Product Mockups with REDMOND LoRA for FLUX.2-Klein-9B

TL;DR with 3 actionable bullets:

• Use the trigger word Mockup. in your prompt to activate the product mockup style.
• Load mockups_flux_klein_9b.safetensors into ComfyUI alongside the FLUX.2-Klein-9B base model at a strength of ~0.8–1.0.
• Generate 1024x1024 images for best fit with the LoRA's training resolution.

What This LoRA Does

MOCKUPS REDMOND FLUXKLEIN is a LoRA (Low-Rank Adaptation) trained on a dataset of product mockup images. It adapts the black-forest-labs/FLUX.2-klein-9B base model to produce staged product scenes — devices, packaging, and consumer goods placed in clean, minimally-styled environments with consistent lighting.

The model was trained with GPU time donated by Redmond.AI, and the weights are distributed as a single Safetensors file under the Apache 2.0 license, making it usable for both commercial and personal projects.

Sample output from MOCKUPS REDMOND FLUXKLEIN. The lighting and staging create a consistent product photography aesthetic.

Trigger Words and Usage

The essential trigger word is Mockup. — placing it at the start of your prompt activates the learned product mockup style. For example:

Mockup. A smartphone on a marble surface, soft natural lighting, high-end product photography

How to Use in ComfyUI

1. Download the Safetensors file from the model's Hugging Face page.
2. Place the file in your ComfyUI/models/loras/ directory.
3. Load the FLUX.2-Klein-9B base model in a CheckpointLoader node.
4. Add a LoraLoader node, select the MOCKUPS REDMOND LoRA, and set the strength between 0.6 and 1.0 depending on how pronounced you want the effect.
5. Connect your prompt to a CLIPTextEncode node and generate at 1024x1024.

Recommended Settings

Setting | Value

|---------|-------

Base model | black-forest-labs/FLUX.2-Klein-9B Trigger word | Mockup. Resolution | 1024x1024 LoRA strength | 0.6–1.0 Format | Safetensors License | Apache 2.0 The LoRA is lightweight — a single Safetensors file with a model strength of 1.0 works well for strong mockup styling, while lower strengths allow more of the base model's creativity through.

Using with Diffusers

If you are using the 🤗 Diffusers library, you can load the LoRA adapter via the LoraLoader:

from diffusers import DiffusionPipeline, FluxTransformer2DModel
import torch

pipe = DiffusionPipeline.from_pretrained(
    "black-forest-labs/FLUX.2-klein-9B",
    torch_dtype=torch.float16,
)
pipe.load_lora_weights(
    "artificialguybr/MOCKUPS-REDMOND-FLUXKLEIN",
    weight_name="mockups_flux_klein_9b",
)
pipe.set_adapters(["mockups_flux_klein_9b"], adapter_weights=[1.0])

prompt = "Mockup. A smart speaker on a wooden desk, soft afternoon light, product photography"
image = pipe(prompt, height=1024, width=1024).images[0]
image.save("mockup_output.png")

Training Background

The LoRA was trained on a curated set of product mockup images, resulting in consistent output for staged product scenes. The training was supported by Redmond.AI, which provided the GPU compute needed for fine-tuning.

Common Errors and Fixes

Error | Cause | Fix

|---|-----|---

Model not triggering | Missing the Mockup. trigger word | Always prefix your prompt with Mockup. Images look unfocused | LoRA strength too low | Increase strength to 0.8 or higher Wrong resolution | Generating at non-standard sizes | Use 1024x1024 for best results File not loading in ComfyUI | Incompatible format | Ensure you downloaded the .safetensors file

FAQ

Can I use this for commercial projects? Yes, the model is licensed under Apache 2.0, which permits commercial use without restrictions.

Do I need the full FLUX.2-Klein-9B model? Yes, this LoRA is an adapter and requires the base model weights to function.

Is there a Civitai page for this model? At the time of writing, this model is hosted on Hugging Face only.

Sources

• Hugging Face model page — Official model card, weights, and sample images.
• FLUX.2-Klein-9B base model — Underlying base model the LoRA adapts.
• ComfyUI — Recommended interface for loading and generating with this LoRA.