Hello there and thanks for checking out the Notorious Secret Fantasy Workflow!
(Compatible with : SDXL/Pony/SD15)
—Purpose—
This workflow makes use of advanced masking procedures to leverage ComfyUI's capabilities to realize simple concepts that prompts alone would barely be able to make happen.
Wrapped up into a tidy pre-configured tool to have fun with!
—Versions—
—v1—
The initial version starts out with the keyhole peeping concept!
I will turn this into a series by adding more concepts.
Potentially even built into the same workflow—depending on how well I can integrate further concepts and compress them into tidy little group nodes—to conveniently switch concepts on the fly.
Upcoming concepts:
changing booth curtain peek
spycam
...
Feel free to suggest ideas for further concepts in comments!
—Features—
2 pass subject generation with optional upscale : txt2img2img
1 pass random keyhole generation
→ leveraging advanced mask processing (no resource download required!)
LoRA support
Teaser!
As usual the workflow is accompanied by notes explaining nodes used and their settings, personal recommendations and observations.
—Custom Nodes—
All of which can be installed through the ComfyUI-Manager
If you encounter any nodes showing up red (failing to load), you can install the corresponding custom node packs in most cases through the 'Install Missing Custom Nodes' tab on the ComfyUI Manager as well.
Detailed instructions and links to recommended models, along with the folders they need to be placed in, are included.
—Thanks—
The workflow would not be possible as is without these custom node packs. If you want to support the custom node creators, give them a ⭐ on their github repos! Thank you!
Feel free to ask any questions, share improvements or suggestions in the comment section!
Also let me know if you encounter any confusing points I can elaborate on and focus on improving for the next update!
Description
v1 — initial release
→ Keyhole Peeping Concept
FAQ
Comments (7)
Soo.. Basically it's an advertisement for comfy ui manager, that otherwise, puts a keyhole layer over your images, which could be done in any image editor in under a minute?
It generates both a keyhole guided by the approximate shape of one, while also generating an image of choice before recombining the both.
Doing that in one generation is something several LoRAs attempted but never managed to get right, probably due to how describing a character that is potentially not fully visible works against the nature of captioning being all that is actually seen. As well as the extreme depth difference.
I am going to update this workflow soon and expand it with more choices and a really nice feature.
Yea, I'm not trying to be an ass. I'm just sayin, a person could go into Gimp, paint tool sai, or Photoshop(if they had a paid PS sub), and add a keyhole layer over their image generation. It would literally take a person under 30 seconds if they had the app open already. I'm not saying that what you've done isn't a bit impressive. Just that it seems like a lot of effort for a simple concept.
Though I guess it would make things simpler for imbeciles (I'm not including you in that), but well, I care a bit less about being dicks to them, lol.. It's like, take a minute, learn a new concept, it's not that hard.
I just wish there was someone as determined as you working on male-oriented Loras. Try to get a guy that's not pissing outta His vagina.. It's NOT easy..
@Lazman Sure you could do that, but you would need to get a whole lot of keyholes and images to lay them over to get any kind of surprise effect, right? And part of the fun in this workflow is not even knowing for sure whether you even get to see anything. I think that adds a nice touch of curiousity to the process. When you enter a prompt you usually know what you are going to get, at least approximately, wildcard randomization aside, but with a concept like this, there is an additional element of anticipation involved.
And the way biasing a latent with a mask works has a lot of potential I find, so I want more people to know about that and explore that direction more, as it also allows complex compositions without need for a controlnet to guide generation.
The feature I am going to add is also taking this to a whole nother level, which should make this a lot more fun I think.
I thought there are a fair amount of male-oriented creators on here. At least I see a lot each day. There are even male oriented checkpoints. One based on pony, but in a realistic direction, called something with "stallion" in the name, I believe.
Though I would not recommend the keyhole part with a pony model. They are so character focused, that a simple textured surface can be a real pain.
Latest comfyui version is giving this error: "can't access property "map", this.node.inputs is undefined"
Hey there, yeah this workflow is pretty outdated. I'm in the process of recreating it entirely as previous experimental versions I planned for updates no longer load on my end.
It will take a bit, but will be far better than it was 👌
@RedPinkRetro awesome!