Download
1 variant available
The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.
IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
π¨ RetroForge StyleCaster

A powerful ComfyUI prompt builder designed for Flux2Klein workflows that transforms any input image into stunning stylized 3D renders with modular control.
β οΈ Important Notice
RetroForge StyleCaster is designed for creative stylization and artistic experimentation.
Because of how Flux2Klein models work, extreme 3D stylization and perfect facial likeness cannot be fully achieved at the same time.
The stronger the collectible figurine effect becomes, the more the model naturally reinterprets facial proportions.
Recommended settings for maximum identity preservation
Stylization: LOW
Identity Strength: STRICT
Animated Stylization: OFF
Final results always depend on:
Source image quality
Model version
Prompt
Overall stylization strength
Designed by artists, for artists.
No coding required.
Choose a preset for instant results or unlock complete creative control with 13 modular parameters.
π Quick Start β Presets
β¨ CLEAN_3D
Vibe
Premium studio-quality render with faithful facial likeness.
Perfect for
β’ Portraits
β’ Product shots
β’ Character references
ποΈ VINTAGE_POSTER
Vibe
Aged paper, gentle creases, noir shadows and timeless poster aesthetics.
Perfect for
β’ Movie posters
β’ Album covers
β’ Retro artwork
Default Style
Uses MEDIUM grunge for a balanced vintage appearance.
π§Έ PIXAR_STYLE
Vibe
Glossy resin materials, expressive eyes and warm cinematic lighting.
Perfect for
β’ Family portraits
β’ Pets
β’ Cheerful characters
π CYBERPUNK_NEON
Vibe
Chrome surfaces, magenta/cyan lighting and dense futuristic detail.
Perfect for
β’ Sci-fi characters
β’ Cyberpunk artwork
β’ Technology-themed renders
ποΈ HORROR_NOIR
Vibe
Hyperreal skin, unsettling atmosphere and dramatic shadows.
Perfect for
β’ Horror icons
β’ Dark portraits
β’ Thriller artwork
πΊ CLAYMATION
Vibe
Soft matte clay with a handcrafted stop-motion feel.
Perfect for
β’ Whimsical characters
β’ Handmade aesthetics
β’ Stylized collectibles
π€ PORCELAIN_DOLL
Vibe
Fragile ceramic glaze with cold, elegant lighting.
Perfect for
β’ Fashion artwork
β’ Elegant portraits
β’ Delicate doll aesthetics
π― MINIMALIST
Vibe
Clean silhouettes with strong shapes and negative space.
Perfect for
β’ Logos
β’ Icons
β’ Posters
β’ Graphic design
πΈ HYPERREAL
Vibe
Photorealistic materials with pore-level detail and cinematic lighting.
Perfect for
β’ Maximum realism
β’ High-end renders
β’ Showcase images
βΆοΈ How to Use
Select a preset.
Connect the node to your Flux2Klein sampler.
Generate.
All compatible parameters are configured automatically.
Want more control?
Switch to CUSTOM mode and fine-tune every setting manually.
ποΈ Full Parameter Guide (CUSTOM Mode)
π¨ Core Stylization
stylization
Controls the overall strength of the 3D transformation.
Options
LOW β subtle realism
MEDIUM β balanced stylization
HIGH β collectible figurine appearance with sculpted resin detail
animated_stylization
Adds cartoon influence.
Options
OFF β realistic 3D
LOW
MEDIUM
HIGH β full Pixar / DreamWorks style
expression_protection
Preserves the original facial expression.
Options
OFF
NORMAL
STRICT β prevents beautification and forced facial symmetry
identity_strength
Controls how closely the generated face matches the original.
Options
LOOSE
BALANCED
STRICT β preserves wrinkles, moles and subtle facial characteristics
π¨ Color & Typography
color_fidelity
Controls color accuracy.
Options
LOW
MEDIUM
HIGH β preserves original skin, hair and clothing colors as closely as possible
typography_mode
Controls text handling.
Options
AUTO
PRESERVE β keeps and stylizes existing text
NONE β removes text entirely
πΌοΈ Composition & Background
composition_strength
Controls framing complexity.
Options
LOW β centered composition
MEDIUM
HIGH β dynamic poster layout
background_mode
Controls how the background is processed.
Options
PRESERVE
CLEAN
REINTERPRET
STUDIO β clean gradient studio backdrop
π‘ Lighting & Atmosphere
lighting_mood
Controls the lighting style.
Options
NEUTRAL
DRAMATIC
NOIR β classic venetian-blind shadows
WARM
COLD
NEON β cyberpunk magenta/cyan rim lighting
π§± Surface & Texture
material_quality
Controls the surface material appearance.
Options
DEFAULT
HYPERREAL
MATTE_CLAY
GLOSSY_RESIN
PORCELAIN
METALLIC
π°οΈ Vintage & Detail
grunge_overlay
Adds aging and weathering effects.
Options
OFF
SUBTLE
MEDIUM β gentle creases and vintage texture
HEAVY β scratches and stains
EXTREME β heavily damaged, torn appearance
detail_density
Controls overall detail level.
Options
MINIMAL β focus on silhouettes
BALANCED
DENSE β micro-details throughout the entire image
β Extra
base_prompt
Your primary transformation prompt.
