What Is ArtCraft?
ArtCraft calls itself the IDE for artists, and the name fits better than most project slogans do. It is an open-source desktop app for interactive AI image and video creation where you compose in 2D, stage scenes in 3D, and choose the models that fit your work, instead of typing a prompt into a box and hoping the result lands where you imagined it. Prompting becomes crafting: visual tools give you precise, repeatable control over composition, character identity, and camera placement.
It is a Rust project, dual-licensed under MIT or Apache-2.0 at your option, and it had 39,133 stars when we checked on November 6, 2026, with 36,700 of them gained this month alone. That kind of surge usually means a project solved a frustration a lot of people share, and here the frustration is obvious: generative models are powerful but hard to direct, and ArtCraft is built around directing them.
Build the Scene Before You Generate
The core idea is scene first, generation second. You place virtual actors in a consistent environment and plan multiple shots in the same room with Image to Location, layer backdrops, foregrounds, and props in 3D to build compositions with real depth, and turn an image into a 3D object you can position, rotate, and frame precisely with Image to 3D Mesh. Character Posing lets you pose characters and set your camera before generating the final shot, Scene Blocking with Kitbashing combines 3D asset kits to control camera angles and object placement, and Character Identity Transfer uses a posed mannequin to guide the placement and pose of your character.
The README makes a point of showing its work: the camera rings each feature row with its real preview image, so you can see the tools doing what they claim before you install anything. Background Removal isolates subjects to use as props, layers, or backdrops, and Mixed Asset Crafting combines image cutouts, worlds, and 3D meshes in one scene. When you are matching a backdrop to a prop, getting the palette right matters, and a quick pass through a Color Converter keeps your HEX, RGB, and HSL values consistent between the scene and the assets you bring in.
Prompting When You Want Speed
You do not have to stage a full scene for every idea. Start with a prompt when you want to explore quickly, then use the canvas and scene tools to refine the result. Text to Image explores ideas across a wide range of image models from a text prompt. Prompted Image Editing describes changes to an image with models such as Nano Banana Pro and GPT Image. Canvas Image Editing uses drawing, masks, and inpainting to guide exactly where an image changes, so you can fix one corner without regenerating the whole frame.
Image to Video animates an image with your choice of video models and generation controls, which is where the staging pays off a second time: a scene you blocked in 3D gives the video model a consistent world to move through instead of a single flat frame. Before you render, it is worth checking your framing, since a vertical 9:16 cut and a 16:9 cut are different compositions, not the same image cropped, and an Aspect Ratio Calculator works out the exact dimensions for each format.
The 62-Model Catalog
ArtCraft does not lock you to one provider. Its own catalog covers 62 models: 16 image models across Nano Banana, GPT Image, FLUX, and Seedream; 25 video models across Seedance, Kling, Veo, Sora, Vidu, MiniMax, and Flux; 5 music and sound models including Suno; 11 3D mesh models including Hunyuan 3D, Tripo3D, Meshy, and Rodin; and 5 world and Gaussian splat models including Marble and TripoSplat. On top of that it connects to Grok Imagine and Grok Video, Midjourney image generation, Sora 1 and 2 plus GPT Image, and World Labs Marble world generation.
A few models in the catalog are currently disabled or restricted, and more integrations are planned for Kling, Google, Runway, and Luma, plus aggregators such as OpenArt and Freepik for existing subscriptions. The catalog is the practical answer to the most common AI art question, which model should I use for this, because you pick inside one workspace instead of opening five tabs and five accounts.
What It Costs
Here is the part to understand before you start: the ArtCraft app itself is free, but you connect your own model subscriptions, and the description says so. Generation runs through the providers you have accounts with, so your cost depends on which models you use and how much you generate, not on ArtCraft. That is the honest trade of a bring-your-own-model tool: no subscription to the app, but no free generation either.
Downloads are straightforward. Stable releases for Windows and macOS are on the project site, the latest builds are on GitHub, and you can build from source including on Linux, with the setup documented in the repo. There is also an active Discord community and a YouTube channel with a two-minute demo that shows the full workflow from staging to finished video.
What to Watch Out For
First, this is a young project moving very fast: 36,700 stars in a month means features, models, and docs change often, so treat the README as the current truth and expect the catalog to keep growing. Second, some models in the catalog are disabled or restricted, so a model you saw in a demo may not be available in the desktop app on the day you try it. Third, because you bring your own subscriptions, costs are yours to manage across providers, and a heavy video session can add up if you are generating with the premium models.
None of that changes what makes it interesting. Most AI art tools ask you to describe a scene and accept what comes back. ArtCraft asks you to build the scene first, then generate inside it, and that inversion is why artists who care about composition keep coming back to it.
