ArtificialGuyBR

Home / Blog / Training LoRAs

Film Grain REDMOND LoRA: Cinematic Texture for Qwen Image

Add authentic film grain and cinematic atmosphere to Qwen Image 2512 generations with this specialized LoRA.

1 sources cited Training LoRAs

On this page

  1. What This Model Does
  2. Trigger Words
  3. Recommended Settings
  4. Examples
  5. License
  6. Common Errors and Fixes
  7. FAQ
  8. Sources
  9. Related Models

Film Grain REDMOND LoRA for Qwen Image

TL;DR

What This Model Does

Film Grain REDMOND is a Low-Rank Adaptation (LoRA) trained on Qwen/Qwen-Image-2512 that applies a consistent, high-quality film grain aesthetic to generated images. Unlike simple grain overlays, this LoRA learns the structural characteristics of analog film — the way grain interacts with highlights, shadows, and midtones — producing results that feel like genuine film scans rather than digital noise.

The model was trained with GPU compute sponsored by Redmond.AI and released by ArtificialGuyBR under the permissive Apache 2.0 license.

Trigger Words

Include both trigger words in your prompt for the full effect:

FILMGRAIN, FilmGrainAF

The model card notes these can be used individually, but combining them yields the strongest and most consistent film grain character.

The LoRA integrates cleanly into standard ComfyUI LoRA loading nodes. No special sampling parameters are required — use your usual Qwen Image settings.

Examples

Film grain portrait with cinematic lighting and authentic grain texture

Atmospheric scene demonstrating consistent grain structure across tonal range

Moody cinematic composition with characteristic film texture

All three samples show the model's hallmark: grain that feels structural rather than additive, preserving detail in shadows and highlights while adding the organic texture of analog film.

License

Apache License 2.0 — free for commercial and non-commercial use with attribution.

Common Errors and Fixes

Error | Cause | Fix

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

No visible grain effect | LoRA weight too low or trigger words missing | Increase weight to 0.8–1.0, verify both FILMGRAIN and FilmGrainAF in prompt Overpowering grain / lost detail | LoRA weight too high | Reduce weight to 0.5–0.7 Wrong base model | Using SDXL or FLUX instead of Qwen Image 2512 | Load Qwen/Qwen-Image-2512 as base model

FAQ

Q: Does this work with other base models? A: No. This LoRA is specifically trained on Qwen/Qwen-Image-2512. Using it with other architectures will produce unpredictable results.

Q: Can I use this commercially? A: Yes, Apache 2.0 permits commercial use with attribution.

Q: What's the difference between the two trigger words? A: FILMGRAIN activates the primary grain style; FilmGrainAF appears to be a secondary token that reinforces the effect. The author recommends using both.

Q: Is there a recommended LoRA weight? A: Start at 0.8. The author doesn't specify an exact value, but 0.7–1.0 is typical for style LoRAs on Qwen Image.

Sources