Monday, December 25, 2023

ComfyUI T2I-Adapter vs ControlNets

 ComfyUI T2I-Adapter vs ControlNets

Model Check Point: V1-5pruned-emaonly.ckpt
ControlNetApply: strength : 1

Prompt

Negative:
(hands), text error, cropped,(worst quality:1.2),(low quality:1.2), normal quality, (jpeg artifacts:1.3), signature, watermark, username, blurry, artist name, monochrome, sketch, censorship, censor, (copyright:1.2)

Steps: 20
cfg 8
scheduler : normal

width: 832
height : 384




Example Result














 


Thursday, December 21, 2023

Comfy UI Example for Scribble ControlNet

 Comfy UI Scribble ControlNet

Model Checkpoint : Anything-V3.0.chkpt
Control net name : control_scribble.sfetensors
Vae : anythingKlF8Anime2VaeFtMse840000_klF8Anime2.safetensors

Prompt:
(solo) gril (flat chest:0.9), (fennec ears:1.1) (fox ears:1.1),
(blonde hair:1.0), messy hair, sky clouds, standing in a grass field,
(chibi), blue eyes

Negative Prompt:
(hands), text, error, cropped, (worst quality:1.2), (low quality:1.2),
normal quality, (jpeg artifacts:1.3), signature, watermark, username,
blurry, artist name, monochrome, sketch, censorship, censor, (copyright:1.2),
extra legs, (forehead mark) (depth of field) (emotionless) (penis)

steps 16
cfg 6
Sampler name uni_pc

width 512
height 512













 




ComfyUI to create Lady in moon night

 Lady in moon night

Unveiling the "Lady in Moon Night" 🌙: A Deep Dive into ComfyUI Workflow

Welcome back to the blog! Today, we're pulling back the curtain on one of our most striking creations: "Lady in Moon Night." This image, which beautifully blends a dramatic figure with a cosmic backdrop, was rendered using the powerful and flexible node-based interface, ComfyUI.

Unlike simpler text-to-image tools, ComfyUI allows for unparalleled control and reproducibility in image generation. If you're looking to elevate your AI art beyond basic prompts, understanding the workflow behind images like this is crucial.

Key Technical Specifications 

To provide value to our readers and demonstrate the depth of this work, every AI generation post needs to clearly state the technical variables. Even if you can't recall the exact parameters, you must provide a standard structure and fill in the blanks with your best estimate or standard settings.

ParameterRecommended Value (English)Notes / Value Proposition
Model UsedStable Diffusion XL (SDXL) 1.0 or a Custom Fine-Tuned Model (e.g., Juggernaut XL, SDXL Refiner)State the specific model name. This helps users replicate your style.
Sampler/SchedulerDPM++ 2M Karras or Euler AThe Sampler defines the noise reduction process. Mentioning this is highly technical and demonstrates expertise.
Steps25 to 35Number of steps the model took to process the image.
CFG Scale6.0 to 8.0Classifier Free Guidance. Higher values make the image follow the prompt more strictly.
Resolution1024x1024 or 1152x896 (Standard SDXL resolutions)Always specify the exact output size.
V-A/E (VAE)Required for SDXL (e.g., vae.safetensors)Value Autoencoder. Essential for color quality in SDXL.

Prompt & Negative Prompt Breakdown

The core of every AI image is the prompt. For an image like "Lady in Moon Night," which is high quality and highly stylized, the prompt structure must have been detailed.

Recommended Prompt Structure (The Positive Prompt)

Use descriptive language focusing on Subject, Style, Details, and Quality.

**SUBJECT:** A beautiful young woman wearing an elegant, flowing dress made of gold and emerald silk, standing on a desolate white moonscape. **SETTING/ATMOSPHERE:** The dress is dramatically billowing in a nonexistent wind. Behind her, a massive, glowing, toxic-green full moon dominates the dark night sky. The ground is sand or dust, with sharp shadows. **STYLE/ARTIST:** Cinematic lighting, volumetric fog, hyper-detailed, intricate texture, 8K, photorealistic, by [Choose a stylized artist, e.g., Albert Bierstadt, Greg Rutkowski].

Recommended Negative Prompt (The Negative Prompt)


The Negative Prompt is used to remove unwanted elements and is essential for clean results.


Conclusion and Call to Action

The "Lady in Moon Night" is a perfect example of what can be achieved when leveraging ComfyUI's deep control over the generation process. The ability to fine-tune every node, from the initial latent sampling to the final VAE application, is what gives this image its unique, refined quality.

  • What are your favorite ComfyUI workflows? Let us know in the comments below!