Kokon 0.9.0 (beta)

First public beta of Kokon: edit and create photos by describing the change, on your own GPU, with Qwen-Image 2.1 Turbo. One installer, no Python, no cloud.

Kokon is a native desktop app built on stable-diffusion.cpp. It picks the mode, the working size and the memory settings for you, and includes an MCP server and a local HTTP API so AI agents can create and edit images on your machine.

Download

	File
Windows 10/11	Kokon_0.9.0_x64-setup.exe (installs for your user only, no admin rights)
Mac (Apple Silicon, beta)	Kokon_0.9.0_aarch64.dmg (macOS 26 or newer, 32 GB of memory recommended)

On first launch Kokon detects your graphics card (or your Mac's chip and memory), recommends a model preset and downloads about 12 GB of models once. Downloads resume if interrupted and every file is checked against its SHA-256. After that Kokon works offline.

Highlights

• Create, edit, combine: a sentence makes a new image; a photo plus a sentence edits it, keeping its size and aspect ratio; several photos can be combined ("image 1", "image 2").
• Brush edits: paint an area and choose Change, Remove, Replace or Add. Kokon marks the area on a second reference image so the model knows exactly where to work.
• Expand (outpainting) by dragging the edges, and upscale to 2×, 4K or a custom width with an ESRGAN upscaler.
• Qwen's official 8-step Qwen-Image 2.1 Turbo model for every preset, with three quality presets: Fast (~1 MP, default), Balanced (~1.5K) and Quality (native 2K).
• Before/after compare (Slider, Flicker, Changes), undo/redo between versions, and a gallery that keeps every image's full recipe (also embedded in the PNG).
• iPhone photos: HEIC opens directly; HDR gain-map photos can be saved as Ultra HDR JPEG.
• Private by design: no account, no uploads, no telemetry. Private mode keeps nothing.
• Time estimates that learn your machine's speed after the first image.
• MCP server (kokon-mcp) for Claude Code, Cursor and other MCP hosts, plus a token-protected HTTP API on 127.0.0.1. Tools: generate_image, edit_image, edit_images, inpaint_image, outpaint_image, upscale_image, get_image_info, get_generator_status.
• One-click updates: Kokon checks once a day for a new version (you can turn this off) and installs it, after verifying its signature, when you click Update.

Requirements

• Windows + NVIDIA: 12 GB VRAM or more (RTX 3060 12 GB, 4070 and up; 16 GB+ recommended), CUDA engine. 16 GB RAM, 15 GB free disk.
• Windows + AMD or Intel: 12 GB VRAM or more, Vulkan engine (beta). Integrated graphics are not supported.
• Mac: Apple Silicon (M1 or newer), macOS 26, 16 GB unified memory minimum, 32 GB or more recommended (beta).

Known limitations

• AMD and Intel are unverified. The Vulkan engine works on NVIDIA cards, but nobody has tried it on a real AMD or Intel card yet. It is slower than CUDA, and its first image takes longer. Please report how it goes.
• Macs are slow. On a 16 GB M2 Pro a 1-megapixel image takes about 4 minutes, in the lightest mode. Chips with more GPU cores and 32 GB or more should be faster, and get native 2K. So far the Mac build has been tested only on a 16 GB M2 Pro.
• 8 GB cards are not supported: 12 GB is the smallest size Kokon is designed for.
• No Linux build yet.

Model licenses

Kokon's code is open source under the GNU AGPL-3.0. The models are downloaded from their authors and keep their own licenses: Qwen-Image 2.1 Turbo is under the Qwen Research License (research and evaluation only, no commercial use without a license from the Qwen team), the Qwen3-VL text encoder is Apache-2.0, and the optional 4x-UltraSharp upscaler is CC BY-NC-SA 4.0 (non-commercial only). Details in NOTICE.

First launch

• Windows: SmartScreen may say "Windows protected your PC" because the app is new and has no download history yet. Click More info → Run anyway. You can check the installer against the SHA-256 below.
• Mac: Kokon is not notarized by Apple, so macOS blocks it the first time. Try to open it once, then go to System Settings → Privacy & Security and click Open Anyway. In-app updates don't trigger this again.

Checksums

SHA-256  Kokon_0.9.0_x64-setup.exe
38a2abca067a390d5f5b9bf11884e628a7292db86e4329a1af47e2b89ae9f193

SHA-256  Kokon_0.9.0_aarch64.dmg
c09a931f94d97616bb8e53cd4bbccfb482162a79012f243db44e5c50bb457762

Feedback

This is a beta, built by one person with the help of AI. Please report problems, and how Kokon runs on your GPU or Mac, in Issues. For questions and ideas, use Discussions.