Consistent Face 3x3 Generator

What is Consistent-Face-3x3-Generator?

Consistent-Face-3x3-Generator is a structured face generation workflow designed to produce a 3×3 grid of the same character rendered across nine different poses. The Consistent-Face-3x3-Generator workflow uses the FLUX base model along with the Flux Depth LoRA to ensure facial consistency while enabling expressive pose variety.

This Consistent-Face-3x3-Generator workflow is perfect for creating pose reference sheets, dataset generation, or consistency testing for stylized characters.

Why Use Consistent-Face-3x3-Generator?

Consistent-Face-3x3-Generator offers:

• Consistent Identity: Preserves the character's face across 9 outputs in the Consistent-Face-3x3-Generator grid
• Pose Variation: Each cell in the Consistent-Face-3x3-Generator's 3×3 grid shows a unique head pose
• Depth-Aware Guidance: Uses Flux Depth LoRA for 3D-aware consistency
• All-in-One Grid Output: Consistent-Face-3x3-Generator outputs a unified 3×3 visual presentation
• Ideal for Stylized Workflows: Perfect for anime, toon, and concept character pipelines

Whether you're generating stylized portraits or testing identity consistency, the Consistent-Face-3x3-Generator is optimized for clarity, control, and batch visual output.

Base

This group in Consistent-Face-3x3-Generator loads all the core models and handles initial generation:

• Load the FLUX model and Flux Depth LoRA for Consistent-Face-3x3-Generator
• Enter your main prompt describing the character and style
• Set CFG, Steps, Sampler, and other KSampler parameters
• The Consistent-Face-3x3-Generator result is a single 3×3 base image containing all 9 variations

This is your starting grid for further upscaling and detailing in the Consistent-Face-3x3-Generator workflow.

SD Upscaler

This stage in Consistent-Face-3x3-Generator uses Stable Diffusion Upscale to enhance each image in the grid:

• The Consistent-Face-3x3-Generator's 3×3 grid is split internally
• Each image is processed one by one for cleaner high-res output
• Helps preserve pose clarity while boosting resolution

Great for preparing final Consistent-Face-3x3-Generator outputs suitable for publishing or design references.

Splitter Group and Refiner

This Consistent-Face-3x3-Generator group:

• Splits the single grid image into 9 separate face crops
• Sends each image into a refiner model to clean up details
• Keeps the identity consistent while enhancing features

The Consistent-Face-3x3-Generator prepares each portrait for high-detail enhancement in the next step.

Adetailer

In the Consistent-Face-3x3-Generator workflow:

• Applies face-specific detailing to each cropped image
• Enhances skin texture, eye sharpness, and facial structure
• Ideal for stylized or photorealistic finishing based on prompt

The Consistent-Face-3x3-Generator's Adetailer is aimed at improving the facial realism and expression clarity across poses.

Chin Detailer + Output

Final touch-up and output for Consistent-Face-3x3-Generator:

• Fixes and adjusts chin/jaw structure if needed
• Saves final upscaled and refined portraits from Consistent-Face-3x3-Generator to your local directory
• Output path: ComfyUI/output/YYYY-MM-DD/

Each image in the Consistent-Face-3x3-Generator set is saved individually for easy access and review.

Acknowledgement

The Consistent-Face-3x3-Generator workflow was developed by tenofas and is based on the FLUX image generation model and its Flux Depth LoRA. Full credit goes to tenofas for designing a reference-consistent, multi-pose portrait generator inside ComfyUI.
Original Workflow:

The Consistent-Face-3x3-Generator workflow makes identity-stable image generation easier and more modular, letting artists and developers explore structured visual variation at scale.