or Python 3.10+
• Speak developer account with API access
• Microphone for speech input testing

Instructions

Step 1: Install Dependencies

set -euo pipefail
# Core Speak SDK
npm install @speak/language-sdk

# Audio processing dependencies
npm install openai          # OpenAI Realtime API for speech
npm install fluent-ffmpeg   # Audio format conversion
npm install node-record-lpcm16  # Microphone capture

Step 2: Configure Authentication

# Speak API credentials
export SPEAK_API_KEY="your-speak-api-key"
export SPEAK_APP_ID="your-app-id"

# OpenAI key for Realtime API (used by Speak for speech processing)
export OPENAI_API_KEY="your-openai-key"

# Create .env file
cat << 'EOF' >> .env
SPEAK_API_KEY=your-speak-api-key
SPEAK_APP_ID=your-app-id
OPENAI_API_KEY=your-openai-key
EOF

Step 3: Initialize the Client

// src/speak/client.ts
import { SpeakClient } from '@speak/language-sdk';

const client = new SpeakClient({
  apiKey: process.env.SPEAK_API_KEY!,
  appId: process.env.SPEAK_APP_ID!,
  language: 'es', // Target language: es, ko, ja, fr, de, pt, zh, id
});

// Verify connection
async function verifySetup() {
  const languages = await client.getLanguages();
  console.log('Available languages:', languages.map(l => l.code).join(', '));
  const health = await client.health.check();
  console.log('API status:', health.status);
}

verifySetup();

Step 4: Configure Speech Recognition

// OpenAI Realtime API for speech-to-text (used by Speak)
import OpenAI from 'openai';

const openai = new OpenAI();

async function transcribeAudio(audioPath: string): Promise<string> {
  const transcription = await openai.audio.transcriptions.create({
    file: fs.createReadStream(audioPath),
    model: 'whisper-1',
    language: 'es', // Match target language
  });
  return transcription.text;
}

Step 5: Supported Languages

Output

• Speak SDK installed and configured
• API key and OpenAI credentials set
• Language support verified
• Speech recognition pipeline ready

Error Handling

Resources

• Speak Website
• OpenAI Realtime API
• OpenAI Whisper
• Speak Blog: GPT-4 Integration

Next Steps

After successful auth, proceed to speak-hello-world for your first lesson session.

Examples

Quick test: Set SPEAK_API_KEY, initialize the client with language: 'ko' for Korean, and call client.health.check() to verify connectivity.

Python setup: Install speak-language-sdk via pip, initialize with api_key from environment, and verify with a health check.