Using this workflow and enjoying my work? Leave a 👍!
Seeing the number of downloads increase is fantastic, but your likes really prove to me that this project is useful to you. That's my main motivation to keep releasing updates and fixing bugs. A huge thank you to everyone who takes the time to do so!
The next version is in the works; I think you're going to love it.
Character Sheet Max Rev3 is out !
v2.0r3
Changes
Fixed 'identity' issue in version 2.0r2
Added Compact Sheet mode (5 image slots ): fewer cells, but better image quality (Toggle ON/OFF on "Global Parameters" Section)
Latent scaling model: switched from FP16 to BF16 format (faster, with minimal loss)
Turbo LoRA model: switched from the base Turbo version to a resized version (faster)
All 12 character sheet slots are now available (instead of 11 + 1 reference image)
Enable/disable Turbo LoRA via the global panel
Enable/disable Reference LoRA via the global panel
New custom nodes replacing two other packs
Reduced the "spaghetti" effect in sub-graphics
Improved prompt syntax
Fixed resizing below maximum limits
Fixed an identity issue in version 2.0r2
Change Save Folder Structures :
Final Sheet Folder :
Sheets/Characters/"SubjectName"/"SubjectName"_0000x.png
Individual Images Folder :
Sheets/Characters/"SubjectName"/IndividualImagesFolder/Frame_0000x.png
Average generation time on 5060 Ti 16 GB (varies by reference image)
1 reference picture + Full mode + Latent Upscale + RTX Upscale + saving all pictures (13 pictures): ~350 sec
1 reference picture + Compact Mode + Latent Upscale + RTX Upscale + saving all pictures (6 pictures): ~200 sec
Hi everyone,
Always with a Modular, Intuitive and Clean approach. Here is finally the New Version of the MICW - Fast Character Sheet Generator Using Reference Images (v2 or Max) !
In this update, I’ve taken your feedback into account and implemented several major improvements:
Support for up to 9 Reference Images: While the base workflow only supported one, the prompt engine is now dynamic. You can now use the Character Sheet Prompt Maker to modify "Subject Definitions" and "Retention analysis" to leverage additional reference images.
Dynamic Prompting & Camera Control: Style, retention, camera angles, and subject definitions are now completely dynamic. While this brings immense flexibility, it can occasionally make the camera unpredictable depending on the subject or style. I spent countless hours fine-tuning the prompts to stabilize generation (hence the delay), but rare issues can still occur—such as an off-angle shot, two characters in the same frame, or a skipped view.
Model Optimization: Styles have been tailored specifically for MiniMax Singularity (results with other models may vary).
Fixing Timeline Shifts (Sheet Composer): Unlike the Compact version, dynamic prompts can introduce slight timing shifts in video frames. To fix this seamlessly without re-rendering, I added two new nodes:
Frames Preview with Index: Helps you identify the exact frame index for every generated image.
Sheet Composer: Allows you to manually assign those frame indexes to each grid cell on the final sheet.
Quick Tip: Because this step happens after the VAE Decoder, changing these frame numbers does NOT trigger a full re-generation. Since images are cached, updating the indexes and pressing
Ctrl + Enter(or clicking Run) will regenerate your final sheet in just 5 seconds!
Legacy Mode: A Legacy Mode toggle on the Character Sheet Prompt Maker lets you revert to the v1 8-slot layout with default camera angles, while still benefiting from the v2 prompt improvements (except camera angles). If timing shifts occur in Legacy Mode, simply toggle Adjust Legacy mode images on the Sheet Composer to adjust manually.
Expanded Grid & Options: v2 now supports 11 customizable slots (Slot 1 is locked to Full Body - Front, while the remaining 10 can be freely selected from 33 available angles).
