rubyllm-image-generation

name	rubyllm-image-generation
description	Generate images from text with RubyLLM. Use this skill for creating images with DALL-E 3, GPT Image 1, Imagen, editing existing images, and working with masks for constrained editing.

RubyLLM Image Generation

Create images from text descriptions using AI models like DALL-E 3 and Imagen.

Basic Usage

# Generate image
image = RubyLLM.paint("A photorealistic red panda coding Ruby")

# Access result
if image.url
  puts "URL: #{image.url}"
  image.save("red_panda.png")  # Save locally
end

if image.base64?
  puts "MIME: #{image.mime_type}"
  File.write("image.png", Base64.decode64(image.data))
end

# Revised prompt (some models optimize your prompt)
puts image.revised_prompt if image.revised_prompt

Models

Model	Provider	Size	Price/Image
dall-e-3	OpenAI	1024x1024	$0.040
dall-e-3	OpenAI	1024x1792	$0.080
gpt-image-1	OpenAI	Various	Supports editing
imagen-3	Google	Various	Base64 output

# DALL-E 3
RubyLLM.paint("Sunset", model: 'dall-e-3')

# GPT Image 1 (supports editing)
RubyLLM.paint("Edit logo", model: 'gpt-image-1', with: 'logo.png')

Image Editing (v1.15+)

Edit existing images:

# Basic editing
image = RubyLLM.paint(
  "Turn the logo green",
  model: 'gpt-image-1',
  with: 'logo.png'
)

# Edit with mask (constrain which parts change)
image = RubyLLM.paint(
  "Change background to blue",
  model: 'gpt-image-1',
  with: 'photo.jpg',
  mask: 'mask.png'  # White areas get edited
)

Options

# Provider-specific params
chat = RubyLLM.chat.with_params(
  size: '1024x1024',
  quality: 'hd',
  style: 'vivid'
)

image = RubyLLM.paint("Sunset", **chat.params)

Active Storage Integration

class GeneratedImage < ApplicationRecord
  has_one_attached :file
end

image = RubyLLM.paint("Sunset")
record = GeneratedImage.create!

if image.base64?
  record.file.attach(
    io: StringIO.open(Base64.decode64(image.data)),
    filename: "sunset.png",
    content_type: image.mime_type
  )
elsif image.url
  # Download and attach
  response = Faraday.get(image.url)
  record.file.attach(
    io: StringIO.new(response.body),
    filename: "sunset.png",
    content_type: response.headers['content-type']
  )
end

Prompt Tips

Be Specific

# Vague
RubyLLM.paint("A dog")

# Specific
RubyLLM.paint("A golden retriever puppy sitting in a sunny meadow, photorealistic, warm lighting")

Style Keywords

# Artistic styles
RubyLLM.paint("City skyline, watercolor style, soft pastels")
RubyLLM.paint("Portrait, oil painting, renaissance style")
RubyLLM.paint("Abstract composition, cubist, bold colors")

# Photography styles
RubyLLM.paint("Product photo, studio lighting, white background")
RubyLLM.paint("Landscape, golden hour, dramatic clouds")

See Also

• Main RubyLLM: rubyllm
• Multi-Modal: See rubyllm for using images in chat