Skip to content

armgit / kokon

Public Imported from GitHub
0

Edit photos by describing the change, on your own GPU. A free, open-source alternative to Nano Banana and ChatGPT image editing: photos stay on your computer, no daily limit. One installer for Windows, Mac and Linux sets up the best open model (Qwen-Image 2.1). MCP server for AI agents.

1.4 KB
Raw Blame

Kokon 0.9.2

Native 2K edits, faster. On graphics cards with 16 GB or more, 2K edits are now drawn at full size instead of enlarged afterwards: sharper, and about a third faster (41 s instead of 66 s for a 2752×1536 edit on an RTX 5070 Ti).

  • Reference detail (Advanced menu): your source images go to the model at 1 MP by default, leaving more room for the result. Choose Full for finer source detail. Every new chat starts at 1 MP.
  • Faster decoding on 16 GB+ cards: tiles shaped to the image (7.8 s → 4.75 s at 2752×1536).
  • Official stable-diffusion.cpp. The MoSR upscaler Kokon contributed was merged upstream (#2128), so Kokon now runs the official release master-964-1d0a58a instead of its own fork. The app checks every few days for a newer engine tested with Kokon and offers it in one click.
  • Prompt tips: a hint under the prompt and a Prompt tips panel. Describing the scene, not only the subject, keeps the action you asked for.

Updating from 0.9.1: the app asks to download the new engine (~340 MB on NVIDIA, ~30 MB with Vulkan or on a Mac). Your images, models and settings stay as they are.

Downloads

File SHA-256
Kokon_0.9.2_x64-setup.exe (Windows) 8766b5aedc8fce50fc8805853ec54d60a72cf0ed5aa609941b4c6d6098e1a03c
Kokon_0.9.2_aarch64.dmg (Mac) cc91171a74d6275a3d0216dcb9529c1065a2617ee75d393be35b91e7f9464d2e