FLUX.2 Klein 9B Consistency LoRA · I2I For AD Film

ABOUT THE WORKFLOW

Edit an Image
Upload a photo, describe the change you want, and get an edited version. FLUX.2 Klein 9B uses the input image as a reference latent, making targeted changes while preserving the structure. The Consistency LoRA tightens adherence to the original so the result stays visually close. The output shows a side-by-side comparison automatically. That's it.

Model

• FLUX.2 Klein 9B by Black Forest Labs. A 9B parameter rectified flow transformer (FLUX Non-Commercial License) that unifies text-to-image generation and single-reference editing in one architecture. Matches or exceeds models 5x its size. Paired with a Consistency LoRA (strength 0.8) that keeps the edited result closer to the original image.

Note: The workflow's embedded sticky note labels this model Apache 2.0. That applies to the 4B Klein variant. The 9B model used here falls under the FLUX Non-Commercial License. Review Black Forest Labs' terms before using outputs commercially.

HOW IT WORKS

Step 1. Upload your image
The photo you want to edit. The workflow auto-scales it to 1 megapixel, preserving its aspect ratio.
Works great with: portraits · product shots · scenes · design mockups

Step 2. Describe the change
Write a specific edit instruction. "Replace the man's clothes with a black formal suit, replace background with a luxurious dark bakery" works better than "make it look different." No trigger word needed.

Step 3. Hit run and download
FLUX.2 Klein 9B applies the edit and returns a side-by-side comparison of the original and the result. Preview it in the workflow, then download.
Ready for: Photoshop · Figma · Canva · any editor

First time? Leave every setting as-is. The defaults (1 megapixel · 8 steps · 1 guidance · 0.8 LoRA strength) are the right starting point for almost everyone.

RECOMMENDED SETTINGS

Quick-start guide. Find the goal that matches yours and copy the settings.

• Standard edit (most people) — 1 megapixel · 8 steps · 1 guidance · 0.8 LoRA strength · random seed. The right starting point for almost everyone.

• Want the edit to stay closer to the original — Raise LoRA strength toward 1.0. Higher values make the Consistency LoRA grip tighter, preserving more of the original structure and color.

• Want more creative freedom — Lower LoRA strength toward 0.5 or below. The model will interpret your instruction more freely and deviate further from the source.

• Want sharper detail — Raise steps from 8 to 12 or 16. More passes add detail but take longer. Keep guidance at 1.

• Want to compare variations — Leave the seed on random. Each run produces a different interpretation of the same edit instruction.

• The edit is too subtle — Make the instruction more direct. "Replace the jacket with a red leather jacket" is clearer than "change the jacket." If the Consistency LoRA is holding back the edit, lower its strength.

• The edit changes too much — Raise LoRA strength and be specific about what to preserve. "Replace only the background with a beach at sunset, keep the person and clothing exactly as they are" anchors the model.

Prompt: Describe the change and what to preserve. Be specific about both. "Replace the man's clothes with a black formal suit, keep the face and pose" works better than "formal outfit." Start with "realistic" if you want photorealistic output.

LEARN

📹 Videos

• Intro to Floyo

• ComfyUI 101 Free Course ft. Sebastian Kamph

• Floyo 101 for Team Collaboration

✨ Quick links

• Help docs

• LoRA Training guide

• Discover more workflows

USE CASES

👔 Wardrobe and Outfit Swaps
Change clothing, accessories, or styling on a subject while keeping the face, pose, and body structure intact. Useful for fashion previews and virtual try-on without reshooting.

🖼️ Background Replacement
Replace or transform the background of a portrait or product shot with a specific setting. The model preserves the subject while generating a new environment around it.

🎨 Style and Material Changes
Swap textures, materials, or finishes on objects in the image. Turn a wooden desk to marble, change a matte wall to exposed brick, or restyle a room without a new render.

🛍️ Product Shot Adjustments
Edit product images for different campaigns, seasons, or colorways. Adjust packaging, labels, or surroundings from a single source photo.

WHAT WORKS BEST / WHAT TO AVOID

✅ Works great

• Specific, targeted edit instructions

• Single subjects on clear backgrounds

• Wardrobe, background, and material swaps

• Photos with good lighting and clean composition

⚠️ May produce softer results

• Vague instructions like "make it better" or "improve it"

• Edits that require generating entirely new subjects

• Crowded scenes with many overlapping elements

• Very low-resolution or heavily compressed source images

FAQ

What is FLUX.2 Klein 9B?
FLUX.2 Klein 9B is a 9B parameter rectified flow transformer by Black Forest Labs. It unifies text-to-image generation and image editing in a single architecture, using the input image as a reference latent to make targeted changes while preserving structure. It is step-distilled with a Qwen3 8B text embedder and matches or exceeds models 5x its size in quality benchmarks.

What does the Consistency LoRA do?
The Consistency LoRA tightens the model's adherence to the original image during editing. At the default strength of 0.8, it preserves color palette, composition, and structural details while still allowing the edit instruction to take effect. Raise it for closer results, lower it for more creative deviation.

Can FLUX.2 Klein 9B do text-to-image as well as editing?
Yes. Klein 9B unifies generation and editing in one model. This workflow is set up for image editing with a reference latent, but the same model supports text-to-image generation and multi-reference editing (combining up to 5 input images) in other configurations.

What license does FLUX.2 Klein 9B use?
The 9B model falls under the FLUX Non-Commercial License. It is open-weights (you can download and run it locally), but commercial use requires a separate arrangement with Black Forest Labs. The 4B Klein variant is Apache 2.0 and fully open for commercial use.

How is FLUX.2 Klein different from FLUX.1 or FLUX.2 Kontext?
FLUX.2 Klein is the compact, fast model in the FLUX.2 family. It focuses on sub-second inference, unified generation and editing, and consumer GPU compatibility (~29GB VRAM). FLUX.2 Kontext is the larger, higher-quality editing model. Klein trades some editing precision for speed and accessibility.

Why does the output show a side-by-side comparison?
The workflow composites the original image and the edited result into one side-by-side output so you can compare them instantly. Download the output and crop to the edited half if you need the result alone.

How to run FLUX.2 Klein 9B editing online?
You can run FLUX.2 Klein 9B editing online through Floyo. No installation, no setup, no API key to wire up. Open the workflow in your browser, upload an image, describe the edit, and hit run. Free to try.

WHY FLOYO?

Floyo is the only platform with team collaboration for ComfyUI in the browser. You run workflows with no install. You share run history, assets, and models across your team. You pay only when you generate. Floyo supports open-source and closed-source models.

A designer runs an edit and likes the result. A teammate opens that exact run from shared history and keeps going. No file handoffs. No version confusion.

For studios and enterprise teams, Floyo adds private workspaces, pooled resources, and a team usage dashboard. Other ComfyUI cloud tools run for one person at a time. Floyo runs for the whole team, with transparent per-generation costs.

Ready to try it?
Upload an image, describe the change, and run it. The settings are already set.

→ Launch Workflow, Free

Questions? Watch the free course or check the FAQ above.