Overview This advanced ComfyUI workflow automatically creates a complete, professional multi-angle character reference sheet from a single reference image. It uses WD14 Tagger to intelligently extract clothing, accessories, and outfit details, then applies sophisticated tag filtering to maintain perfect clothing consistency across different body views. The workflow generates 6+ high-quality, consistent camera angles (full-body front/rear/side + upper-body close-ups) and automatically stitches them into one clean, organized reference sheet.
It is ideal for character designers, LoRA trainers, animators, and anyone who needs reliable multi-view consistency without manually prompting every angle.
Key Features
Smart Outfit Extraction — Load any reference image; WD14Tagger + custom TagFilter/TagRemover nodes automatically pull clothing tags while intelligently removing lower-body elements for upper-body close-ups (and vice versa).
Multi-View Consistency — Generates full front, full rear, side profile, upper-body close-up, upper rear, and upper side views in one pass using carefully crafted pose and camera prompts.
Automatic Reference Sheet Stitching — Final ImageBatchMulti + Image Stitch nodes combine all views into one professional-looking sheet (with optional brightness/contrast post-processing).
Modular & Customizable — Separate prompt boxes for each view, bypass options, detailed notes inside the workflow, and easy LoRA swapping.
Quality-Focused — Built-in high-quality base prompts, perfect hands emphasis, cel-shading style, neutral color balance, and post-processing adjustments.
Character LoRA Support — Includes an easy loraStack node (example uses a Carlotta Wuthering Waves LoRA) with a MangoTriggerExporter helper.
How It Works (High-Level Flow)
Load reference image → WD14 Tagger extracts detailed tags.
Parallel tag filtering paths: one optimized for full-body outfits, another that strips lower-body clothing for clean upper-body close-ups.
Multiple KSamplers run in parallel with view-specific prompts (front/rear/side/close-up).
VAE decode + brightness/contrast adjustment.
Text concatenation merges view-specific prompts.
Image stitching produces the final reference sheet.
Inputs
Reference image (outfit or character photo) via LoadImage node.
Base style/quality prompt (highly detailed, pre-filled with excellent anime/Illustrious-friendly tags).
Character description / LoRA trigger words.
Optional: Your own character LoRA via the easy loraStack node.
Outputs
Individual high-resolution views (great for dataset curation).
Final stitched multi-view reference sheet (PreviewImage node).
Recommended Usage & Tips
Best results with clear, well-lit reference images of outfits or characters.
The workflow is pre-configured for anime/Illustrious-style models (example checkpoint: cstatiANIME25DXL_v1025DSOFT).
Swap the example Carlotta LoRA with your own character LoRA — the MangoTriggerExporter node helps you quickly grab trigger words.
Expand the collapsed nodes and read the in-workflow Notes for customization guidance (especially the tag exclude lists).
Use fixed seeds or the same seed across views for maximum consistency when creating training data.
Excellent for generating LoRA training datasets — the clean background + consistent character across angles makes labeling much easier.
Works for both SFW and NSFW depending on your base prompts.
Requirements
ComfyUI + ComfyUI Manager (strongly recommended).
Custom nodes (install via Manager): comfyui-wd14-tagger, comfyui-easy-use, was-node-suite-comfyui, ComfyUI-Impact-Pack, ComfyUI-Chibi-Nodes, rgthree-comfy, comfyui_layerstyle, and others listed in the workflow.
Recommended base model: Illustrious series or any high-quality anime/SDXL checkpoint compatible with your LoRAs.
Customization Notes
Edit the “Full Body Outfit” and “Upper body Only Outfit” Textbox nodes to manually override extracted clothing.
Modify the exclude tag lists (two PrimitiveStringMultiline nodes) to fine-tune what gets removed for different outfit styles.
Change the view-specific prompt boxes (Front View, Rear View, Close-up View, etc.) to alter poses or add expressions.
The workflow contains many helpful notes and a “Hidden nodes [leave on]” group — keep those enabled.
Edge Cases & Considerations
Very complex or heavily layered outfits may require manual tag tweaking.
Extremely loose or flowing clothing can sometimes be partially filtered — adjust the exclude lists accordingly.
For maximum face/body consistency across angles, pair this workflow with a strong character LoRA or IPAdapter workflow in a second pass.
The stitched output is optimized for reference use; you can further upscale individual views if needed using the built-in PixelKSampleUpscalerProvider.
This workflow saves hours of manual prompting and iteration while delivering reliable, production-ready character reference sheets.