GLabs SDK

TypeScript SDK and CLI for Google Labs AI media generation — Imagen 4 image generation and Veo 3.1 video generation.

Packages

| Package | Description | Install | |---------|-------------|---------| | @getvrex/glabs-sdk | SDK for programmatic usage | npm install @getvrex/glabs-sdk | | @getvrex/glabs-cli | Global CLI | npm install -g @getvrex/glabs-cli |

CLI

npm install -g @getvrex/glabs-cli

glabs <command> [subcommand] [options]

Commands:
  auth      extract             Extract tokens via browser automation
  config    show | set          Manage CLI configuration
  projects  list | get          Manage projects
  images    generate | upload | upsample | credits
  videos    generate | i2v | r2v | extend | reshoot | upsample | status | poll
  whisk     generate            Whisk image generation (Imagen 3.5)
  serve                         Start OpenAI-compatible server

Setup

# Extract tokens automatically
glabs auth extract --email [email protected] --password "pass" --save

# Or set tokens manually
glabs config set --bearer-token "tok" --session-token "stok" --account-tier pro

Usage Examples

# Generate images
glabs images generate -p "A sunset over mountains" -a 16:9 -n 4

# Generate video from text
glabs videos generate -p "A cinematic drone shot of a city" -a 16:9

# Generate video from image
glabs videos i2v -p "Camera pans slowly" --start-media-id "abc123"

# Poll and download video
glabs videos poll --operation-name "operations/xyz" -o ./videos

# Start OpenAI-compatible server
glabs serve --port 8000 --api-key "sk-my-key"

SDK

npm install @getvrex/glabs-sdk

Quick Start

import { GLabsClient } from '@getvrex/glabs-sdk';

const client = new GLabsClient({
  bearerToken: 'your-bearer-token',
  sessionToken: 'your-session-token',
  accountTier: 'pro',
  recaptcha: { provider: 'chrome' },
});

// Generate an image
const image = await client.images.generate({
  prompt: 'A beautiful sunset over mountains',
  sessionId: GLabsClient.generateSessionId(),
  aspectRatio: '16:9',
});

// Generate a video
const operation = await client.videos.generateTextToVideo({
  prompt: 'A cinematic drone shot of a city',
  sessionId: GLabsClient.generateSessionId(),
  aspectRatio: '16:9',
});

// Poll until ready
const video = await client.videos.pollOperation({
  operationName: operation.operationName,
  onProgress: (status, attempt) => console.log(`[${attempt}] ${status.status}`),
});

console.log('Video URL:', video.videoUrl);
await client.close();

Image Generation

Models: nanobanana2 (default, NARWHAL), nanobananapro (GEM_PIX_2), imagen-4 / imagen-4-fast / imagen-4-ultra (IMAGEN_3_5)

client.images.generate(opts)       // Text-to-image (up to 4 per batch)
client.images.upload(opts)         // Upload for video generation
client.images.upsampleImage(opts)  // Upscale to 2K/4K
client.images.getCreditStatus()    // Check account credits

Video Generation (Veo 3.1)

client.videos.generateTextToVideo(opts)          // Text-to-video
client.videos.generateImageToVideo(opts)          // Image-to-video (start or first+last frame)
client.videos.generateReferenceImagesVideo(opts)  // Multi-reference image video (1-3 images)
client.videos.extend(opts)                        // Extend existing videos
client.videos.reshoot(opts)                       // Camera control reshoot (14 motion types)
client.videos.upsample(opts)                      // Upscale to HD/4K
client.videos.checkStatus(opts)                   // Check generation status
client.videos.pollOperation(opts)                 // Poll until completion

Project Management

client.projects.list()              // List projects
client.projects.get({ projectId })  // Get project details
client.projects.getFirstProjectId() // Auto-resolve (cached)

All generation methods auto-select the first project if none is provided.

OpenAI-Compatible Server

import { GLabsClient } from '@getvrex/glabs-sdk';
import { OpenAIServer } from '@getvrex/glabs-sdk/openai';

const server = new OpenAIServer(new GLabsClient({ ... }), {
  port: 8000,
  apiKey: 'sk-xxx',
});
await server.start();
// POST /v1/chat/completions
// GET  /v1/models

Whisk (Imagen 3.5)

import { WhiskService } from '@getvrex/glabs-sdk';

const whisk = new WhiskService('your-cookie-string');
const result = await whisk.generateImage('A cute robot');

Configuration

const client = new GLabsClient({
  bearerToken: string,           // Required: auth token
  sessionToken?: string,         // Auto token refresh (ST -> AT)
  accountTier?: 'pro' | 'ultra', // Default: 'pro'
  projectId?: string,            // Auto-resolved if omitted
  recaptcha?: RecaptchaConfig,   // Required for generation
  timeout?: number,              // Request timeout ms (default: 120000)
  maxRetries?: number,           // Network retries (default: 2)
  retryDelay?: number,           // Retry delay ms (default: 1500)
});

reCAPTCHA Providers

| Provider | Type | Note | |----------|------|------| | chrome | Browser | Recommended — real Chrome, highest scores | | yescaptcha | Cloud | Recommended — no local browser needed | | playwright | Browser | Playwright-managed browser | | regotcha | Cloud | Optimized for Google Labs | | capsolver | Cloud | Proxy support | | veo3solver | Token | Pre-solved tokens via JWT | | custom | Self-hosted | Your own solver endpoint |

Fallback chains:

recaptcha: {
  provider: 'chrome',
  fallback: { provider: 'yescaptcha', apiKey: 'key' },
}

Token Management

With sessionToken configured, the SDK automatically refreshes the bearer token before expiry, retries on 401, and deduplicates concurrent refresh calls.

Account Tiers

| Feature | Pro | Ultra | |---------|-----|-------| | Default Video Mode | fast | quality | | HD/4K Upscaling | Yes | Yes | | Max Images/Batch | 4 | 4 |

Types

import type {
  AccountTier, AspectRatio, GenerateImageOptions,
  GenerateTextToVideoOptions, VideoStatusResult,
  RecaptchaConfig, Project, OpenAIServerConfig,
} from '@getvrex/glabs-sdk/types';

Error Handling

import { GLabsError } from '@getvrex/glabs-sdk';

try {
  await client.images.generate({ ... });
} catch (error) {
  if (error instanceof GLabsError) {
    console.error(`[${error.code}] ${error.message}`, error.statusCode);
  }
}

Claude Code Skill

Install the glabs-cli skill for Claude Code AI assistance:

curl -fsSL https://raw.githubusercontent.com/getvrex/glabs-sdk/main/scripts/install-skill.sh | bash

Documentation

Full docs: docs/

| Guide | Description | |-------|-------------| | Getting Started | Installation and setup | | CLI | Command-line interface | | Client | Client configuration | | Image Generation | Image API | | Video Generation | Video API | | Project Management | Project API | | reCAPTCHA | reCAPTCHA integration | | OpenAI Server | OpenAI-compatible server | | Token Management | Auto token refresh | | Whisk | Whisk image generation | | Error Handling | Error codes | | API Reference | Complete API reference |

License

MIT