Use this field to customize the core request or improve facial preservation by adding phrases like:
sharp facial features
detailed face
accurate face likeness
extra_positive
Append additional prompts, LoRA triggers or custom instructions without modifying the main prompt.
π‘ Pro Tips
π» Horror
Use:
Animated Stylization β OFF
Expression Protection β STRICT
Pixar-style eyes immediately reduce the horror atmosphere.
ποΈ Vintage Posters
Recommended combination:
Grunge Overlay β MEDIUM
Lighting Mood β NOIR
Background Mode β STUDIO
Increase to HEAVY or EXTREME for distressed poster effects.
π Maximum Identity
Use:
Identity Strength β STRICT
Color Fidelity β HIGH
Expression Protection β STRICT
Then add:
sharp facial features
accurate likeness
to base_prompt.
This improves facial preservation, although it may slightly reduce stylization strength.
π Cyberpunk
For vibrant neon artwork combine:
Lighting Mood β NEON
Material Quality β METALLIC
Color Fidelity β LOW
This allows the model to freely recolor the image into a cyberpunk palette.
π€ Typography
If the source image already contains text:
Use PRESERVE or AUTO.
Choose NONE if you want to remove all text.
βοΈ Identity vs Style
Flux2Klein naturally prioritizes stylization at higher settings.
If facial accuracy is more important than artistic transformation, start with:
Stylization β LOW
Identity Strength β STRICT
Then reinforce the prompt using face-specific keywords inside base_prompt.
π§ Advanced Techniques

π― Multi-Reference Trick for Better Identity Preservation
The included workflow uses two reference images instead of one to achieve stronger identity preservation while maintaining stylized 3D results.
πΌοΈ Reference Image #1
Your original image provides:
Color
Texture
Lighting
Overall appearance
This tells the model what to generate.
βοΈ Reference Image #2
The same image is processed using AIO_Preprocessor before being supplied as a second reference.
This version provides:
Facial structure
Body proportions
Pose
Overall outlines
Unlike ControlNet, this image is not used as a control signal.
Instead, it is passed directly into Flux2Klein's native multi-reference conditioning, allowing the model to blend information from both references simultaneously.
π‘ Why It Works
Flux2Klein attempts to satisfy both conditioning sources at the same time.
The first image describes:
"What should be generated?"
The second image describes:
"How should it be structured?"
The result is often:
Better facial proportions
Stronger pose consistency
Improved identity preservation
Cleaner 3D stylization
...all without relying on ControlNet.
π¨ Choosing the Right Preprocessor
βοΈ Lineart
Best for
Portraits
Faces
Clear outlines
Why use it
Provides the strongest facial structure preservation and is the recommended starting point for most portraits.
π Canny
Best for
Hard edges
Architecture
Mechanical objects
Why use it
Excellent for preserving sharp contours and geometric detail.
π Depth
Best for
Complex scenes
Interior spaces
Environmental renders
Why use it
Adds spatial awareness and improves perspective consistency.
π§± Normal Map
Best for
Surface relief
Material detail
Mechanical assets
Why use it
Helps preserve surface topology and fine material information.
β Recommended Starting Points
For most projects:
Portraits β Lineart
Objects & Architecture β Canny
Complex Scenes β Depth
Different source images respond differently, so don't hesitate to experiment.
βοΈ Recommended Settings
For the best balance between identity preservation and stylization:
Preset: CLEAN_3D (or any preset you prefer)
Stylization: LOW
Identity Strength: STRICT
Expression Protection: STRICT
Animated Stylization: OFF
β οΈ Technical Note
This workflow takes advantage of an undocumented behavior in Flux2Klein's native multi-reference blending.
Because this behavior is not officially documented:
Results may vary between model versions.
Different preprocessors may produce different outcomes.
Future updates could change how this behaves.
Treat it as an advanced optimization technique rather than a guaranteed feature.
π§ Recommended Companion Nodes
RetroForge StyleCaster works best as part of a larger workflow.
Here are two companion nodes that pair exceptionally well with it.
π ZImageResolutionScaler
π https://civitai.red/models/2344436/z-image-resolution-picker
Automatically resizes images to resolutions that work best with Flux2Klein models.
Why use it?
Flux models are surprisingly sensitive to input resolution.
Feeding very large images (such as 4K photos) often introduces facial artifacts or inconsistent stylization.
ZImageResolutionScaler automatically finds a more suitable working resolution before generation begins.
π€ Simple Text Overlay Node
π https://civitai.red/models/2711044/simple-text-overlay-node
Adds crisp text overlays after rendering.
Perfect for:
Movie titles
Logos
Captions
Credits
Poster typography
Why use it?
StyleCaster can preserve existing text reasonably well, but generating new readable text is still unreliable.
Adding typography as a post-process produces much cleaner and more professional results.
π§ Installation
Installation takes only a few seconds.
Extract the RetroForge-StyleCaster folder.
Copy it into:
ComfyUI/custom_nodes/
Restart ComfyUI.
Find the node under:
prompt β builders
You're ready to create.
β οΈ Limitations & Notes
3D Stylization vs Facial Accuracy
This is a limitation of Flux2Klein modelsβnot a bug in StyleCaster.
As stylization increases, the model naturally begins to reinterpret facial anatomy.
Higher stylization creates stronger collectible and animated aesthetics, while lower stylization preserves facial identity more faithfully.
Finding the right balance between style and likeness is part of the creative process.

