A simple ComfyUI alternative for editing photos
ComfyUI gives you every knob. Kokon gives you the result. If you want to edit photos with AI on your own computer without building workflows, this page explains what Kokon decides for you, and when ComfyUI is still the right tool. Updated October 21, 2026.
ComfyUI is great, if you want to build
ComfyUI is the most flexible way to run image models locally: a node graph where you wire up every step. That flexibility is the point, and it's also why so many people bounce off it. To edit one photo you need Python or the portable build, the right custom nodes, four or five model files in the right folders, a workflow, and settings that fit your VRAM.
Kokon is built on a different idea: most people want the edited photo, not the pipeline. So we make the technical choices, test them, and ship them as one app.
What Kokon decides for you
| Choice | Kokon | ComfyUI |
|---|---|---|
| Model and quantization | The best open model we've tested, in the right size for your GPU | You pick and download |
| Text encoder, VAE, vision projector | Downloaded and checked with SHA-256 | You find the matching files |
| Steps, CFG, sampler | Tested defaults | You tune them |
| Resolution and VRAM split | Chosen per job for your GPU | You avoid out-of-memory errors |
| Inpainting (paint an area) | Built in: Change, Remove, Replace, Add | An inpainting workflow |
| Upscaling | One button: 2×, 4K or custom | An upscaling workflow |
| Updates | In-app, one click | Nodes and dependencies to keep in sync |
What's left for you is three choices: the photo, the sentence, and Fast, Balanced or Quality.
Already have ComfyUI files?
Kokon's setup can use model files you already have on disk, for example in a ComfyUI models folder, instead of downloading them again. Your own Qwen-Image 2.1 LoRAs work too (beta): Kokon makes its own prepared copy and never changes your file.
When to stay with ComfyUI
- You want custom pipelines: ControlNet chains, video, many models in one graph.
- You want to tune samplers, schedulers and steps yourself.
- You use models Kokon doesn't ship. Kokon is opinionated on purpose: one tested model, not a model zoo.
Many people use both: ComfyUI for experiments, Kokon for quick edits and for friends who will never open a node graph.
Is it the same engine?
No. Kokon runs the model through stable-diffusion.cpp, a native C++ engine (Kokon ships its own tested build), so there is no Python at all. It uses CUDA on NVIDIA, Vulkan on AMD and Intel, and Metal on Macs.
What you need
| Works | Recommended | |
|---|---|---|
| Windows or Linux + NVIDIA | 12 GB of VRAM (8 GB in beta) | 16 GB or more |
| Windows or Linux + AMD / Intel Arc | 12 GB of VRAM, Vulkan (beta) | 16 GB or more |
| Mac | Apple Silicon, 16 GB, macOS 15 (slow) | 32 GB or more, Pro/Max/Ultra chip |
| Disk | 15 GB free | 25 GB free |
Speed with the default model: on an RTX 5070 Ti, about 24 seconds for an edit at 1.5K and about 36 seconds for a new image at native 2K. 12 GB cards are a little slower, 8 GB cards take 1 to 2 minutes per edit, and an M2 Pro Mac takes about 4 minutes per image.
Questions
Is there a simpler alternative to ComfyUI?
For editing and creating photos, yes: Kokon is a free desktop app with no nodes and no Python. It downloads the model and picks the settings for your graphics card; you describe the change.
Can Kokon use my ComfyUI model files?
Yes, if they are the same model files Kokon uses. Setup can point at files you already have on disk instead of downloading them again.
Can I change the sampler, steps or CFG in Kokon?
Not next to the prompt: Kokon uses tested defaults so you never have to. If you need full control over the pipeline, ComfyUI is the better tool.
Does Kokon support LoRAs?
Yes, in beta: your own Qwen-Image 2.1 LoRA files, loaded in Settings and turned on per project, up to three at a time.
Try Kokon
Free and open source for Windows, macOS and Linux. One installer; Kokon sets up the model for your graphics card.