skill-image-gen

"Imagine it. Generate it. Everywhere." Powered by Chitra Engine (चित्र — Sanskrit for picture/image) By Darshj.me

Description

Generate stunning images from text prompts using a unified interface across DALL-E 3, Stability AI, Replicate (Flux), Fal.ai, and Google Imagen. Automatically selects the best provider based on your quality/speed/cost preference.

Core capabilities:

• 🎨 5 providers: OpenAI, Stability AI, Replicate, Fal.ai, Google Imagen
• ✨ 12 style presets: photorealistic, anime, oil-painting, cinematic, cyberpunk, and more
• 🔁 Intelligent provider auto-selection (quality / speed / cost mode)
• 💡 Automatic prompt enhancement — turns short descriptions into rich prompts
• ⚡ Result caching — no re-generation for identical prompts
• 🔄 Automatic fallback on provider failure
• 📐 Aspect ratio normalization across all providers
• 🔢 Parallel batch generation

Setup

Set at least one of the following environment variables:

# OpenAI DALL-E 3 (best quality)
OPENAI_API_KEY=sk-...

# Stability AI (strong prompt adherence)
STABILITY_API_KEY=sk-...

# Replicate (Flux, SDXL, open-source models)
REPLICATE_API_TOKEN=r8_...

# Fal.ai (fastest, cheapest Flux access)
FAL_KEY=...

# Google Imagen via Vertex AI (photorealistic)
GOOGLE_PROJECT_ID=my-gcp-project
GOOGLE_ACCESS_TOKEN=ya29...
GOOGLE_LOCATION=us-central1    # optional, default: us-central1

Optional configuration:

CHITRA_VERBOSE=1    # Enable debug logging

Usage

As an OpenClaw Skill

The agent can invoke this skill by importing and using the chitra engine:

import { chitra } from 'skill-image-gen';

// Basic generation — auto-selects best provider
const result = await chitra.generate('a serene Japanese garden at dawn');
console.log(result.images[0].url);

// With style and quality options
const artResult = await chitra.generate('a lone samurai under cherry blossoms', {
  style: 'anime',
  quality: 'hd',
  aspectRatio: '16:9',
  preferMode: 'quality',
});

// Batch generation — multiple prompts in parallel
const batch = await chitra.batchGenerate([
  'a dragon in the mountains',
  'a futuristic city at night',
  'an ocean at sunset',
], { style: 'cinematic', quality: 'standard' });

Quick Reference

// Import
import { chitra, Chitra, STYLE_PRESETS } from 'skill-image-gen';

// Generate with all options
const result = await chitra.generate(prompt, {
  provider: 'openai',        // 'openai' | 'stability' | 'replicate' | 'fal' | 'google'
  count: 1,                  // 1–10 images
  aspectRatio: '1:1',        // '1:1' | '16:9' | '9:16' | '4:3' | '3:4' | '3:2' | '2:3' | '21:9'
  style: 'photorealistic',   // see Style Presets below
  quality: 'hd',             // 'draft' | 'standard' | 'hd' | 'ultra'
  preferMode: 'quality',     // 'quality' | 'speed' | 'cost' (for auto-selection)
  enhancePrompt: true,       // Auto-expand short prompts
  negativePrompt: 'blurry',  // Elements to exclude
  seed: 42,                  // For reproducibility
  cache: true,               // Cache results
  retryOnFailure: true,      // Try alternate provider on error
  model: 'dall-e-3',         // Model override within provider
});

// Result shape
result.images[0].url          // Public image URL
result.images[0].provider     // Which provider generated it
result.images[0].model        // Which model was used
result.images[0].fromCache    // Was this a cache hit?
result.images[0].latencyMs    // Generation time
result.enhancedPrompt         // The prompt actually sent to the provider
result.totalLatencyMs         // Total wall time

Style Presets

| Preset | Description | Best For | |--------|-------------|----------| | photorealistic | DSLR photography quality | Product shots, portraits | | anime | Japanese animation style | Characters, fantasy | | oil-painting | Classical canvas painting | Art, portraits | | digital-art | Modern digital illustration | Concept art, characters | | sketch | Pencil/charcoal drawing | Quick drafts, thumbnails | | cinematic | Movie-quality cinematography | Scenes, storytelling | | watercolor | Soft watercolor painting | Nature, dreamy scenes | | pixel-art | Retro 8/16-bit game art | Games, nostalgia | | fantasy | Epic fantasy illustration | Worlds, creatures | | minimalist | Clean, negative-space design | Logos, icons | | cyberpunk | Neon dystopian aesthetic | Sci-fi, urban | | vintage | Retro nostalgic film | Memories, retro vibes |

Provider Selection Guide

| Provider | Quality | Speed | Cost | Best For | |----------|---------|-------|------|----------| | OpenAI DALL-E 3 | ⭐⭐⭐⭐⭐ | Medium | $$$$ | Best quality, presentations | | Google Imagen | ⭐⭐⭐⭐⭐ | Medium | $$$ | Photorealism, text in image | | Stability AI | ⭐⭐⭐⭐ | Fast | $$ | Creative, prompt adherence | | Replicate Flux | ⭐⭐⭐⭐ | Medium | $$ | Open-source, custom models | | Fal.ai Flux | ⭐⭐⭐⭐ | ⚡ Fast | $ | Prototyping, high volume |

Agent Invocation Pattern

When an agent receives an image generation request, use this pattern:

import { chitra } from 'skill-image-gen';

// Map user intent to options
async function generateForUser(userRequest: string) {
  // Extract intent keywords
  const wantsHighQuality = /high quality|detailed|best|hd/i.test(userRequest);
  const wantsFast = /quick|fast|draft|preview/i.test(userRequest);

  const result = await chitra.generate(userRequest, {
    quality: wantsHighQuality ? 'hd' : 'standard',
    preferMode: wantsFast ? 'speed' : 'quality',
    enhancePrompt: true,
    retryOnFailure: true,
  });

  return {
    imageUrl: result.images[0]?.url,
    provider: result.images[0]?.provider,
    prompt: result.enhancedPrompt,
  };
}

Chitra Engine v0.1.0 — By Darshj.me "Imagine it. Generate it. Everywhere."