ArtificialGuyBR

Home / Blog / Training LoRAs

Product Mockups REDMOND LoRA for Flux, Qwen, ZImage

Train a product-mockup LoRA on Flux.2 Klein 9B, Qwen, and ZImageTurbo. Trigger word Mockup., ComfyUI setup, Apache 2.0 license.

1 sources cited Training LoRAs

On this page

  1. What This LoRA Does
  2. Trigger Words and Usage
  3. Sample Output
  4. Recommended Settings
  5. License
  6. Common Errors and Fixes
  7. FAQ
  8. Sources

Product Mockups REDMOND LoRA: Turn Any Base Model Into a Product Photography Machine

TL;DR —

What This LoRA Does

Product Mockups REDMOND LoRA is a concept LoRA trained on a curated dataset of product mockup imagery — the clean, high-detail shots you see on e-commerce sites, app showcases, and design portfolios. The training, made possible by GPU time donated by Redmond.Ai, teaches the base model to emit product-mockup aesthetics: crisp lighting, consistent surface reflections, controlled depth of field, and neutral-to-matte backgrounds that put the product forward.

The model supports three base backbones:

  1. Flux.2 Klein 9B — distilled Flux variant, fast and efficient.
  2. Qwen / Qwen Image — Qwen's image generation pipeline.
  3. ZImageTurbo / Z-Image — a turbo-optimized base for speed.

This means you can drop the same LoRA into three different pipelines and get style-coherent output in each.

Trigger Words and Usage

The single trigger word is:

Mockup.

Including Mockup. at the start of your prompt (ideally as the first token sequence) tells the LoRA network to bias activation toward the learned product-mockup manifold. Omit it and the effect drops off sharply.

How to Load in ComfyUI

  1. Download the .safetensors weight file from the Civitai page.
  2. Place it in ComfyUI/models/loras/.
  3. In your workflow, add a LoraLoader (or LoraLoaderAdvanced) node.
  4. Select product-mockup-redmond-lora from the dropdown.
  5. Start with a strength of 0.7 and adjust upward for more pronounced styling or downward to blend with the base model's native character.
  6. Prepend Mockup. to your prompt.

Sample Output

Product mockup style sample 1

Product mockup style sample 2

Product mockup style sample 3

The samples show the style's signature traits: clean white or gradient backgrounds, consistent lighting direction, and surfaces that read as professional product photography rather than stylized illustration.

Setting | Value

|---|---

Trigger word | Mockup. (first in prompt) Strength | 0.6 – 0.8 (start at 0.7) Sampler | DPM++ 2M Karras, Euler a, or base model default Steps | 20 – 30 Base model | Flux.2 Klein 9B, Qwen Image, or ZImageTurbo

License

This model is released under the Apache License 2.0. You may use it for commercial projects, resale artwork, and derivative works without additional permission. No attribution is required, though crediting the author is appreciated.

Common Errors and Fixes

Error | Cause | Fix

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

Style not appearing | Trigger word missing or misspelled | Ensure Mockup. is at the start of the prompt Over-saturated / artificial look | Strength too high | Reduce to 0.5–0.6 and increase steps Wrong base model | Using SDXL or SD 1.5 weights | Use the Flux.2 Klein 9B, Qwen, or ZImageTurbo variant only LoRA not loading in ComfyUI | File placed in wrong folder | Put .safetensors in models/loras/ and restart ComfyUI

FAQ

Q: Can I use this for print-on-demand products? A: Yes. Apache 2.0 permits commercial use including POD, merchandise, and resale.

Q: What resolution should I generate at? A: 1024×1024 is the native training resolution for most base models. Upscaling afterward is recommended for large-format mockups.

Q: How do I combine it with other LoRAs? :A: Stack the LoraLoader nodes in series. Keep the combined strength under 1.0 total to avoid over-borrowing artifacts.

Sources