Available Camera Angles:
"Full Body - Back"
"Full Body - Both Sides"
"Full Body - Left Profile"
"Full Body - Right Profile"
"Upper Body - 3/4 Front Left"
"Upper Body - 3/4 Front Right"
"Upper Body - 3/4 Back Left"
"Upper Body - 3/4 Back Right"
"Lower Body - 3/4 Front Left"
"Lower Body - 3/4 Front Right"
"Lower Body - 3/4 Back Left (may not work sometimes)"
"Lower Body - 3/4 Back Right (may not work sometimes)"
"Torso - Front (Shoulder to Belt)"
"Torso - Back (Shoulder to Belt)"
"Arms - Left Shoulder (Close-Up)"
"Arms - Right Shoulder (Close-Up)"
"Arms - Left Upper Arm"
"Arms - Right Upper Arm"
"Arms - Left Hand (Close-Up)"
"Arms - Right Hand (Close-Up)"
"Legs - Left Thigh (may not work sometimes)"
"Legs - Right Thigh (may not work sometimes)"
"Legs - Left Shin Profile"
"Legs - Right Shin Profile"
"Legs - Left Knee (Close-Up)"
"Legs - Right Knee (Close-Up)"
"Legs - Left Leg (Full)"
"Legs - Right Leg (Full)"
"Legs - Left Foot (Close-Up)"
"Legs - Right Foot (Close-Up)"
"Facial Expression - Neutral"
"Facial Expression - Angry"
"Facial Expression - Smiling"
I hope you enjoy this update and find it useful for your projects. Have fun creating!
Extra Custom Nodes
- RTX Video Super Resolution Pro
- ComfyUI-SpaceGremlin-Toolkit v2.0r3 ***NEW***
Environment & Props Sheet Generator: Currently in development! It will cover full landscape scenes (Exterior + Interior) alongside item/prop turnarounds. Stay tuned!
v1.0 (Compact) / v2.0 (Max) Note :
Key Features:
Lightning Fast & Standalone: Powered exclusively by MiniMax H3—no extra checkpoints, ControlNets, or external models required.
Flexible Export: Automatically composites the final grid sheet OR/AND directly exports individual raw frames so you can compose your own custom layout.
8 Precise Angles & Shots: Full-body (Front, Back, Side profiles), 3/4 Views (Front & Rear), and 3 Headshot Close-ups with distinct facial expressions. (v1.0)
Style Versatility: Fully customizable—seamlessly handles Photorealism, 3D Renders, Manhwa/Anime, Stylized Concept Art, and more.
1-Pass or 2-Pass Latent Upscale: Run it blisteringly fast in a single pass, or toggle the optional Latent Upscale pass for extra texture refinement.
Modular & Clean: Zero canvas clutter. Built-in padding logic for distortion-free alignment.
Feel free to check it out and let me know your thoughts!
Description
v2.0r3
Changes
Fixed an identity issue in version 2.0r2
Added Compact Sheet mode (5 image slots ): fewer cells, but better image quality
Latent scaling model: switched from FP16 to BF16 format (faster, with minimal loss)
Turbo LoRA model: switched from the base Turbo version to a resized version (faster)
All 12 character sheet slots are now available (instead of 11 + 1 reference image)
Enable/disable Turbo LoRA via the global panel
Enable/disable Reference LoRA via the global panel
New custom nodes replacing two other packs
Reduced the "spaghetti" effect in sub-graphics
Improved prompt syntax
Fixed resizing below maximum limits
Fixed an identity issue in version 2.0r2
Change Save Folder Structures :
Final Sheet Folder :
Sheets/Characters/"SubjectName"/"SubjectName"_0000x.png
Individual Images Folder :
Sheets/Characters/"SubjectName"/IndividualImagesFolder/Frame_0000x.png
Average generation time on 5060 Ti 16 GB (varies by reference image)
1 reference picture + Full mode + Latent Upscale + RTX Upscale + saving all pictures (13 pictures): ~350 sec
1 reference picture + Compact Mode + Latent Upscale + RTX Upscale + saving all pictures (6 pictures): ~200 sec
FAQ
Comments (26)
Could this be used to generate a simple character sheet, like, only 4 views (main face, front, side, back)?
