What's New

Ideogram 4 Support: Added support for the Ideogram 4 model.

Prompt Composer: Updated the Prompt Composer to allow visual prompt composition. It works best with the Ideogram 4 model with JSON output, but also works with other models. Learn More.

Size Presets: Now reorderable to keep your most-used options within easy reach.

Lens-Turbo: Updated download link of Lens-Turbo model. Now points to the correct 6-bit quanlitzed version.

Studio Assistant: Added Lens model knowledge for Studio Assistant for better prompt suggestion.

Cloud Model: Updated dimension presets for some cloud models.

Image-to-Image for Lens Model: Added image to image mode for Lens/Lens-Turbo Models.

Prompt History: Added prompt history for Inpainting/Outpainting modes

Smart Suggestion: Fixed a bug where smart suggestion text might be cutoff when streaming.

Lens Models Support: Support for MicroSoft's Lens and Lens-Turbo models.

HEIC/HEIF Support: Support importing HEIC/HEIF images.

Memory Optimization: Optimized memory usage during model loading.

ERNIE-Image Support: Support for ERNIE-Image and ERNIE-Image-Turbo models.

Studio Assistant Enhancement: Use the Assistant directly within the Canvas workspace, with support for external providers (e.g., Ollama, LM Studio).

Edit and resend chat messages to refine instructions.

Describe canvas images and save descriptions as notes.

Model-Aware Prompt Enhancement: Studio Assistant and Prompt Enhancer now automatically adapt to the currently selected model for optimized prompt enhancement.

Arrow Key Navigation: Added keyboard support for canvas panning using Arrow keys (hold `Shift` for accelerated panning).

Fixed: Fixed an issue that upscaler model couldn't be loaded.

Qwen-Image-Edit: Support Qwen-Image-Edit model

Fixed: Fixed an issue with Crop and Outpaint modes.

Qwen-Image Integration: Support Qwen-Image model (requires 32GB+ memory, model download links can be found here).

In-App Quantization: Import full-weight models and quantize them on-the-fly.

Prompt Composer: Compose structured prompts using JSON or natural language for precision.

Custom Size Presets: Save and reuse personalized dimensions for faster generations.

Resizable Action Panels: Dynamically resize prompt inputs within canvas action panels.

Enhanced Memory Management: Refined memory loading strategies to improve stability and performance.

Fixed: Variation mode couldn’t use the z-image model

Cloud Models: Offload heavy computational tasks to the cloud for a smoother experience

Vectorize: Convert raster images into clean, scalable vector graphics.

Image-Guided Inpainting: Use visual references to guide your inpainting edits.

Multi-Selection: Select multiple regions in Inpainting and Eraser tools.

Quick Look (Q): Preview images in full-screen mode in Canvas.

True Aspect Ratio: View images in their true, original proportions inside the Gallery view.

Resizable Prompt Editor: Resize the prompt editor for more editing spaces.

Productivity Shortcuts: Keyboard shortcuts for Eraser and Rearrange.

Batch Asset Management: Multi-select and delete multiple images in Gallery.

Cached Camera Analysis: Drastically reduced camera angle adjustment times following the initial pass.

Responsive Interruptions: Cancel ongoing generation tasks faster and more reliably.

Canvas Optimization: Improved canvas performance with large images.

HDR Color Accuracy: Fixed an issue causing washed-out Display P3 / HDR images

Fixes: Resolved an issue where double-clicking the canvas would unexpectedly close the editing panel.

Note Editing: Fixed a bug where saved notes could not be edited.

Modify Selection (Inpaint): Modify selected area of images on canvas.

Erase Objects: Erase objects from images on canvas.

Expand Image (outpainting): Expand images beyond their original borders.

Sections: Group related items on canvas into sections.

Rearrange: Automatically rearrange items on canvas.

Notes: Add notes on canvas to record your inspiration

Crop Image: Crop images on canvas

Export/Import Canvas: Import and export your canvas with full editing history

Canvas Perfomance: Improved image rendering performance on canvas

Canvas Visual Lineage: Trace the evolution of your ideas and regenerate them with full history restored.

