ComfyUI-DonutNodes Introduction
ComfyUI-DonutNodes is a collection of nodes built around Donut Workflow V5, an image-making workflow for ComfyUI. The workflow brings model selection, prompts, LoRAs, image editing, generation, refinement, and saving into one organized setup. You can start with the complete workflow or use individual DonutNodes in workflows you build yourself.
V5 is designed to make a complex creative process easier to navigate. Its numbered control cards keep everyday choices visible, while collapsible Advanced settings leave more room to focus on your images. It also gives you ways to inspect what went into a generation, follow its progress, and review the result.
Download Donut Workflow V5 · Read the workflow setup and usage guide
Overview of Donut Workflow V5's redesigned controls
How ComfyUI-DonutNodes Works
Think of the workflow as a series of creative stations. You choose a model and any LoRAs you want to use, prepare your prompt or source images, generate an image, optionally refine it, and save the result. DonutNodes supplies controls for several of those steps; the workflow connects them into a ready-made process.
For example, you might begin with a prompt and optional reference images to guide a new Krea2 generation. If you are editing an existing image, Edit Studio groups your source images, crop controls, and output sizing choices together. After generation, the workflow can show intermediate and final images so you can see how the result developed.
You do not have to use every capability on every run. The workflow includes optional editing and refinement paths, and the individual nodes can also be placed in your own ComfyUI graphs.
ComfyUI-DonutNodes Features
- Organized workflow controls: V5 replaces V3’s spread-out controls with numbered cards. Common choices remain accessible in both Graph Mode and App Mode, while less frequently used settings can stay collapsed under Advanced.
- Edit Studio: Bring together a base image and, optionally, an identity image. You can paste, drop, or upload images, adjust the crop visually, and set the output size in the same area.
- Reference guidance: Add optional image references to guide native Krea2 generation. This is a separate path from the edit-LoRA controls, so a reference used to guide generation is not the same thing as choosing an edit LoRA.
- Wildcard library: Create reusable prompt choices, insert wildcard tokens into a prompt, and inspect the final expanded prompt. This can help when you want variation without rewriting the whole prompt each time.
- Model downloads: Check selected files against a reviewed model catalog and download missing catalogued files. The downloader verifies file size and hash, giving you a way to check that a download matches the catalogued file.
- Generation progress and latest result: Follow the stages of a run and view intermediate as well as final images. This makes it easier to tell where the workflow is and what each stage contributed.
- Flexible image saving: Donut Image Save supports numbered filenames and configurable output formats, including WebP. It does not depend on WAS.
- Additional creative nodes: The pack includes block-weighted LoRA stacking, Krea2 model merging and Fusion Control, CFG sampling curves, detailers, tiled upscaling, SDXL TeaCache, and spectral sharpening. V3 already included face detailing, editing, model merging, LoRA stacking, and upscaling; those capabilities are not all new to V5. Experimental execution modes are optional.
For explanations of the individual nodes and their advanced controls, see the node guide.
ComfyUI-DonutNodes Models
The provided documentation describes a Krea2 model setup and model-related controls, but does not list specific model filenames or a complete set of supported models. The workflow lets you select models and LoRAs, use optional image-reference guidance for native Krea2 generation, and use DonutNodes for Krea2 model merging.
If you need to identify the files required by the workflow, consult the Donut Workflow V5 guide. Select the files appropriate to the path you intend to use rather than assuming that every optional editing or refinement feature is needed for every image.
What's New with ComfyUI-DonutNodes
Donut Workflow V5 introduces a redesigned interface with numbered cards, collapsible Advanced settings, and source and generation subgraphs you can inspect. For everyday use, this means fewer controls competing for your attention while the more detailed parts of the workflow remain available when you need them.
V5 also adds Edit Studio, optional reference guidance for native Krea2 generation, a reusable wildcard library with expanded-prompt inspection, catalog-based model downloads with verification, a latest-result view, and Donut Image Save with numbered filenames and formats including WebP. These changes help you prepare inputs, understand prompts and generation stages, manage missing files, and save outputs from within the workflow.
The V5 workflow release name is separate from the DonutNodes package version. An older DonutNodes package may not contain the nodes used by V5, including DonutImageSave and DonutEditStudio. For the distinction between V5 additions, capabilities already present in V3, and installation fixes, see the workflow changelog.
Troubleshooting ComfyUI-DonutNodes
The workflow reports missing nodes. What should you check?
- Check that your DonutNodes version includes
DonutImageSaveandDonutEditStudio. An older package is not sufficient for V5. - In ComfyUI, use Install Missing Nodes → Install All, then apply the changes and restart.
- Keep the workflow’s collapsed Required node packs subgraph in place; it helps ComfyUI detect companion-pack requirements.
- Refresh the browser after restarting ComfyUI. The workflow guide provides the full dependency and usage details.
A selected model or LoRA is missing. What can you do?
Check the filenames in the workflow guide. For files included in the reviewed catalog, use Download missing and let the workflow perform its size and hash checks.
Do you need WAS to save images in V5?
No. Donut Image Save supports the workflow’s saving path, including WebP, without a WAS dependency. The documented fresh-install test did not require WAS installation or manual WAS version selection.
Does one successful test mean every mode and computer is covered?
No. The documented beta test completed a fresh Linux/Python 3.12.7 path on an RTX 4070, including base generation, a first upscale, two face refinements, and a 1728 × 1344 WebP save. That run reused models and did not test editing or the second upscale. Other modes, platforms, and GPU configurations still need coverage. Read the validation report for the test scope and remaining warnings.
What information helps diagnose an error?
Use Donut Dependency Check to inspect dependency problems. When reporting an error, include the workflow, relevant versions, your operating system and GPU, and the full traceback. The dependency and compatibility guide is another place to check when a workflow will not run as expected.
Learn More about ComfyUI-DonutNodes
- Donut Workflow V5 setup and usage: model filenames, companion packs, editing, wildcard files, saving, and migration from V3.
- Workflow changelog: changes introduced in V5 and capabilities carried forward from earlier versions.
- Node guide and advanced settings: details for using DonutNodes beyond the main workflow.
- Dependency diagnostics and compatibility: guidance for investigating missing nodes and related problems.
- Validation report: the documented beta-test setup, tested path, and known limits.
Two separate, optional packages are also available: ComfyUI-DonutLocalAutomation provides local Prompt Receiver and Image Reporter nodes, while ComfyUI-DonutCivitaiLocal provides local CivitAI library and workflow-recovery tools. They are distinct from the companion packs required by V5. Older personal workflows containing DonutPromptReceiver or DonutImageReporter need DonutLocalAutomation to retain those nodes.