Yes, but the workflow needs to be slightly modified.
You can also choose your camera angle and enable the "Export Individual Image" option to obtain a full-resolution image, then compose the sheet using Photoshop or other software.
Available Camera Angles:
"Full Body - Back"
"Full Body - Both Sides"
"Full Body - Left Profile"
"Full Body - Right Profile"
"Upper Body - 3/4 Front Left"
"Upper Body - 3/4 Front Right"
"Upper Body - 3/4 Back Left"
"Upper Body - 3/4 Back Right"
"Lower Body - 3/4 Front Left"
"Lower Body - 3/4 Front Right"
"Lower Body - 3/4 Back Left (may not work sometimes)"
"Lower Body - 3/4 Back Right (may not work sometimes)"
"Torso - Front (Shoulder to Belt)"
"Torso - Back (Shoulder to Belt)"
"Arms - Left Shoulder (Close-Up)"
"Arms - Right Shoulder (Close-Up)"
"Arms - Left Upper Arm"
"Arms - Right Upper Arm"
"Arms - Left Hand (Close-Up)"
"Arms - Right Hand (Close-Up)"
"Legs - Left Thigh (may not work sometimes)"
"Legs - Right Thigh (may not work sometimes)"
"Legs - Left Shin Profile"
"Legs - Right Shin Profile"
"Legs - Left Knee (Close-Up)"
"Legs - Right Knee (Close-Up)"
"Legs - Left Leg (Full)"
"Legs - Right Leg (Full)"
"Legs - Left Foot (Close-Up)"
"Legs - Right Foot (Close-Up)"
"Facial Expression - Neutral"
"Facial Expression - Angry"
"Facial Expression - Smiling"
@SpaceGremlin I see, good 'ol Photoshop never fails when needed, thanks for the info!
@ArttTaku I'm going to consider creating several types of forms for the next version. It's a good idea, anyway. PS: I said Photoshop, but I use Affinity ;)
@SpaceGremlin Thanks for considering it then... I've been learning that for most cases, less is more, so, these character sheets padded with too many images might not be ideal, and most generators can do fine just with 4 images in one single sheet, so, if this could be an option in an update of your workflow, then awesome.
@ArttTaku I have just published the latest version (rev3). It includes the compact sheet mode you requested, featuring five images from the angles of your choice.
@SpaceGremlin awesome, thanks so much!
Awesome WF, thank you so much for sharing it and continuing to develop it.
Thank you so much :)
Huge thanks for this workflow, excellent samples I've done. How do I know if its using RTX VSR or any of the denoise/deblur functions?
Hi, and thank a lot.
There’s no way to know for sure without comparing it yourself, but it works fine on my end—I ran all the tests before release. If you left the image size at 1 MP, disabling supersampling (which doesn’t really add much) together with enabling frame export will get you 2 MP images instead of 1 MP for a full-resolution result. The sheet is limited by the "Largest max size" option.
Been attempting to use this, but I've been having problems with the RTX Denoise/Deblur nodes - I installed the node pack, it's in my custom nodes folder (the node pack code seems very sparse, which surprised me), even ran the requirements text file, but for some reason the nodes won't activate or even appear in the Extensions Manager. What can I do to fix this issue? Did I install it in the wrong place?
From the little digging I've done, the RTX VSR PRO node pack somehow isn't recognizing my CUDA or my Nvidia-VFX install. Do I need to repair my CUDA or Nvidia-VFX somehow? And how would I do that if I do need to do that? What CUDA version do those node packs require?
@Ambrosia_AI
Hello,
Could you please provide me with the model of your graphics card?
I'll take a look on my end, because if even VSR isn't working, you have a problem with your NVIDIA drivers.
@SpaceGremlin NVidia RTX 4070 TI Super. By all rights it should be able to support this. I can see about updating my drivers though; I'm genning using a gaming PC, not a dedicated industrial PC, so my drivers are gaming optimized.