Camera Control: Better following of camera instructions

Sutdio Assistant: Now you can config System Prompt for Studio Assistant.

Lora compatibility: Expand LoRA mapping coverage to support more Z-Image and Flux.2 Loras.

Override Lora types: Manually override Lora type in case the system detects it incorrectly.

Import remote models: import models from Huggingface or ModelScope directly.

Studio Assistant performance: Now responds faster than before.

Smart Suggestion: Fix a problem that Smart Suggestion may not respond in some edge cases.

Infinite Canvas Architecture: A completely redesigned experience centered around a boundless workspace for organizing, comparing, and evolving creative projects.

In-place Power Tools: Remix, modify upscale and remove backgrounds directly on the canvas without switching modes.

Image Modification: Reference images directly on the canvas to guide AI edits like style transfers, object additions, and face swaps.

Advanced Camera Navigator: Professional 3D-native control over perspective, tilt, and rotation for precise shot composition.

Variations: Visual exploration tool for navigating aesthetic variations and complexity levels to land on the perfect creative direction.

Contextual Suggestions: Intelligently analyzes canvas context to recommend optimal prompts and workflows.

Custom Project Storage: You can now choose a custom location to store your studio projects.

Bug Fix: Resolved an issue where Remix (img2img) mode was incorrectly triggered as image edit mode when using FLUX.2 Klein models.

FLUX.2 Klein Improvements: Support for resizing and cropping input images for Klein edit mode.

Import images into Projects: Add images to your projects by simply dragging and dropping them into the app.

Stability Improvements: Enhanced the generation engine to more robustly handle edge cases and improve overall stability.

FLUX.2 Klein: Add support for FLUX.2 Klein 4B & 9B models.

Multi-Image Remix: Use up to 7 images as reference prompts to edit and remix (FLUX.2 Klein only), allowing for complex subject blending and consistent character transitions.

Prompt Translation: Translate prompt into English from any language.

i18n: Added Chinese support.

Model Switching: Fixed an issue where the model type wasn't correctly updated when switching between different models.

Studio Assistant: your new AI companion for image creation

Turn ideas into high-quality image prompts through natural language chat

Analyze images, edit visuals, and transfer styles with simple instructions

Plan and run batch tasks effortlessly

SeedVR2 Upscaler Improvement: Now runs smoothly with less memory consumption

Sequential Seeds: Auto-increase seed on each generation to find the best candidate

Memory Optimization: Smarter memory management keeps everything responsive

Faster & Smoother: Faster app startup and readiness checks

Bug Fix: Fix for Mac AppStore Version where imported models could appear missing after relaunch

Custom Model Library: Point your model library to an external storage device to store your models.

Improved Upscaler Model: New SeedVR2 model for higher quality 2x and 4x upscales.

Remove Background: Remove image background with a single click.

Compare View: View original and the upscaled/removed background image side by side.

Presets: Create and save your own presets with Model, Size, Seed, Lora, and all other possible settings, and reapply them with a single click!

Prompt Library: Add your favorate prompts to the library, and the on-board AI will learn from them and enhace your prompts in the way you want!

Better memory management: Smarter model loading and unloading to free up memory when not in use.

Improved high-res generation: Efficient handling of large images by dividing them into tiles for processing.

Model Loading: Fix a bug that models may not be able to reload after a while of idle time.

MindCraft Studio Launch: Official check-in of the comprehensive MLX-native AI design studio.

Unified Pipeline: Support for Z-Image, FLUX.1, and SDXL models (in MLX/MFLUX formats)

Creative Layers: Support for LoRA, ControlNet(Canny), and Redux.

Quick Upscaling: Instantaneous upscaling of images to 4x.

Model Presets: Pre-configured settings for each model so they can be automatically applied when loading.

Lora Trigger Words: Pre-configure and automatically add LoRA trigger words to your prompts when using LoRA layers.

Onboarding Experience: Mission-based walkthrough for first-time use.