AnalogRedmond: Analog Film Photography LoRA for SDXL

TL;DR:

• Use trigger word AnalogRedmAF to activate the analog film style.
• Add style tags like detailed, minimalist, colorful, or black and white to control the look.
• Run on SDXL 1.0 with multiple generations for best results.

AnalogRedmond is a LoRA for Stable Diffusion XL 1.0 trained on a large dataset of analog photographs. It replicates the look of film-based photography, including natural grain, soft focus, warm color grading, and occasional halation or light leaks. The model works across portraits, architecture, still life, and automotive subjects, consistently applying a cohesive analog aesthetic rather than mimicking a single camera or film stock.

How It Was Trained

The model was fine-tuned on analog photographs and is inspired by Nitro's earlier Dreambooth model for SD 1.5. It is released under the CreativeML OpenRAIL-M license. A v2 version is also available from the same author on Hugging Face.

Trigger Words and Prompting

The primary trigger word is AnalogRedmAF. The author recommends pairing it with descriptive tags to steer results:

• detailed — sharper, more defined images
• minimalist — simpler compositions
• colorful — vibrant analog tones
• black and white — monochrome film look

How to Use

ComfyUI / A1111 / Diffusers

Load the LoRA with weight around 0.7–1.0 on SDXL 1.0. Use AnalogRedmAF in the prompt. The author notes the model is not perfect and sometimes needs more than one generation to create good images, so expect to sample a few outputs.

Recommended Settings

• Base model: SDXL 1.0
• LoRA trigger: AnalogRedmAF
• CFG scale: standard SDXL range
• Sampler: DPM++ 2M Karras or similar
• Steps: 30–40

Examples

Common Errors and Fixes

Error | Cause | Fix | Source

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

No analog style | Missing trigger word | Add AnalogRedmAF to prompt | HF README Overly saturated colors | LoRA weight too high | Lower weight to 0.7–0.8 | Author notes Blurry outputs | Low step count or bad seed | Increase steps, retry generation | Author notes Digital-looking results | Missing style keywords | Add analog, film grain, film photography | Civitai tags

FAQ

Q: Does this work with SD 1.5? A: No. This LoRA is trained for SDXL 1.0. For SD 1.5 analog styles, see Nitro's Dreambooth model, which inspired this release.

Q: Can I use it commercially? A: Yes, under the CreativeML OpenRAIL-M license terms.

Q: Why are my results inconsistent? A: The author notes LoRA generation can vary. Use multiple seeds and prompt variations to find strong outputs.

Sources

• Hugging Face model card — base model, license, and usage notes.
• Civitai model page — community samples, trigger words, and download stats.
• AnalogRedmond v2 on Hugging Face — follow-up version.
• FindGoodPrompt — prompt library for image and language models.
• ArtificialGuyBR portfolio — more models and projects.