Check earlier versions for more examples!
Runtime profile
Source description
Check earlier versions for more examples!
Development Status: I am using this daily and taking notes for future improvements. Next version will come in the form of a custom node for comfyui.
This will improve computation speed by probably 1000% 😁
I want to add many more options, including Natural language based instead of just tags.
Like my stuff? Want to use a powerful GPU? You can send me some love by signing up to RunPod using my referral link: https://runpod.io?ref=slmzn8qv
WHY a Random Prompt Generator? So the other day I was thinking of a way to automate my prompts so that I could go somewhere come back and have a load of images to choose from. I've seen quite a few implementations using different approaches but none where very pleasing as they where simply loading pre made prompts and this would lead to repetition.
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
MPI Random Prompt Gen is a ComfyUI workflow for generating varied image prompts by randomly combining editable character and scene lists with details entered in Pre and style or camera terms in Post. It provides a positive prompt output and a negative post-prompt output.
It can randomly select clothing, makeup, expressions, hairstyles, locations, colors, texture patterns, lighting, effects, and ambience.
Optional General and Character list nodes let you change scene and character choices separately; without those connections, it uses its included lists.
Connect the positive prompt output to a positive text encoder; the negative post-prompt has its own output.
A Save Image node is included to save prompts.
Copy and paste the node into an existing ComfyUI workflow.
Open the subgraph by double-clicking it or using its top-right icon to edit the lists.
Listed node packs include comfyui-kjnodes, rgthree-comfy, comfyui-logic, ComfyUI-Jjk-Nodes, ComfyMath, comfyui-impact-pack, comfyui-easy-use, and was-ns.
Other listed node packs are comfyui-custom-scripts, comfyui-art-venture, comfyui_ultimatesdupscale, comfyui_stringessentials, controlaltai-nodes, comfyui-impact-subpack, and kaytool.
Other named files include 4x_NMKD-Siax_200k.pth, bbox/face_yolov8m.pt, bbox/hand_yolov8s.pt, lustifySDXLNSFW_oltONELASTTIME.safetensors, sam_vit_b_01ec64.pth, and sd_xl_base_1.0.safetensors.
Most controls use probability sliders from 0 (off) to 100 (always on).
Separate seeds let you vary character and environment independently; SFW and NSFW probability sliders affect action lists and clothing choices.
Before setup, verify which ComfyUI packages are meant by ComfyMath and was-ns.
Suggestion · not verified
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
1 excerptSource context: 820 downloads · Type Workflows · Base model Other
So I though what if I just give it a list of locations, leg wear, top wear, colours, texture patterns, light types and effects, ambient?
And so here it is the solution .
TIP: Set your KSampler seed to random , click the arrow next to RUN and select " instant ", press RUN and go for a walk or to work. Come back to hundreds of images to chose from 😉
Using the new ComfyUI Subgraph feature I built a system where it will randomly pick clothes, makeup, expression, etc from a list.
The above image is the provided workflow.
1 - I added a Save Image Node so that you can save your prompts as well as all is randomly generated you wont get the same prompt when you reload a image into comfy.
2 - The tool was built with the intention of generating pictures for a model , so in the top text box where it says Pre, that's where you would write your model's identity, like trigger word, hair colour, etc. For the images I am sharing here, I simply used "1 woman".
3 - The buttons will turn on/off generation for what they mean. SFW and NSFW will will trigger different lists of actions and also be used for logic of what to wear in different settings.
4 - The box at the end where it says Post is whatever you want to go after the generation. Typically this is where you can write the style of the composition, like anime, realistic, and cameras, Cannon, phone, etc
5 - To use simple copy and paste the node into an existing workflow and connect the prompt output to a text encoder positive.
Enter the subgraph to edit the lists.
You can edit this lists by entering the Subgraph clicking twice on it or by clicking on the top right icon.
Once inside the subgraph look for the section depicted in the image above .
Its all the lists you can alter.
If you feel adventurous check the code and change that to your will 😉
All the images I am showing here where done with all the buttons ON and "1 Woman" in the pre prompt, for some models I turned off SFW and clothes to be able to generate NSFW content, Krea wasn't having it though 🤣🤣
Send me some love and I'll release other versions, hope you like and use it and ask/suggest away, I'll answer when I can.
V1.0 added seed control for recall!
V1.1 bug fix: solved issue of regenerating even when the seed was the same!
V1.2
I noticed that for V1.1 the colour variation has stopped, this is due to the seed being locked and the same for all nodes, in version 1.2 I added colour and pattern variation sliders.
V2.0
Added Hair styles and 2 separate ( Optional ) nodes that separate the lists from the main subgraph and are broken up into 2 main categories, General and Character .
These 2 nodes where created so you can change characters and environment by simply plugin them into the main subgraph. Each character may have different styles of clothes, hair, colours, etc so I though it could be handy to have the system broken up.
Although these nodes are optional, if not plugged in, it will default to the lists it contains.
Here are a few examples on how you could use them, but these are just copies of the other 2.
Simple copy/paste or Alt+Drag a lists node, enter the subgraph and change it to your needs.
As keep using it on a daily basis I will probably release more versions but for now I think we good 😉
V2.1
Bug fix: Fixed a node that was giving error in the console.
Removed the lists from the main subgraph, the subgraph is now dependant of them.
Created a node with the combined lists for easy access.
V3.0
About this version
Changed the colour of the groups to no colour so it wouldn't interfere with workflows that have groups bypasser/muter based on colours
Added Bra types (And a node to show how to update your existing lists)
Changed SFW and NSFW to a probability slider
Changed most buttons to a slider that controls probability (0 = off, 100 = always on)
Added my SDXL tests workflow so you can see it in action
Added a negative post prompt and its output
The tests workflow has a detailer, high-res fix, face detailer and hand detailer, so you can use it for generations, maybe add an upscale at the end.
V3.1
Fixed clothed <> Lingerie Slider has it was inverted
Fixed Bra input
Added separate seeds for character and environment to be able to change one or the other
Remaining hardware or download-size estimates are not available yet.
Requirements
Upscaler · 63.9 MB · PT · Unknown
Object Detector · Hugging Face · bbox/face_yolov8m.pt
Object Detector · Hugging Face · bbox/hand_yolov8s.pt
comfyui_ultimatesdupscale
File unverifiedUpscaler · Registry
lustifySDXLNSFW_oltONELASTTIME.safetensors
Not resolvedCheckpoint · Unknown
Segmentation · 358 MB · PTH · Unknown
Checkpoint · 6.46 GB · SAFETENSORS · fp16 · pruned · Hugging Face · stabilityai/stable-diffusion-xl-base-1.0
ComfyMath
PossibleNode pack · Registry
comfyui_stringessentials
FoundNode pack · Registry
Node pack · Registry
comfyui-custom-scripts
FoundNode pack · Registry
comfyui-easy-use
FoundNode pack · Registry
comfyui-impact-pack
FoundNode pack · Registry
comfyui-impact-subpack
FoundNode pack · Registry
ComfyUI-Jjk-Nodes
FoundNode pack · Registry
Node pack · Registry
Node pack · Registry
controlaltai-nodes
FoundNode pack · Registry
kaytool
FoundNode pack · Registry
rgthree-comfy
FoundNode pack · Registry
was-ns
Not resolvedNode pack · Registry