1.2.0

Added

• GPT Image 1.5 - OpenAI's image generation model with transparent background support, quality control, and masked API key input
• Nano Banana Pro - Google's Gemini 3 Pro model with 4K resolution, text rendering, and multi-image fusion
• Qwen Image - Qwen's generation model with guidance control, img2img support, and prompt enhancement

Changed

• Flux - Upgraded from 1.1 Pro to 2 Pro with improved generation and editing capabilities
• Nano Banana - Updated to full release (Gemini 2.5 Flash Image), removed "Preview" designation
• Seedream - Upgraded to 4.5 with stronger spatial understanding and 4K resolution presets
• Renamed model files for consistency (imagen4→imagen, seedream4→seedream, etc.)

1.1.5

[1.1.5] - 2025-11-06

Added

• New Model: Flux 1.1 Pro - Added support for Black Forest Labs' Flux 1.1 Pro model
  • 6x faster generation than predecessor
  • Excellent image quality with superior prompt adherence
  • Support for one reference image for image-to-image generation
  • Custom dimensions up to 1440px with multiple aspect ratio presets
  • Configurable safety tolerance (1-6 scale)
  • Configurable parameters including seed, prompt upsampling, and output quality
  • Supports PNG, JPEG, WebP, and GIF formats

Changed

• Improved Seed Handling - Standardized random seed convention across all models
  • Changed from 0 = random to -1 = random for consistency
  • Seeds are now only included in API calls when explicitly set (not -1)
  • Removed random.randint() at module load time for cleaner architecture
  • Applied to: Flux 1.1 Pro, Qwen Image Edit Plus, and Stable Diffusion 3.5 Large Turbo

Fixed

• Reference Image Limit Display - Fixed bug where dialog always showed "10 reference images" on initial load
  • Now correctly displays the maximum reference images for the selected model
  • Ensures self.model is synchronized with the dropdown selection on dialog initialization
• Code Quality - Comprehensive code formatting improvements
  • Standardized to 80 character line length (updated from 79)
  • Consistent double-quote usage throughout codebase
  • Added trailing commas for better version control diffs
  • Improved line wrapping and import formatting
  • Added blank lines after imports for better readability
  • Fixed all flake8 linting issues

Translations

• Updated all 10 language translations to include new Flux 1.1 Pro strings
  • All translations now 100% complete (135 messages each)
  • Fixed fuzzy translations from seed convention changes
  • Updated version numbers in all .po files
  • Languages: Bengali, Spanish, French, Hindi, Japanese, Korean, Portuguese, Russian, Chinese (Simplified), Chinese (Traditional)

Documentation

• Updated README with Flux 1.1 Pro information
• Added Flux 1.1 Pro to available models list with full specifications
• Updated installation examples to reflect version 1.1.5
• Updated credits to include Black Forest Labs

Technical

• Added Ruff configuration for modern Python linting
• Improved code consistency across all model implementations
• Better API call efficiency (conditional parameter inclusion)
• Enhanced model selection initialization for more reliable UI state

1.1.4

• Improved error handling

1.1.3

• Added Stable Diffusion 3.5 Large Turbo
• Minor code cleanups
• Updated translations

1.1.2

• Added Imagen 4 and Qwen Image Edit Plus models
• Updated translations
• Code cleanup to allow for generate/edit only models

1.1.1

• Update to two column layout
• Misc code cleanup

1.1.0

• Updated API to use Replicate instead of Google
• Now supports both Nano Banana and Seedream 4
• Updated to support future models, with full model factory
• Dynamic settings per model
• Significant code cleanup

1.0.3

• Simplified UI
• Extensive code cleanup and refactoring
• Updated translations

1.0.2

• Resize edit layers to match image size

1.0.1

• Allow editing a selection within an image
• Add alpha channel support to provide transparent layer backgrounds
• Code cleanups and new stubs to make development more enjoyable