<h1 id="donut-workflow-v5">🍩 Donut Workflow V5</h1>Krea2 • DonutNodes 3.0.23+Generate, edit, inpaint and outpaint in one workflow—with organized panels, flexible references and previews for every finishing stage.💬 <a target="_blank" rel="ugc" href="https://discord.gg/5djEeQ3wpU">Discord — help, artwork &amp; updates</a> · ☕ <a target="_blank" rel="ugc" href="https://ko-fi.com/donutsdelivery">Support on Ko-fi</a> · 🛠️ <a target="_blank" rel="ugc" href="https://github.com/DonutsDelivery/ComfyUI-DonutNodes">DonutNodes on GitHub</a><h2 id="v5-highlights">✨ V5 highlights</h2><ul><li>Cleaner layout — separate panels for setup, generation, upscaling and face detail.</li><li>Inpainting &amp; outpainting — paint or draw a selection, invert it, adjust the seam, or expand with directional canvas presets. Outpainting stays within your megapixel budget.</li><li>Smart masks — describe what to select with SAM3.1, or use BiRefNet for foreground/background separation. Refine masks manually.</li><li>Independent A/B crops — frame each reference separately, with clipboard paste and optional reference guidance.</li><li>Flexible finishing — first and second upscales, Face Detailer and SeedVR2, with individual result previews.</li><li>Prompt variants — add, duplicate and reorder prompts; keep one active or cycle automatically. Includes autosizing editors and shared wildcard controls.</li><li>Searchable LoRAs — find models by typing, then adjust strengths and block weights.</li></ul><h2 id="quick-start">🚀 Quick start</h2><ol><li>Install/update DonutNodes 3.0.23+. Restart ComfyUI and refresh your browser.</li><li>Open DonutWF_v5.json.</li><li>Use Manager → Install Missing Nodes, then restart.</li><li>Install the models using one of the options below.</li><li>Review your models, LoRAs, prompt and resolution. Leave Editing off for your first text-to-image run, then hit Run.</li></ol>📦 GitHub installation: use Download missing for the supported model catalog.📦 Registry/Manager installation: download the optional <a target="_blank" rel="ugc" href="http://DonutNodes-model-installer.zip">DonutNodes-model-installer.zip</a> from this version’s files. Extract it, follow the included README and run the Windows or Linux/macOS launcher. It downloads catalog models into the appropriate ComfyUI folders and reuses matching existing files. Some downloads may require an access token.The Registry package provides model links and the separate installer link instead of the built-in downloader.<h2 id="editing-masks-and-outpainting">🎨 Editing, masks &amp; outpainting</h2>Enable Editing, add your base image to A, and optionally place a second reference in B. Describe the change in the prompt.Open Paint / outpaint… to edit a selected area or extend the canvas. Use Overlap to let generation reach into the original image and Seam width to soften the transition.For B, choose Prompt selection to mask a named object, or Auto subject for the foreground. Use full-resolution references for better results.Reference guidance is a separate option for borrowing visual elements during generation; it pauses while Editing is enabled.<h2 id="saving-and-moving-workflows">💾 Saving &amp; moving workflows</h2>The supplied preset saves to output/Final with seed-based filenames. Saving supports PNG/WebP workflow metadata.Keep your old workflow when upgrading, then copy your personal settings into V5. Reference images and wildcard files are stored separately—copy those too when moving installations.<h2 id="feedback">🐛 Feedback</h2>Found a bug? Share your workflow with private content removed, ComfyUI/DonutNodes versions, GPU/VRAM and the full error on <a target="_blank" rel="ugc" href="https://discord.gg/5djEeQ3wpU">Discord</a>.High-resolution editing with multiple references and NAG can be VRAM-heavy. Seam quality depends on the image, prompt and overlap settings.

Professional Krea 2 Workflow for high IQ individuals

# Donut Workflow changelog## V4 Beta — compared with the original V3This comparison uses the original V3 JSON supplied by the author, not anintermediate development file that was also named V3. The source fingerprintsand structural comparison are recorded in [comparison.json](comparison.json).### Redesigned interface- Replaced the spread-out loader/settings layout and rgthree labels with numbered Donut control cards and collapsible Advanced sections.- Moved source loaders and generation internals into inspectable subgraphs.- Added direct block-weight slider controls, automatic card sizing and shared controls across Graph and App Mode. LoRA stacking and ordering existed in V3.- Added a consolidated latest-result panel, stage progress and final expanded prompt display while retaining the individual stage outputs.### Images and editing- Replaced the separate Load Image, editing toggle, edit-LoRA loader and resolution selector with Edit Studio.- Added two image slots: A for the base/scene and optional B for subject/identity, including B-based face-reference routing.- Added paste/drop/upload, visual crop controls and integrated preset/custom, reference-aspect and crop-aligned sizing.- Added Reference guidance as a separate native image-conditioning path for generation, without an edit LoRA. It pauses while Editing is enabled.V3 already had a single-image edit path and a face detailer. These changes areediting/control additions, not a claim of universally smarter face detailing.### Prompts and guidance- Added a persistent Wildcard library, picker and expanded-text preview; explicit wildcard tokens replace the old automatic prompt-addition controls.- Added shared seed-variance controls for general and face positives, including reapplication to freshly encoded edit conditioning.- Integrated NAG controls into the sampler, upscale and face-detail paths, including edit-mode negative conditioning. V3 already used a standalone NAG node; NAG itself is not new to the workflow.### Installation and models- Added Download missing with a reviewed model catalog, size/hash checks, reuse of matching local files and cancellation.- Added a collapsed Required node packs subgraph so Manager can discover internally called Krea2 Edit, NAG and seed-variance dependencies.- Fixed DonutFaceDetailer registration depending on Impact Pack loading first. Dependency loading failures are distinct from image-quality improvements.- Removed WAS and rgthree label requirements from this workflow. Seven companion packs remain discoverable through the normal missing-node installer.### Saving- Replaced WAS Image Save with Donut Image Save, without adding dependencies.- Preserved the active Final folder, seed-based filenames, numbering, WebP quality, overwrite setting and secondary core save output.- Retained MIT attribution for the adapted WAS filename numbering code.- Removed unused WAS-only history, color-profile and high-bit-depth/EXR controls.### Defaults and compatibility- Existing V3 concepts remain: model merging/Fusion Control, LoRA stacking, Turbo sampling, face detailing, two upscale stages and seed planning.- The beta is a new preset as well as a new layout; saved settings differ from the original V3. Carry personal settings over explicitly rather than assuming the same prompt/seed will reproduce a V3 image.- The distributed configuration has editing off, first upscale on and second upscale off. It includes optional seed variance and integrated NAG controls.- Requires the updated DonutNodes code shipped with this workflow. A node-pack release number is not the workflow's V4 version number.### Validation- Fresh default missing-node installation, restart and full generation passed with the development DonutNodes code; no WAS selection or repair override.- Verified base generation, first upscale, two face refinements and a 1728 × 1344 WebP save. Output matched the preceding WAS save for the same tested preset.- Six saver tests and 34 registration/dependency-isolation tests passed.- Editing, the disabled second upscale, model downloads and Windows/macOS were not exercised by that final fresh-install run. These remain beta test areas.

Files with SHA256: 60b198ede3739f1a91822dbdf1347e59c40bf218709315d3768ea96194334c7f