v1.4

• added Qwen 3.8 27B model for Ask mode
• added reference image list for Ask mode

v1.3

• added video player
• improved options menu
• CPU offloading options can be specified manually

v1.2

• added negative prompt (optional, if negative_prompt.txt exists)
• improved reference image scaling (don't allow higher resolution than generation resolution)
• RAM and VRAM usage display in the title bar
• window size and placement remembered across runs
• improved CPU usage when idle

v1.1

Mini-AI v1.1

• added option to open output folder to right-click menu
• added batch image generation option to right-click menu
• changed how window resize works, the current image will keep aspect ratio fixed
• improvements for model downloader
• added nicer visual feedback for generation
• added LTX 2.3 video
• added Gemma 4 LLM
• added Qwen image and image edit
• added Ernie-image Turbo
• added Minimax H3 video

v1.0

Mini-AI v1.0

A tiny graphical application for running simple AI generation tasks locally on your machine.
Made in C++ and WinAPI, no external dependencies.

Features

• Image generation: You simply type a text and the image will be generated.
  • Z-Image Turbo
  • Flux 2 Klein
  • Stable Diffusion 3.5 Large
• Image Editing: Edit the current image with a prompt and other reference images
  • Flux 2 Klein
• Ask: enter a text and the AI will answer, also supports image if one is opened
  • Qwen 3 VL
• Video: generate a video from text prompt, and/or from the reference images
  • Wan 2.2
• Copy-paste, undo-redo, resize, save, load, drag and drop, PNG, JPG, TGA, ICO support

How to use

• Write a text prompt into the textbox and hit the bottom right button to generate image (or choose an other mode from its drop-down menu)
• On first use the required AI models will be downloaded from the internet for the currently selected mode into the models/ folder, which can take long as they are often several GB large.
• Generated images and videos will be saved to the output/ folder near the exe.
• You can resize the window to set your generation resolution by dragging the window bounds, or choosing from resolution presets in the right-click menu.
• You can use Ctrl+C and Ctrl+V to copy images from/to the mini-ai application.
• You can create a seed.txt and write your random seed in there if you would like to fix the generation seed.
• To change image generation model, right click -> Image generation model... -> choose model
• You can save any image as .ico to use for a windows application icon, it will have multiple resolutions embedded in the file.