What this workflow does

This is inspired from "Comfy Dungeon" Which I urge you to check out (It's Really Good) → Comfy Dungeon

I built this because I felt some elements were missing from the "Comfy Dungeon", for example - some races, classes, etc, as well as the option to include your own photo as a likeness base.

I do know the original developer for "Comfy Dungeon" is working on including these elements, but in the meantime I have added them to this workflow.

Basically this is a Dungeons & Dragons Portrait generator.

Please also Note, this is an SDXL based Workflow - although it works with SD, it may not yield expected results, and some of the "Processing Group" nodes may have to be amended.

How to use this workflow

This takes advantage of "NX_PromptStyler" and needs some setting up to work correctly (sorry - if you know of a better way, please message me your ideas)

You can download the "NX_PromptStyler" from the Comfyui Manager, or you can get it here → NX_PromptStyler

If you have previously used (or currently use) NX_PromptStyler, then I suggest you make a backup of "..\ComfyUI\custom_nodes\NX_PromptStyler\CSV" and then clear this directory of all files.

Once installed you will need to download the Assets from my Workflow Page:

Copy presets.json (or amend your existing one if you know what you are doing) into "..\ComfyUI\custom_nodes\NX_PromptStyler"

Unzip ADnD_CSV_Files_1.1.zip into "..\ComfyUI\custom_nodes\NX_PromptStyler\CSV"

I suggest restarting ComfyUI and refreshing your browser.

After restarting ComfyUI - you will need to Select the Preset "ADnD Character Portrait Generator"

Hopefully this should all work for you.

🟨 Character Configuration 🟨

🟦 Character Options - This is where you will primarily build your character Portrait.  The options should be self explanatory.  You can even control the weight of each option.  You can ignore the Prompt fields - these are not necessary.  You can utilise the Preset options if you wish.

🟦 Checkpoint - This is where you chose your checkpoint Model - I would suggest using a RPG/Fantasy based Model here.  For the Best results I found Proteus v0.3 to be good, you can get it here → Proteus v0.3

🟦 Include Likeness Image.? - Decide weather or not if the image will try to build the character around your real life looks - This will inject your likeness into the image regardless of what race you choose - Prepare to be Uncanny.! 😊

🟦 Classic / Realistic - Decide if you want a classic AD&D style illustration portrait or a more realistic looking portrait.

🟦 Seed Control - You can keep this random if you wish, but if you like the style of the result but want to tweak a few things, then you can set "Control_after_generate" to fixed.

🟦 Enable/Disable Animation - See 🟪 Bonus Content 🟪 Below

🟦 Likeness Image (Your Photo) - Basically, a photo of you (or whomever you want a likeness of)

🟪 Bonus Content 🟪

I have added an animated GIF creator for those of you who want a live portrait.

This uses ComfyUI-LTXVideo which can be found here with install instructions -> ComfyUI-LTXVideo

🟪 Animation CLIP - This needs to be t5xxl_fp8_e4m3fn.safetensors to work correctly.

🟪 Animation Checkpoint - This needs to be ltx-video-2b-v0.9.safetensors to work correctly.

🟪 Sampler Pass - Fixed Seed "562396681440748" is a nice casual idol breathing animation - but feel free to switch to Random seed.

⬛ Processing Group ⬛

There are a few things in here that you may have to amend - But Most things you don't have to touch.  I have un-pinned and coloured "blue" the ones you might need to amend.

"IPAdapter InsightFace Loader" you may have to select your provider (CPU is default)

"IPAdapter Unified Loader FaceID" you may have to select your provider (CPU is default)

"Face Cropper" you may have to select your onnx_device (CPU is default)

"Load Clip Vision" should be "CLIP-ViT-H-14-laion2B-s32B-b79K"

Tips about this workflow

If you are going to use a likeness image, ideally it should be 1024x1024 square image, with your face in the centre.  But don't worry to much as I have added a process that will detect and correct any image you place here.

Everything should work as is - The Dragon race maybe an issue, so I suggest trying a different Checkpoint Model if this is the case.  If this is still an issue please tell me and I may have to inject a lora into the mix.

If you think you are brave enough, you can edit the CSV files provided to include your own race's and other styles.  But remember to restart your Comfyui server and refresh the browser.

Version Changes

v1.2

• Updated Nodes to latest versions
• Added Bonus Animated GIF Content

v1.1

• Updated Nodes to latest versions
• Fixed issue with NX_PromptStyler (Newer NX_PromptStyler requires "presets.json", previously this didn't)
• Tweaked Illustration configuration to be more like an Illustration.
• Tweaked Realistic configuration to compensate changes to Illustration.
• Added more Races (by request)
• Added more Clothes (By Request)
• Added more Beard Styles (By Request)

v1.0

• Release Workflow