ArtCraft is an open-source desktop crafting environment for artists, designers, and filmmakers who want precise, repeatable control over AI-generated images, video, and 3D scenes instead of relying on free-form prompting alone.

What it is

ArtCraft describes itself as the IDE for interactive AI image and video creation: you compose in 2D, stage scenes in 3D, and choose the models that fit your work, turning prompting into crafting with visual tools. It is written in Rust, released under the Apache-2.0 licence, and lives in the AI interaction and interfaces space alongside tools for image generation, AI video, 3D graphics, and filmmaking. The project ships as a desktop application, distributed through GitHub releases or built from source, with a homepage at getartcraft.com.

The concrete problem it addresses is that artists need control over composition, character identity, and camera placement, and prompting alone does not give them that. ArtCraft's answer is to have you build the scene before you generate it: layer backdrops, foregrounds, and props in 3D, pose characters and set the camera, block scenes by combining 3D asset kits, and place virtual actors in a consistent environment so multiple shots can be planned in the same room. In doing so it replaces a workflow where scene layout, subject isolation, and shot planning happen across separate image editing and 3D tools, then get re-described in prose prompts.

Key capabilities

• 3D image compositing for layering backdrops, foregrounds, and props to build compositions with depth.
• 2D image compositing using image layers, background removal, and drawing tools, plus canvas image editing with drawing, masks, and inpainting to guide exactly where an image changes.
• Image to 3D Mesh, which turns an image into a 3D object you can position, rotate, and frame precisely.
• Scene Blocking with Kitbashing, combining 3D asset kits to control camera angles, object placement, and depth.
• Character Posing and Character Identity Transfer, where a posed mannequin guides the placement and pose of your character before you generate the final shot.
• Background Removal to isolate subjects for use as props, layers, or backdrops in both 2D and 3D scenes, supporting Mixed Asset Crafting of image cutouts, worlds, and meshes in one scene.
• Model choice across a 62-model catalog spanning images (16, including Nano Banana, GPT Image, FLUX, and Seedream families), video, music and sound, 3D meshes, and worlds, with providers including Grok, Midjourney, Sora, and World Labs (Marble world generation with Gaussian splats).

Who uses it and how

• Filmmakers and directors who block shots first: posing characters, setting the camera, and planning multiple shots in the same room through Image to Location before any generation happens.
• Illustrators and compositors who build layered scenes with 2D and 3D image compositing, background removal, and inpainting on the canvas.
• Concept and environment artists who start from a text prompt to explore ideas, then refine using the canvas and scene tools.
• Teams and individuals comparing output across providers — Grok Imagine and Grok Video, Midjourney, Sora 1, Sora 2, GPT Image 1, and the ArtCraft catalog — rather than committing to one vendor's models.

Getting started

Download ArtCraft from GitHub releases or build from source, and refer to the README for the build instructions; a two-minute demo video is linked from the project page.

How it compares

This registry names no comparable or paid alternatives to ArtCraft, so the project stands alone here; its differentiators in the supplied facts are its Apache-2.0 licence, Rust implementation, and self-run desktop distribution rather than a hosted service.

When to use it — and when not

Pick it if you want scene construction — posing, camera, layers, meshes — to precede generation, and accept that you are operating a desktop application whose output depends on external model providers for several workflows. Do not choose it if you need a particular model guaranteed: the catalog notes models that are currently disabled and a restricted model not currently available in the desktop app, the README indicates previews for scene blocking, canvas editing, scene relighting, image inpainting, and image ingredients are still coming, and 171 issues remain open.