@Ambrosia_AI Try the Studio drivers, not the gaming drivers. I think it can be the problem
@SpaceGremlin Just installed the Studio Drivers, looks like that didn't do the trick - it's still asking for the old Comfy-UI-RTX-Nodes, and it's still claiming it doesn't see RTX Denoise or RTX Deblur despite those being in the custom nodes folder. Error I get for the VSR PRO:
Traceback (most recent call last):
File "F:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\nodes.py", line 2274, in load_custom_node
module_spec.loader.exec_module(module)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^
File "<frozen importlib._bootstrap_external>", line 1023, in exec_module
File "<frozen importlib._bootstrap>", line 488, in callwith_frames_removed
File "F:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\custom_nodes\nvidia-rtx-vsr-pro\__init__.py", line 14, in <module>
import nvvfx
ModuleNotFoundError: No module named 'nvvfx'
@Ambrosia_AI
Sorry,
You are the first person to report this issue; since I am not the creator of these nodes, I cannot help you resolve it. You should report it on the nodes' GitHub repository; they might be able to help or have a solution.
In the meantime, disable the RTX node (bypass: Ctrl+B) and use the workflow without RTX upscaling.
Alternatively, remove RTX VSR Pro from your custom nodes and get the official version directly within Comfy. Both nodes share the same ID, so there is no need to swap them out.
If you encounter the same issues with Deblur and Denoise, bypass both nodes: they don't really offer any significant improvement.
hope it helps.
I have the same problems, the RTX nodes can't be installed, tried alot to fix this, no way.
@Jojo0Gojo Ok, wierd, can you provide me with the model of your graphics card, please ?
Which Series (5000, 4000, 3000) ?
Ambrosia have an 4000 series, I will create a new instance to try this, but I have an 5000 series.
Really love this workflow❤️ but It seems like I'm having some understanding problems with the new workflow. I can't find "Full body - Front" view. It's essential for me because the workflow itself creates a "new style" and I want the character looks the same on every image.
Also I'm wondering where is Shot 1 Frame? My Compact mode shot frame has only 5 views in total.
Hello,
This "Full Body - Front" view is always rendered and generated as Shot 1. That's why you can't find it in the enumeration list. Selecting it from the list is pointless.
When your sheet is generated, locate this view in the "Frames Preview with Index." You'll see that the first images are always "Full Body - Front."
You simply need to adjust the index of the target cell to indicate the index of this view's image in the "Sheet Composer / RTX Suit / Save Frames" window.
So in Compact Mode, Set the views as follows in the "Character Sheet Prompt Maker":
Shot 2 Frame (Compact Mode) -> Full Body - Back
Shot 3 Frame (Compact Mode) -> Full Body - Left Profile
Shot 4 Frame (Compact Mode) -> Full Body - Right Profile
Shot 5 Frame (Compact Mode) -> Any, it will not be used for sheet in this case
Shot 6 Frame (Compact Mode) -> Facial Expression - Neutral
And set indexes in "Sheet Composer / RTX Suit / Save Frames"
Cell 1 Image Index (Compact Mode) at ~120 (Face expression view)
Cell 2 Image Index (Compact Mode) at ~10 <-- This index corresponds to an index in the Full-body front view shots.
Cell 3 Image Index (Compact Mode) at ~31 (Full Body - Back view)
Cell 4 Image Index (Compact Mode) at ~53 (Full Body - Left Side view)
Cell 5 Image Index (Compact Mode) at ~78 (Full Body - Right Side view)
You must have on your final sheet :
- Face expression view
- Full Body - Front view
- Full Body - Back view
- Full Body - Left Side view
- Full Body - Right Side view
I hope this has been helpful.
@SpaceGremlin wow this helped a lot to also understand the Workflow. Its so powerful!!! Great job SpaceGremlin!
@Dannyone Thanks a lot ;)
I see that is quite useful for designers but i can use this also in H3 as a reference?
Yes, the sheet works on Minimax or others reference model